1. INTRODUCTION Electrical power systems are designed and operated to meet the continuous variation of power demand.

In power system, minimization of the operation cost is very important. Economic Load Dispatch (ELD) is a method to schedule the power generator outputs with respect to the load demands, and to operate the power system most economically, or in other words, we can say that main objective of economic load dispatch is to allocate the optimal power generation from different units at the lowest cost possible while meeting all system constraints. Over the years, many efforts have been made to solve the ELD problem, incorporating different kinds of constraints or multiple objectives through various mathematical programming and optimization techniques. The conventional methods include Newton-Raphson method, Lambda Iteration method, Base Point and Participation Factor method, Gradient method, etc. However, these classical dispatch algorithms require the incremental cost curves to be monotonically increasing or piece-wise linear. The input/output characteristics of modern units are inherently highly nonlinear (with valve-point effect, rate limits etc) and having multiple local minimum points in the cost function. Their characteristics are approximated to meet the requirements of classical dispatch algorithms leading to suboptimal solutions and therefore, resulting in huge revenue loss over the time. Consideration of highly nonlinear characteristics of the units requires highly robust algorithms to avoid getting stuck at local optima. The classical calculus based techniques fail in solving these types of problems. In this respect, stochastic search algorithms like genetic algorithm (GA), evolutionary strategy (ES), evolutionary programming (EP), particle swarm optimization (PSO) and simulated annealing (SA) may prove to be very efficient in solving highly nonlinear ELD problem without any restrictions on the shape of the cost curves. Although these heuristic methods do not always guarantee the global optimal solution, they generally provide a fast and reasonable solution (sub optimal or near global optimal). The economic and environmental problems in the power generation have
1

received considerable attention. The apparition of the energy crisis and the excessive increase of the consumption have obliged production companies to implant renewable sources. However, this production poses many technical problems for their integration in the electric system. The economic dispatch is a significant function in the modern energy system. It consists in programming correctly the electric production in order to reduce the operational cost. Recently, the wind power attracted much attention like promising renewable energy resources. The problem is formulated as a multi-objective optimization problem. It consists in distributing the active and renewable productions between the power stations of the most economic way, to reduce the emissions of the polluting gases and to maintain the stability of the network after penetration of renewable energy. This work proposes evolutionary optimization techniques namely Genetic Algorithm (GA) and Differential Evolution (DE) to solve ELD in the electric power system, which are generic population, based probabilistic search optimization algorithms and can be applied to real world problem. Both techniques are respectively applied to solve an ELD problem by using the proposed algorithms. And at the last the comparison between the two methods has been presented.

2. ECONOMIC LOAD DISPATCH 2.1. Economic Load Dispatch The Economic Load Dispatch (ELD) can be defined as the process of allocating generation levels to the generating units, so that the system load is supplied entirely and most economically. For an interconnected system, it is necessary to minimize the expenses. The economic load dispatch is used to define the production level of each plant, so that the total cost of generation and transmission loss is minimum for a prescribed schedule of load. The objective of economic load dispatch is to minimize the overall cost of generation. The method of economic load dispatch for generating units at different loads must have total fuel cost at the minimum point. In a typical power system, multiple generators are implemented to provide enough total output to satisfy a given total consumer demand. Each of these generating stations can, and usually does, have a unique costper-hour characteristic for its output operating range. A station has incremental operating costs for fuel and maintenance; and fixed costs associated with the station itself that can be quite considerable in the case of a nuclear power plant, for example. Things get even more complicated when utilities try to account for transmission line losses, and the seasonal changes associated with hydroelectric plants. There are many conventional methods that are used to solve economic load dispatch problem such as Lagrange multiplier method, Lambda iteration method and Newton- Raphson method. In the conventional methods, it is difficult to solve the optimal economic problem if the load is changed. It needs to compute the economic load dispatch each time which uses a long time in each of computation loops. It is a computational process where the total required generation is distributed among the generation units in operation, by minimizing the selected cost criterion, and subjects it to load and operational constraints as well. 2.1.1. Load Dispatching The operation of a modern power system has become very complex. It is necessary to maintain frequency and voltage within limits in addition to ensuring
3

reliability of power supply and for maintaining the frequency and voltage within limits it is essential to match the generation of active and reactive power with the load demand. For ensuring reliability of power system it is necessary to put additional generation capacity into the system in the event of outage of generating equipment at some station. Over and above it is also necessary to ensure the cost of electric supply to the minimum. The total interconnected network is controlled by the load dispatch centre. The load dispatch centre allocates the MW generation to each grid depending upon the prevailing MW demand in that area. Each load dispatch centre controls load and frequency of its own by matching generation in various generating stations with total required MW demand plus MW losses. Therefore, the task of load control centre is to keep the exchange of power between various zones and system frequency at desired values. 2.1.2. Necessity of Generation Scheduling: In a practical power system, the power plants are not located at the same distance from the centre of loads and there fuel costs are different. Also under normal operating, the generation capacity is more than the total load demand and losses. Thus, there are many options for scheduling generation. In an interconnected power system, the objective is to find the real and reactive power scheduling of each power plant in such a way so as to minimize the operating cost. This means that the generators real and reactive powers are allowed to vary within certain limits so as to meet a particular load demand with minimum fuel cost. This is called the Economic load dispatch (ELD) problem. The objective functions, also known as cost functions may present economic cost system security or other objectives. The transmission loss formula can be derived and the economic load dispatch of generation based on the loss formula can also be obtained. The Loss coefficients are known as B-coefficients. A major challenge for all power utilities is not only to satisfy the consumer demand for power, but to do so at minimal cost. Any given power system can be comprised of multiple generating stations having number of generators and the cost
4

of operating these generators does not usually correlate proportionally with their outputs; therefore the challenge for power utilities is to try to balance the total load among generators that are running as efficiently as possible. The economic load dispatch (ELD) problem assumes that the amount of power to be supplied by a given set of units is constants for a given interval of time and attempts to minimize cost of supplying this energy subject to constraints of the generating units. Therefore, it is concerned with the minimization of total. Therefore, the main aim in the economic load dispatch problem is to minimize the total cost of generating real power (production cost) at various stations while satisfying the loads and the losses in the transmission links. 2.2. Generator Operating Cost The total cost of operation includes the fuel cost, cost of labour, supplies and maintenance. Generally, costs of labour, supplies and maintenance are fixed percentages of incoming fuel costs. The power output of fossil plants is increased sequentially by opening a set of valves to its steam turbine at the inlet. The throttling losses are large when a valve is just opened and small when it is fully opened.

Figure 2.1 Simple model of a fossil plant

Figure 2.1 shows the simple model of a fossil plant dispatching purposes. The cost is usually approximated by one or more quadratic segments. The operating cost of the plant has the form shown in Figure 2.2. For dispatching purposes, this cost is usually approximated by one or more quadratic segments. So, the fuel cost curve in the active power generation, takes up a quadratic form, given as:

The fuel cost curve may have a number of discontinuities. The discontinuities occur when the output power is extended by using additional boilers, steam condensers, or other equipment. They may also appear if the cost represents the operation of an entire power station, and hence cost has discontinuities on paralleling of generators. Within the continuity range the incremental fuel cost may be expressed by a number of short line segments or piece-wise linearization. The is the minimum loading limit below which, operating the unit proves to be uneconomical (or may be technically infeasible) and limit.
6

is the maximum output

2.3. Economic Dispatch With Piecewise Linear Cost Functions Many electric utilities prefer to represent their generator cost functions as single or multiple segment linear cost functions. The curves shown in Figure are representative of such functions. Note that were we to attempt to use the lambdaiteration search method on the single segment cost function, we would always land on Pmin or Pmax unless exactly matched the incremental cost at which point the

Figure 2.3 Piece-wise cost curve value of P would be undetermined. To resolve this problem, we perform the dispatch differently. For all units running, we start with all of them at Pmin, and then begin to raise the output of the unit with the lowest incremental cost segment. If this unit hits the right-hand end of a segment, or if it hits Pmax ,we then find the unit with the next lowest incremental cost segment and raise its output. Eventually, we will reach a point where a units output is being raised and the total of all unit outputs equals the total load, or load plus losses. At that point, we assign the last unit being adjusted to have a generation which is partially loaded for one segment. Note, that if there are two units with exactly the same incremental cost, we simply load them equally. To make this procedure very fast, we can create a table giving each segment of each unit its MW contribution (the right-hand end MW minus the left-hand end MW). Then we order this table by ascending order of incremental cost. By searching
7

from the top down in this table we do not have to go and look for the next segment each time a new segment is to be chosen. This is an extremely fast form of economic dispatch. 2.4. The Economic Load Dispatch Problem 2.4.1. Economic Load Dispatch without Losses The simplest economic load dispatch problem is the case when transmission line losses are neglected. Due to this the total demand PD is the sum of all generations. A cost function Fi (Pgi) is assumed to be known for each plant. The problem is to find the real power generation, Pgi for each plant such that the total operating cost F(Pgi) is minimum and the generation remains within the lower generation generation and upper

. Suppose there is a station with NG generators committed and the

active power load demand PD is given, the real power generation Pgi for each generator has to be allocated so as to minimize the total cost. The optimization problem can be therefore be stated as Minimize: Subject to: (i) The energy balance equation (ii) And the inequality constraints Where, is the decision variable, i.e. real power generation PD is the real power demand NG is the number of generation plants
8

( i = 1,2,.........., NG )

is the lower permissible limit of real power generation is the upper permissible limit of real power generation Fi (Pgi) is the operating fuel cost of the i equation F(Pgi)=ai + biPgi + ci Rs/hr
th

plant and is given by the quadratic

The above constrained optimization problem is converted into an unconstrained optimization problem. 2.4.2. Economic Load Dispatch with Losses Transmission losses may be neglected when transmission losses are very small but in a large interconnected network where power is transmitted over long distances, transmission losses are a major factor and affect the optimum dispatch of generation. The economic load dispatch problem considering the transmission power loss PL for the objective function is thus formulated as: Minimize: Minimize: Subject to: (i) The energy balance equation

And the inequality constraints

( i = 1,2,.........., NG )

The general form of the loss formula using B- coefficients is

 where Pgi and Pgj are the real power generations at ith and jth buses respectively. Bij are the loss coefficients or B-coefficients.

10

3. OPTIMIZATION 3.1. Definition Throughout human being history, mankind has faced optimization problems and made great efforts to solve them. Loosely speaking, optimization is the process of finding the best way to use available resources, while at the same time not violating any of the constraints that are imposed. More accurately, we may say that we wish to define a system mathematically, identify its variables and the conditions they must satisfy, define properties of the system, and then seek the state of the system (values of the variables) that gives the most desirable (largest or smallest) properties. This general process is referred to as optimization. 3.2. Optimization Parameters Optimization parameters x are critical for an optimization problem. They affect the value of objective and constraint functions. If there are no optimization parameters, we cannot define the objective and constraint functions. In the investment fund management problem, the optimization parameters are the amounts of money invested in each fund. In experimental data fitting problems, the optimization parameters are the parameters that define the model. In the random design problem, the optimization parameters might include the material index in the materials database, material thickness, and some other parameters. An optimization parameter can be continuous, discrete, or even symbolic. 3.3 Objective Functions An objective function f(x) is what we want to optimize. It is either fi min (x) or fimax(x), depending on the desirable properties of the optimization problem. For instance, in the investment fund management problem, the fund manager wants to maximize the return. In fitting experimental data to a user-defined model, we might minimize the total deviation of observed data from predictions based on the model. In the random design problem, we have to maximize the strength and minimize the distortion and cost.

11

An objective function has at least one global optimum, and may have multiple local optima as shown in the figure. In the remainder of this book, optimum stands for the global optimum of an objective function, f(x*), the objective function value at the optimal solution point x*. For an optimization problem with multiple objective functions, optimal solution points corresponding to different objective functions may be inconsistent.

A one-dimensional continuous multimodal objective function 3.3.1. Objective Functions characteristics features An objective function has some characteristic features. These features are very important for choosing optimization algorithms to solve the optimization problem of interest. 3.3.1.1. Continuity An objective function can be continuous as shown in figure, or discontinuous, as shown

One-dimensional continuous multimodal objective function 3.3.1.2. Decomposability Decomposability is sometimes also referred to as nonlinearity, separability, or epitasis. If each optimization parameter xi is independent of the other parameters x j
12

( j i) in f(x), it is decomposable and is relatively easy to optimize, since the landscape can be decomposed into simpler components. The optimization process of a decomposable objective function can be performed in a sequence of N independent optimization processes, where each parameter is optimized independently. 3.3.1.3. Differentiability A continuous objective function may be non-differentiable or differentiable of order n ( 1).

A continuous non-differentiable function. 3.3.1.4. Modality An objective function f(x) is unimodal if there is some path from every point x to the optimal solution point x* along which it is monotonous. Otherwise, it is multimodal. Figure shows a one-dimensional multimodal objective function and Figure shows a two-dimensional multimodal objective function where the dot is the optimal point x*.

One-dimensional multimodal objective function

Two-dimensional multimodal objective function

13

One-dimensional unimodal objective function

Two-dimensional unimodal objective function

3.3.1.5. Noise Some optimization problems are dynamic or even noisy. In another word, their objective and/or constraint functions are random to a certain extent. Noise is most often used to represent randomness. 3.3.1.6. Scalability An objective function f(x) is not scalable if its number of parameters, or dimension, is fixed. On the other hand, scalable objective functions can be scaled to any dimension. As dimension increases, the search space size also increases exponentially. The difficulty of finding an optimal solution increases accordingly. 3.3.1.7. Symmetry A symmetric objective function f(x) does not change its value if x is replaced by any of its N! Permutations. 3.3.1.8. Uniqueness An optimal solution point x* is unique if there is only one optimal solution. Some objective functions, for example, f(x) = sin(x), x [-2, 2], may have multiple optimal solution points. Objective function values at all optimal solution points are equal. In this regard, no optimal solution is distinguishable from others.

14

4. EVOLUTIONARY PROGRAMMING 4.1. Introduction Evolutionary Algorithms were inspired by Darwins theory of evolution. Natural selection is the foundation of Darwins theory of evolution. The study of evolutionary algorithm began in 1960s. A number of creative researchers independently came up with idea of mimicking the biological evolution mechanism and developed three main stream evolutionary algorithm, namely genetic algorithm, evolutionary programming, evolution strategies. Other evolutionary algorithm includes memetic algorithms, scatter search, self-organizing migrating and tabu search. In recent decades, swarm algorithms including ant colony optimization, bees algorithm, cultural algorithm, Particle swarm optimization, have emerged within the evolutionary computation community. The Swarm algorithm, as their name implies, imitate human or animal behaviors. 4.2. Evolutionary Terminologies For the convenience of the following description, several essential terminologies frequently encountered in evolutionary computation are defined. 4.2.1. Gene The gene is the basic building block of all evolutionary algorithms. There are usually two classes of genes: Real, where a gene is a real number; and alphabetic, where a gene takes a value from an alphabet set. Common alphabet sets are the binary, octal, decimal, and hexadecimal sets. 4.2.2. Chromosome The chromosome is another essential building block of all evolutionary algorithms. It is a symbolic representation of optimization parameters X.

Where gij is the jth gene representing xi and Li is the number of genes in substring for xi .
15

4.2.3. Fitness Fitness is the measure of goodness of a chromosome. It is directly related to objective and constraint function values through a scaling operation. It is not absolutely necessary for some evolutionary algorithms such as differential evolution. 4.2.4. Individual An individual p is an aggregate of a chromosome g, optimization parameter values X, and objective function values f. Additional attributes of an individual may include fitness value, generation, velocity, age gender and even memory. 4.2.5. Population A population is a congregation of individuals. An important characteristics feature of a population is its age, expressed in terms of generation. It is the object all evolutionary algorithm work on. On average, statistically speaking, a later population is better than an earlier population. 4.3. Genetic Algorithm Genetic algorithms, a class of search technique inspired by evolutionary biology, were originated by J.H. Holland. Genetic algorithm involves two stages: Initialization and evolution. Initialization generates an initial population P0. Then P0 evolves P1, P1 evolves into P2... until the termination conditions are fulfilled. When evolving from P n to Pn+1, the three evolutionary operations selection, crossover, and Mutation are executed in sequence.

16

Figure 4.1 Flowchart of genetic algorithm I. Initialization: Initialization generates an initial population P0 which contains Np Individuals p0,i, 1 i Np. II. Selection: The selection operator is believed to be responsible for the convergence of genetic algorithms. It selects good individuals on the basis of their fitness values and produces a temporary population, namely, the mating pool with member individuals. The mating pool is usually, but not necessarily, the same size as Pn-1. This can be achieved by many different schemes, but the most common methods are roulette-wheel, ranking and tournament selection. III. Crossover: The crossover operator has been believed to be the main search tool of Genetic algorithms. It mates individuals qn,i and qn,j in the mating pool by pairs and generates children by crossing over the mated pairs with probability pc, one of the intrinsic control parameters of genetic algorithms. Analogously to human society, the two mating partners are defined as parents, among which individual q n,i is the mother and
17

individual qn,j is the father. Usually, but not necessarily, every pair of parents delivers two children. IV. Mutation: After crossover, some of the genes in the child chromosome are inverted with probability pm, another intrinsic control parameter of genetic algorithms. The mutation operator is included to prevent premature convergence by ensuring the population diversity. V. Elitism Obviously, there is no direct comparison between Pn-1 and Pn. All children are unconditionally admitted to Pn and all parents in Pn-1 are unconditionally abandoned. Consequently, the best individual (in terms of fitness) in Pn may be worse than that in Pn-1. In this case, elitism will be implemented by replacing the worst individual in Pn with the best individual in Pn-1. 4.4. Differential Evolution Differential evolution evolved from the Genetic algorithm with slight modification in search technique. It is more competitive with search space. The brief study of DE is continued in next chapter.

18

5. DIFFERENTIAL EVOLUTION 5.1. Introduction Differential evolution was proposed by KennethV. Price and R. Storn in 1995 while trying to solve the Chebyshev polynomial fitting problem. It stems from the genetic annealing algorithm which was also developed by Kenneth V. Price. Troubled by the slow convergence of the genetic annealing algorithm and the difficulties it faced in determining effective control parameters, Price modified the genetic annealing algorithm by using real code with arithmetic operations instead of binary code with Boolean operations. During this process, he discussed the differential mutation operator which was later shown to be the key to the success of differential evolution. Thus, differential evolution came to being. Differential evolution has been the subject of intensive performance evaluation since its appearance. Apart from the performance evaluation conducted by its originators, to this authors knowledge, the first performance evaluation of differential evolution by other researchers was carried out in 1995 on a test bed of 15 test functions. Differential evolution soon earned its reputation by ranking third among all entries in the first international contest on evolutionary optimization held in Nagoya, Japan, in May 1996 and best among all qualified entries in the second international contest on evolutionary optimization held in Indianapolis, Indiana, USA, in April 1997 (the actual contest was cancelled due to a lack of valid entries). In late 1997, Storn established a website

(http://www.icsi.berkeley.edu/_storn/code.html) to publicize differential evolution. Since then we have witnessed explosive growth in differential evolution research.

5.2. Differential Evolution Optimization Process An optimization task consisting of D parameters can be presented by a Ddimensional vector. In DE, a population of NP solution vectors is randomly created
19

at the start. This population is successfully improved over G generations by applying mutation, crossover and selection operators, to reach an optimal solution. The main steps of the DE algorithm are given below: Initialization Evaluation Repeat Mutation Crossover Evaluation Selection Until (Termination criteria are met) I. Initialization Initialization generates an initial population P0 which contains Np Individuals p0,i, 1 i Np. II. Mutation The mutation operator creates mutant vectors by perturbing a randomly selected vector xa with the difference of two other randomly selected vectors xb and xc,

where xa, xb and xc are randomly chosen vectors among the NP population, and a b c. xa, xb and xc are selected a new for each parent vector. The scaling constant F is an algorithm control parameter used to adjust the perturbation size in the mutation operator and improve algorithm convergence.

20

III.

Crossover The crossover operation generates trial vectors xi by mixing the parameters

of the mutant vectors xi with the target vectors xi according to a selected probability distribution,

where i=1, , NP and j=1,, D; q is a randomly chosen index {1,,Np} that guarantees that the trial vector gets at least one parameter from the mutant vector; j s a uniformly distributed random number within [0 , 1] generated anew for each value of j. The crossover constant CR is an algorithm parameter that controls the diversity of the population and aids the algorithm to escape from local minima. xj,i(G) and xj,i(G) are the jth parameter of the ith target vector, mutant vector, and trial vector at generation G, respectively. IV. Selection

The selection operator forms the population by choosing between the trial vectors and their predecessors (target vectors) those individuals that present a better fitness or are more optimal according to

This optimization process is repeated for several generations, allowing individuals to improve their fitness as they explore the solution space in search of optimal values. DE has three essential control parameters: the scaling factor (F), the crossover constant (CR) and the population size (NP). The scaling factor is a value in the range [0, 2] that controls the amount of perturbation in the mutation process. The crossover constant is a value in the range [0, 1] that controls the diversity of the population.
21

The population size determines the number of individuals in the population and provides the algorithm enough diversity to search the solution space.

Figure 5.1 Flowchart for Differential Evolution 5.3. Differential Evolution in Electrical And Electronic Engineering Advances in electrical and electronic engineering stimulate evolutionary computation. Conversely, electrical and electronic engineering benefits greatly from differential evolution. 5.3.1. Electrical Engineering Applications of differential evolution in electrical engineering focus on motor modeling and design,

22

5.3.2. Control Theory and Engineering Control problems occur in almost all engineering fields such as aeronautics, agriculture, chemical engineering, civil engineering, environmental engineering, food engineering, manufacturing industry, mass rapid transit system, material engineering, and power engineering. Differential evolution has been successfully applied to solve numerous control-related application problems. 5.3.2.1. System Modeling System modeling is fundamental in the simulation and controller design for complex dynamic systems. Different systems have been modeled by using differential evolution. 5.3.2.2. Controller Design Differential evolution plays a significant role in controller design. Controllers designed by using differential evolution are summarized in Table.

23

5.3.3. Robotics The history of implementing differential evolution in robotics is only slightly shorter than that of differential evolution itself. Problems in robotics solved by using differential evolution are summarized in Table.

24

6. ECONOMIC LOAD DISPATCH WITH LOSSES 6.1. Dispatch Model The objective of ED is to minimize the total generation cost of a power system over appropriate period (1 h generally) while satisfying various constraints. It is the most important sub-problem of power system operation for economic issue. Due to the highly nonlinearity of power system and generators, ED belongs to the class of nonlinear programming optimization with equality and inequality constraints. Practically, while the scheduled combination units at each specific period of operation are listed, the ED planning must perform the optimal generation dispatch among the operating units to satisfy the system load demand, practical operation constraints of generators that include the ramp rate limit, maximum and minimum limits and prohibited operating zones. Generally, the generation cost function is usually expressed as quadratic polynomial below Fi(Pgi)=aiPgi2+biPgi+ci $/hr where Pi, the output power of generator i for dispatched hour Fi(Pi), the fuel cost function of generator i; aibici, the coefficients of the fuel cost function for generator i. Although this quadratic form is used in this paper, any other programmable cost functions including piecewise continue or non-differentialable can also be easily applied on this paper due to the flexibility of DE. 6.2. Operational limitation and constraints For convenience in dealing with ED problems, the generation output is assumed adjusted continuously, smoothly, instantaneously, and boundlessly. However, the available operating range of generators is restricted by their mechanical and electrical characteristic, such as ramp rate limits, prohibitive zones, and bounded power output. Hence, these practical constraints of generator operation

25

must be taken into account to achieve true economic operation. The operational limitations and constraints considered in this paper are listed and stated below. 6.2.1. Power balance The total power output of generators must be always equal to the sum of the power demands and the network losses. Since power plants are spread out geographically, the transmission network losses must be taken into account to achieve true economic dispatch. Network loss is a function of unit generation. To calculate network losses, the B coefficients method is commonly used by the power utility industry. It can be expressed as a quadratic function.

where Bij, the B coefficients for losses calculation. 6.2.2. Ramp rate limits The generation output of a practical generator cannot be adjusted instantaneously without limits. The operating range of all on-line units is restricted by their ramp rate limits during each dispatch period. Therefore, the subsequent dispatch output of a generator should be limited between the constraints of up and down ramp rates. Pipre-DownRi Pi Pipre +UpRi where Pipre, the output power of generator i before dispatched hour; UpRi, the up ramp limit of generator i; DownRi the down ramp limit of generator i. 6.2.3. Prohibited operating zone and generation limits Each generator has its generation capacity, which cannot be exceeded at any time. Moreover, a typical thermal unit may have a steam valve in operation, or a vibration
26

in a shaft bearing, which may result in interference and discontinue inputoutput performance-curve sections, called prohibited zones. Therefore, in practical operation, adjusting the generation output of a unit must avoid all capacity limits and unit operations in prohibited zones. Pimin Pi Pilower j=2,3,, PZi

where Pimax, the maximum power output limit of generator i; Pimin , the minimum power output limit of generator i; for generator i; , the lower bound of the jth prohibited zone

, the upper bound of the jth prohibited zone for generator i;

PZi, the number of prohibited zones for generator i. The above problem is solved with reference to data in Appendix 2 (A.2). The Problem is solved using MATLAB with GODLIKE tool. 6.3. Results and Analysis This system contains 6 thermal units, 26 buses and 46 transmission lines. The load demand is decision hour is set as 1263 MW. As shown in below table DE has better results than the GA in terms of cost and line losses.

P1 (MW) P2 (MW) P3 (MW) P4(MW) P5 (MW) P6 (MW) Line Loss (MW) Total output Load supplied Cost ($/h)

GA 474.9448 157.2653 246.0471 121.8457 181.5663 94.1546 12.8222 1275.8 1263 15458

DE 446.7958 172.9763 263.7102 138.8322 165.7803 87.2215 12.3169 1275.3 1263 15441

27

7. ECONOMIC LOAD DISPATCH WITH RENEWABLE ENERGY In this section, the ELD problem of a power system that includes wind generation will be formulated. The paper will use a number of case simulations to investigate the impact of renewable energy (wind power) on the load dispatching calculation. This paper reports an initial investigation of topic; therefore some simplifications have been made. 7.1. Dispatch Model The dispatch model in the simulation uses centralized dispatch in a deregulated power system. Generators in the systems are thermal, wind and diesel. Fuel for diesel generators is assumed to be bio-diesel. Constraints included in the calculation are the maximum and minimum values of generator output, the ramp rate of the generators, and reserve requirements. For simplification, transmission losses are neglected. The economic dispatch process aims at cost Minimization subject to these constraints. The solution flowchart is shown in figure. This paper will use as a basis the ELD models suggested by Doherty. Initially, a no wind scenario is used as a base case and then a forecasted scenario is also applied. In the forecasted scenario, two approaches to deal with wind generators will be applied as follows 1. Negative load approach This is the simpler of the two approaches, where the wind forecast is treated as a negative load. Therefore, load demand is reduced by the forecast wind power producing a new load demand. This new load demand is then used in the ELD process. 2. Inclusive approach In this approach, wind turbines are included in the calculation. In order to maximize the wind output for the purpose of reducing emission, wind output should be used as much as possible.

28

Figure 7.1 Flow chart for ELD problem The objective function of ELD is formulated as follows: Min( Cpi Pi + CriRi), i=1,2,,N Subject to constraints (PLOAD + PLOSS ) Pi = 0 Pimin Pi Pimax Ri Rmin DRi.t Pit+1 Pit, for Pit+1 < Pit URi.t Pit+1 Pit, for Pit+1 > Pit Where: i, N generator number and total number of generator. t, t, T time interval, duration of time and maximum time horizon respectively. Cpi, Cri the price of power output and reserve of generation I respectively. Pi, Ri, - power output and reserve of generation I respectively. Pimin Pimax minimum and maximum output of generator i respectively.
29

 Rmin minimum reserve requirements. DRi, URi, - down and up ramp limit of generator. 7.2. Generator Price Model 7.2.1. Thermal Generators It is assumed that the prices are cost-reflective and that each generator has a four-step incremental price curve. The curves of operating and incremental prices are shown in figure For all units running, we start with all of them at Pmin then begin to raise the output of the unit with the lowest incremental cost segment. If this unit hits the righthand end of a segment, or if it hits Pmax, we then find the unit with the next lowest incremental cost segment and raise its output. Eventually, we will reach a point where a units output is being raised and the total of all unit outputs

Figure 7.2 Piece-wise cost curve equals the total load, or load plus losses. At that point, we assign the last unit being adjusted to have a generation which is partially loaded for one segment. Note, that if there are two units with exactly the same incremental cost, we simply load them equally.

30

The operating limit and ramp limit for each generator is as shown in Table

7.2.2. Wind generator The wind source consists of a number of wind turbines. For simplicity, all wind turbines are considered to be identical. The total output of the wind (Pwt) generators is equal to the product of the number of wind turbines (Nw) and the power output of the individual wind turbine (Pw), which is itself a function of the wind speed (w). Therefore, the total wind generator output is: Pwt = Nw . Pw(w) The wind cost has a high capital cost and low operational cost and its price is assumed as follows: Cw = 3082 + 0.01 Pw 7.2.3. Bio-Diesel Generators Bio-diesel generators are assumed to be used only as back up when the reserve requirement provided by committed generators is not adequate. The price characteristic (Cd) is assumed as a simple linear function of generator output (Pd). Cd = 80 Pd Total capacity of all bio diesel generators is 150 MW, which is estimated being sufficient of compensating wind power variability. 7.3 Reserve Requirement Due to the uncertainty of wind output, reserve should be increased to take this into account. The additional reserve will also increase with the increase of the time horizon. For simplicity, the normal reserves (Rn) are based on load demand, and
31

additional reserves (Ra) due to wind power are based on wind output and time horizon. Both are determined as follows: Rn = 25% Pload
20% of wind output, for t=1 22.5% of wind output, for t=2 24% of wind output, for t=3

Ra =

25% of wind output, for t=4 26% of wind output, for t=5 27% of wind output, for t=6,7,..,12

The above equation is derived from the graph of typical standard deviation by Doherty and OMalley. The assumed reserve cost for each generator is shown in Table. Bio-diesel generators are not included in the reserve cost, as these generators are considered as back up.

7.4. Load Demand and Wind Power Forecast In the simulation, load and wind power forecast are carried out for a 12 hour period including peak hours. The load forecast is derived from a typical peak load pattern in Indonesia. Wind power is derived from the wind speed forecast using the wind generators characteristic function of Figure. In the absence of typical wind profiles, in order to test the approach of profile based on randomized sinusoidal has been used for representing fluctuation and uncertainty of wind turbine. The load demand and wind power forecast used in the simulation is shown in Figure.
Time (Hr) Demand (Mw) 1 2 3 4 5 6 7 8 9 10 11 12

575

725

675

750

760

850

1180

1210

1125

1050

940

900

32

Time (Hr)

10

11

12

Wind power (Mw)

68

40

60

118

169

150

70

41

48

85

159

142

Demand curve
1400 1200

Load (Mw)

1000 800 600 400 200 0 1 2 3 4 5 6 7 8 9 10 11 12 Demand curve

time (hrs) Figure 7.3 Demand Curve for 12 hrs

wind power
180

wind power (Mw)

160 140 120 100 80 60 40 20 0 1 2 3 4 5 6 7 8 9 10 11 12 wind power

time (hrs) Figure 7.4 Forecasted Wind power for 12 hrs

33

7.5. System Operation Scenario In order to satisfy demand and other constraints, all thermal generators are considered to be committed. All wind output should be fully utilized based on wind speed forecasting in order to minimize fuel consumption and reduce emissions. Bio-diesel generators are committed when the reserve are not fulfilled by thermal reserves. Reserve requirements are based on the next hour availability, therefore the maximum reserve provided by thermal generators is limited by the up ramp limit (UR) of each generator. All bio-diesel generators are assumed to be able to reach a maximum output within a few minutes when these are committed. This model will use as a basis the ELD models suggested by Doherty. Initially, a no wind scenario is used as a base case and then a forecasted scenario is also applied. In the forecasted scenario, two approaches to deal with wind generators will be applied as follows: 1. Negative load Approach 2. Inclusive Approach 7.5.1. No wind Case In no wind case, ELD is completely based on the Thermal generators, reserve and Bio-diesel as back-up. Bio-diesel is used whenever adequate reserve is not available. The minimum reserve considered for this model is 25% of full load. Whenever the reverse is less than 25% of full load, then back-up plant is added to compensate the reserve. 7.5.2. Negative Load Approach In negative load case, wind is treated as the negative load. Load demand is reduced by forecasted wind power producing a new load demand. This new load demand is used for ELD Process. In this approach, extra reserve also taken in case of forecasted wind fails. This reserve is considered with reference to Doherty and OMalley.

34

7.5.3. Inclusive Approach In Inclusive approach, wind power generators are included in ELD problem. In order to maximize the wind output for purpose of reducing the emission, wind power is completely used as much as possible. Since wind is used as the part of power generation, the maintenance and operational is taken as default, even wind power is doesnt share with load demand. 7.6. Algorithm for ELD Using DE The step-wise procedure is outlined below: 1. Read data, namely cost coefficients, no. of iterations, length of string, population size, probability of crossover and mutations, power demand and Pmin and Pmax . 2. Create the initial population randomly in the binary form. 3. Decode the string, or obtain the decimal integer from the binary string. 4. Calculate the power generated from the decoded population. 5. Check Pij > pimax, then set pij = pimax Pij < pimin, then set pij = pimin 6. Find the fitness or cost function. 7. Find population with maximum fitness and average fitness of the population. 8. Perform the Reproduction Process, which includes the following steps: 8 (a). Set selection rate and number of mating in a pool. 8 (b). Define total fitness as sum of values obtained by using above step for all chromosomes which are selected. 8 (c). Select percentage of each chromosome which is equal to the ratio of its fitness value to the total fitness value, i.e. find probability which can be written as: Probability = fitness / Fitnesss. 8 (d) Calculate cumulative sum (CS) to normalize the values between 0.0 and 1.0. 9. Perform Crossover Process:
35

9 (a) Choose a pair of random numbers between 0 and 1 to select one mother and one father chromosome, so as to produce new offspring. 9 (b) Pairing the chromosomes from different location, for different locations, crossover point has to be selected which can be selected randomly. Generate offspring by applying crossover. 10. Perform mutation by randomly selecting the mutation points from the total number of bits in the population matrix. 11. Update the population. 12. If the number of iterations reaches the maximum, then go to step 13. Otherwise, go to step 6. 13. The fitness that generates the minimum total generation cost is the solution of the problem.

36

8. SIMULATIONS AND RESULTS 8.1. No wind case: 8.1.1. Result: GA

NO WIND CASE - GENETIC ALGORITHM THERMAL GEN 1 THERMAL GEN 4 GA TOTAL COST 13260.54595 19231.58467 17723.8481 20141.04717 20454.60213 22909.2535 40987.86609 42314.02221 36337.55332 32979.13637 27303.79594 26041.08698 THERMAL COST RESERVE COST THERMAL GEN 2 THERMAL GEN 3 BIO-DIESEL GENERATION BIO-DIESEL COST In Rs 12933.41376 18990.21405 17121.90155 19766.49851 19967.92535 22550.34141 35275.38051 41093.53715 35711.04254 32352.85208 26644.04402 25476.40989 327.132191 239.759825 601.946551 374.548658 486.676779 358.912091 104.352444 410.64895 626.510777 626.284298 659.751913 564.677094 1.610791 5608.133133 809.836109 RESERVE % 74.9388 127.7425 74.9879 124.8717 144.766 165.3897 168.8967 224.8572 211.1214 174.4977 174.9885 169.3833 Not applicable 0.0201 Not applicable Not applicable Not applicable Not applicable 70.1017 1.01E+01 Not applicable Not applicable Not applicable Not applicable 33.913078 23.451121 54.811996 32.666337 40.789599 27.058923 5.093152 19.011835 35.098013 37.619485 45.744866 39.999858 DEMAND MW 575 725 675 750 760 850 1180 1210 1125 1050 940 900 199.8466 199.822 197.58 199.3977 199.6133 199.8472 268.8773 331.9163 298.3767 285.1457 278.8245 276.2425 68.1134 122.2292 102.4143 115.1854 112.1032 124.2036 213.0728 166.9633 121.828 109.3541 123.2889 124.1514

Hr 1 2 3 4 5 6 7 8 9 10 11 12

TIME

MW 232.101 275.1856 300.0161 310.5458 303.5173 364.5583 459.0528 476.1197 493.6776 481.0018 362.8956 330.2216

37

8.1.2 Result: DE
NO WIND CASE - DIFFERENTIAL EVOLUTION ALGORTIHM BIO-DIESEL GENERATION BIO-DIESEL COST THERMAL GEN 1 THERMAL GEN 2 THERMAL GEN 3 THERMAL GEN 4 THERMAL COST DE TOTAL COST 13248.71994 18727.46693 16702.46685 18867.46219 19814.94926 22874.95572 39427.2596 39789.96768 35844.93836 31629.9429 27307.45608 25739.95723 RESERVE RESERVE COST In Rs 12924.9681 18494.94765 16099.96816 18499.96136 19329.94882 22499.95665 37809.94546 39349.96356 35174.94141 31024.94211 26599.95725 25149.95723 323.75184 232.501308 602.498689 367.500826 485.000439 374.999073 17.500206 440.003187 669.996943 605.000797 707.498829 590.000001 0.017971 1599.813937 0.000932 DEMAND MW 575 725 675 750 760 850 1180 1210 1125 1050 940 900 200 199.9999 200 200 200 200 270 299.9995 300 300 300 300 75 139.9986 75 124.9987 134.9984 125 214.9999 224.9941 175 199.9983 125 75

Hr 1 2 3 4 5 6 7 8 9 10 11 12

TIME

MW 224.9988 249.9998 224.9988 250 250 349.9986 449.9984 460.0077 474.9983 375 339.9986 349.9986 75 134.9999 175 175 175 175 225 224.9973 175 175 175 175 Not applicable 0.0002 Not applicable Not applicable Not applicable Not applicable 19.9977 1.17E-05 Not applicable Not applicable Not applicable Not applicable

% 33.913256 23.448338 54.814748 32.666681 40.789502 27.058812 0.847467 22.314041 36.000019 37.619046 45.744656 40

38

8.1.3 Discussion As a base case, the results of the calculation using GA and DE for the system without wind power are shown in 8.1.1 and 8.1.2.
Hr 1 2 3 4 5 6 7 8 9 10 11 12 GA COST ($) 13260.54595 19231.58467 17723.8481 20141.04717 20454.60213 22909.2535 40987.86609 42314.02221 36337.55332 32979.13637 27303.79594 26041.08698 DE COST ($) 13248.71994 18727.46693 16702.46685 18867.46219 19814.94926 22874.95572 39427.2596 39789.96768 35844.93836 31629.9429 27307.45608 25739.95723

45000 40000 35000 30000

Cost ($)

25000 20000 15000 10000 5000 0 1 2 3 4 5 6 Hr 7 8 9 10 11 12 GA COST ($) DE COST ($)

Figure 8.1 No wind case Results From above graph it is clear that DE has the best Economic value than that of GA.

39

8.2. Negative Load: 8.2.1 Result: GA

NEGATIVE LOAD - GENETIC ALGORITHM THERMAL GEN 1 THERMAL GEN 2 TOTAL DEMAND THERMAL COST RESERVE COST THERMAL GEN 3 THERMAL GEN 4 ACTUAL LOAD WIND POWER BIO-DIESEL GENERATION TOTAL COST 11606.6047 17144.78844 15223.90387 15353.54876 14383.41316 18316.82033 39085.96015 39669.25266 33535.98486 29212.78892 21447.7466 21404.60111 BIO-DIESEL COST In (Rs) 11178.73343 16839.00823 14591.65213 14874.79931 13815.19088 17833.53045 29360.41807 36982.21375 32929.21572 28420.06368 20636.54893 20079.25504 427.871273 199.743632 632.251737 475.446954 568.222285 313.842349 52.281787 286.598766 606.768088 710.702325 811.197665 509.228481 106.03658 3.302501 169.447525 9673.260292 2400.440153 0.001057 82.022913 816.117591 RESERVE % 73.9987 133.7217 131.6544 69.881 68.4816 123.9414 174.7285 199.6844 213.3038 215.6189 174.5882 163.948 Not applicable 1.33E+00 Not applicable 4.13E-02 Not applicable 2.12E+00 120.9158 30.0055 1.32E-05 1.03E+00 Not applicable 10.2015 51.873833 20.924866 63.197716 47.948868 61.076581 30.43871 2.598853 14.550024 35.468331 44.872642 64.401632 46.597039

Hr 1 2 3 4 5 6 7 8 9 10 11 12 575 725 675 750 760 850 1180 1210 1125 1050 940 900

TIME

MW 68 40 60 118 169 150 70 41 48 85 159 142 507 685 615 632 591 700 1110 1169 1077 965 781 758 198.8588 189.4306 198.8867 199.3446 199.3458 199.9222 268.7572 297.6759 299.8284 295.6771 290.0108 231.3128 70.1179

MW 164.0243 249.1652 211.1262 249.8437 213.1352 229.6533 325.7643 418.2873 362.7331 329.1987 241.5351 239.8859 111.3469 73.3311 112.8923 110.0364 144.4111 219.8308 223.3433 201.1317 123.4813 74.8652 112.647

40

8.2.2 Result: DE
NEGATIVE LOAD - DIFFERENTIAL EVOLUTION ALGORITHM TOTAL DEMAND THERMAL COST RESERVE COST THERMAL GEN 1 THERMAL GEN 2 THERMAL GEN 3 THERMAL GEN 4 ACTUAL LOAD WIND POWER BIO-DIESEL GENERATION TOTAL COST 11582.71901 16985.46314 14939.96154 15273.96775 13923.46796 17516.46126 38444.71543 38158.6455 33486.94622 28061.46126 21414.46084 19733.95876 BIO-DIESEL COST In (Rs) 11156.96705 16774.96239 14309.96145 14802.96824 13340.96822 17209.96035 29609.93982 37914.64606 32819.9441 27349.96157 20617.96218 19263.95649 425.751957 210.500609 630.000092 470.999432 582.499743 306.500878 35.000695 243.998931 667.000842 711.499508 796.498664 470.001278 0.000138 0.000085 0.000036 8799.774916 0.000508 0.001277 0.000183 0.00099 RESERVE % 75 135 75 75 75 135 175 224.9999 224.9997 174.9996 175 174.9999 Not applicable 1.73E-06 Not applicable 1.06E-06 Not applicable 4.50E-07 109.972 0.00000635 1.60E-05 2.29E-06 Not applicable 0.000012375 51.874019 20.729934 63.414636 47.94301 61.082922 30.14288 1.801834 12.917008 38.254417 44.766814 64.532645 45.250669

Hr 1 2 3 4 5 6 7 8 9 10 11 12 575 725 675 750 760 850 1180 1210 1125 1050 940 900

TIME

MW 68 40 60 118 169 150 70 41 48 85 159 142 507 685 615 632 591 700 1110 1169 1077 965 781 758 200 200 200 199.9999 200 200 270 299.9999 300 299.9998 280.9986 200 75 99.9988 89.9987 106.9992 75 114.9987 204.9984 225 176.9987 124.9988 75 132.999

MW 156.9988 250 250 249.9997 240.9988 250 349.9999 418.9986 375 365.0003 250 249.9996

41

8.2.3 Discussion In this approach, wind power forecast is treated as negative load, which reduces the load demand. Using GA and DE method, results of this approach are given in figure below
Hr 1 2 3 4 5 6 7 8 9 10 11 12
45000 40000 35000 30000 Cost (Rs) 25000 20000 15000 10000 5000 0 1 2 3 4 5 6 7 8 9 10 11 12 Hour GA COST (Rs) DE COST (Rs)

GA COST (Rs) 11606.6047 17144.78844 15223.90387 15353.54876 14383.41316 18316.82033 39085.96015 39669.25266 33535.98486 29212.78892 21447.7466 21404.60111

DE COST (Rs) 11582.71901 16985.46314 14939.96154 15273.96775 13923.46796 17516.46126 38444.71543 38158.6455 33486.94622 28061.46126 21414.46084 19733.95876

Figure 8.2 Negative Load case Results

Negative Load approach reduces the system cost due neglecting wind cost, except in hour 7 when bio-diesel engine is added more for compensating high load ramp and additional reserve.

42

8.3 Inclusive Approach : 8.3.1 Result: GA

INCLUSIVE APPROACH - GENETIC ALGORITHM WIND GENERATION BIO-DIESEL POWER COST WIND POWER COST THERMAL GEN 3 THERMAL GEN 4 THERMAL GEN 1 THERMAL GEN 2 TOTAL DEMAND THERMAL COST RESERVE COST

WIND POWER

BIO-DIESEL GENERATION

Hr 1 2 3 4 5 6 7 8 9 10 11 12

MW 575 725 675 750 760 850 1180 1210 1125 1050 940 900 68 40 60 118 169 150 70 41 48 85 159 142 180.5196 244.1431 189.5176 188.5612 197.7316 199.8855 269.6782 299.8526 283.982 295.8296 243.4216 189.2247

MW 53.1274 124.0807 72.0661 68.7737 71.3886 100.173 187.2657 223.4784 174.4525 117.4851 68.3062 62.5507 205.8176 248.4683 277.3608 245.8653 217.3131 249.2864 338.5759 372.7282 404.859 411.8782 311.1582 373.9513 68.9223 68.2995 74.462 128.6852 104.1342 152.6782 206.6255 266.2434 213.0225 143.914 165.3194 124.2266 0.0214 0.0177 2.0116 0.3214 0.6492 0.0344 110.1123 10.2627 1.0738 1.3638 0.0268 10.1143 66.5916 39.9884 59.5855 117.7838 168.7867 147.9379 67.7371 37.4536 47.6234 79.5317 151.7789 139.9614

% 45.497926 19.778656 58.012607 40.202806 47.541855 24.53463 1.684562 13.210148 36.087677 40.686597 53.287379 39.80579 11392.59177 17987.31643 15104.22189 15100.1973 13992.84094 17201.12813 30667.95493 38049.19434 34482.19396 29703.88189 21991.19253 19672.46749 426.385139 236.583691 609.53154 464.893785 576.392669 312.365244 32.679941 246.840544 663.862428 717.498537 769.386093 548.430663

In Rs 1.712423 1.415099 160.92747 25.709767 51.933709 2.750372 8808.984224 821.013181 85.902991 109.102208 2.147299 809.146458 3082.665916 3082.399884 3082.595855 3083.177838 3083.687867 3083.479379 3082.677371 3082.374536 3082.476234 3082.795317 3083.517789 3083.399614 14903.35525 21307.71511 18957.27676 18673.97869 17704.85519 20599.72312 42592.29646 42199.4226 38314.43562 33613.27795 25846.24371 24113.44422

43

TOTAL COST

RESERVE

TIME

8.3.2 Result: DE
INCLUSIVE APPROACH - DIFFERENTIAL EVOLUTION ALGORITHM THERMAL GEN 3 THERMAL GEN 4 THERMAL GEN 1 THERMAL GEN 2 TOTAL DEMAND THERMAL COST RESERVE COST WIND POWER WIND POWER COST BIO-DIESEL GENERATION WIND GENERATION BIO-DIESEL POWER COST TOTAL COST 14665.42528 20068.465 18235.24771 18544.9089 17406.30381 20545.56563 42036.18724 41261.59538 36569.59432 31144.43931 24498.12897 23585.90839

Hr 1 2 3 4 5 6 7 8 9 10 11 12

MW 575 725 675 750 760 850 1180 1210 1125 1050 940 900 68 40 60 118 169 150 70 41 48 85 159 142 199.9996 199.994 199.9935 199.9966 199.9981 199.998 269.994 299.9817 299.9712 299.9992 281.0124 258.0827 74.9966 100.1771 74.9799 74.8938 91.0146 100.0108 190.0063 224.7459 177.0646 124.9903 74.991 74.9537

MW 157.0049 249.8705 165.0296 182.1155 124.9962 224.9904 324.9794 419.273 374.9996 365.0134 249.9991 249.9806 74.997 134.9593 174.9956 174.9955 174.9944 174.9993 224.9983 224.998 224.9621 174.9973 174.9963 174.9824 0.0001 0.0003 0.0003 0.0003 0.0002 0.0006 100.0197 4.69E-05 2.49E-05 0.0002 0.0003 0.0027 68 39.996 59.9998 117.9976 168.9954 149.9994 69.9979 40.9993 47.9993 84.9986 158.9999 141.9972

% 45.739468 19.585809 57.778111 40.399654 47.499721 24.824103 0.849199 13.304078 36.622328 41.142593 53.617162 38.111063 11156.9854 16775.63256 14540.02933 14982.16494 13764.16436 17125.00713 30934.38172 37915.08845 32820.26615 27350.03168 20618.00902 19974.21213 425.755171 210.412021 612.59267 479.540716 558.437458 337.007921 17.531307 264.093177 666.846189 711.538124 796.50847 528.056642

RESERVE

TIME

In Rs 0.004708 0.020463 0.025709 0.023271 0.012037 0.050578 8001.574238 0.003755 0.001992 0.019514 0.021482 0.219644 3082.68 3082.39996 3082.599998 3083.179976 3083.689954 3083.499994 3082.699979 3082.409993 3082.479993 3082.849986 3083.589999 3083.419972

44

8.3.3 Discussion Wind power forecast is included in the calculation and results using GA and DE method are presented in table below.
Hour GA Cost (Rs) DE cost (Rs)

1 2 3 4 5 6 7 8 9 10 11 12

14903.35525 21307.71511 18957.27676 18673.97869 17704.85519 20599.72312 42592.29646 42199.4226 38314.43562 33613.27795 25846.24371 24113.44422

14665.42528 20068.465 18235.24771 18544.9089 17406.30381 20545.56563 42036.18724 41261.59538 36569.59432 31144.43931 24498.12897 23585.90839

45000 40000 35000 30000 Cost (Rs) 25000 20000 15000 10000 5000 0 1 2 3 4 5 6 7 8 9 10 11 12 Hour GA Cost (Rs) DE cost (Rs)

Figure 8.3 Inclusive Approach case Results

In this approach the cost reduces when wind power output is high, otherwise, the cost will be higher when wind power output is relatively low.

45

8.4 Comparison of all 3 methods:

DE Cost (Rs)

Hour
1 2 3 4 5 6 7 8 9 10 11 12

No wind case
13248.71994 18727.46693 16702.46685 18867.46219 19814.94926 22874.95572 39427.2596 39789.96768 35844.93836 31629.9429 27307.45608 25739.95723

Negative Load
11582.71901 16985.46314 14939.96154 15273.96775 13923.46796 17516.46126 38444.71543 38158.6455 33486.94622 28061.46126 21414.46084 19733.95876

Inclusive Method
14665.42528 20068.465 18235.24771 18544.9089 17406.30381 20545.56563 42036.18724 41261.59538 36569.59432 31144.43931 24498.12897 23585.90839

45000 40000 35000 30000

Cost (RS)

25000 20000 15000 10000 5000 0 1 2 3 4 5 6 7 8 9 10 11 12 Hour

No wind case Negative Load Inclusive Method

Figure 8.4. Comparison of results of all methods

46

8.5. GUI for ELD Problem 8.5.1. No Wind Case

47

8.5.2. Negative Load

48

8.5.3. Inclusive Approach

49

9. CONCLUSIONS As there will be an increasing use of renewable energy in the future, it is important to carry out the study of its impact on the operation of power systems and in particular ELD problem. This paper reports an initial study focusing on the use of the DE method for the detailed investigation of this problem and Comparison of results between GA and DE. Several basic models dealing with the problem of centrally dispatched wind generators as reported in the literature have been combined and implemented for a simple case. It has been demonstrated that the inclusion of renewable energy in the power system in this case has an impacts on cost, the variability of load seen by other plant, and reserve calculations. The results presented provide an initial verification of the hypothesis that to compensate for the variability of renewable energy, the ramp rate capability of the plant plays an important role and that fast ramp rate plant such as bio-diesel engines can be used to compensate for this variability. Because wind power cost is assumed to be zero for negative load approach, the result does not reflect to the real cost. Therefore, the actual wind power costs should be included in the negative load approach.

50

APPENDIX

51

A.1.1.Economic Load Dispatch with Losses:

%Evolutionary Programming Function for Economic Load dispatch function val=check_final1(x) pd=1263; pin=[440 170 200 150 190 110]; f1=240 f2=200 f3=220 f4=200 f5=220 f6=190 + + + + + + 07.0*x(1) 10.0*x(2) 08.5*x(3) 11.0*x(4) 10.5*x(5) 12.0*x(6) + + + + + + 0.0070*x(1)^2; 0.0095*x(2)^2; 0.0090*x(3)^2; 0.0090*x(4)^2; 0.0080*x(5)^2; 0.0075*x(6)^2;

%loss calculation b1=(10^-2)*[0.0017 0.0012 0.0007 -0.0001 -0.0005 -0.0002; 0.0012 0.0014 0.0009 0.0001 -0.0006 -0.0001; 0.0007 0.0009 0.0031 0.0000 -0.0010 -0.0006; -0.0001 0.0001 0.0000 0.0024 -0.0006 -0.0008; -0.0005 -0.0006 -0.0010 -0.0006 0.0129 -0.0009; -0.0002 -0.0001 -0.0006 -0.0008 -0.0002 0.0150]; b2=(10^-2)*(10^-3)*[-0.3908 -0.1297 0.7047 0.0591 0.2161 -0.6635]; B00=(10^-2)*0.056; rpow=x; cpow=x'; B=rpow*b1*cpow; B01=b2*cpow; loss=B+B01+B00; if(((pin(1)-120)<x(1) && (pin(1)+80)>x(1))) if(((pin(2)-90)<x(2) && (pin(1)+50)>x(2))) if(((pin(3)-100)<x(3) && (pin(3)+65)>x(3))) if(((pin(4)-90)<x(4) && (pin(4)+50)>x(4))) if(((pin(5)-90)<x(5) && (pin(5)+50)>x(5))) if(((pin(6)-90)<x(6) && (pin(6)+50)>x(6))) %Total cost cost=f1+f2+f3+f4+f5+f6; %Power Balance equation error= ((x(1)+x(2)+x(3)+x(4)+x(5)+x(6))-(pd+loss))^2; %Fitness Equation val=((error*10000))+cost; else val=10^10; end else val=10^10; end else val=10^10; end else val=10^10; end else val=10^10; end else val=10^10;

52

end %Main Function calling ELD computing function clear clc pd=1263; ub=[500 200 300 150 200 120]; lb=[100 50 80 50 50 50]; x=GODLIKE(@check_final1,100,lb,ub,'DE'); x tot_gen=x(1)+x(2)+x(3)+x(4)+x(5)+x(6) f1=240 f2=200 f3=220 f4=200 f5=220 f6=190 + + + + + + 07.0*x(1) 10.0*x(2) 08.5*x(3) 11.0*x(4) 10.5*x(5) 12.0*x(6) + + + + + + 0.0070*x(1)^2; 0.0095*x(2)^2; 0.0090*x(3)^2; 0.0090*x(4)^2; 0.0080*x(5)^2; 0.0075*x(6)^2;

b1=(10^-2)*[0.0017 0.0012 0.0007 -0.0001 -0.0005 -0.0002; 0.0012 0.0014 0.0009 0.0001 -0.0006 -0.0001; 0.0007 0.0009 0.0031 0.0000 -0.0010 -0.0006; -0.0001 0.0001 0.0000 0.0024 -0.0006 -0.0008; -0.0005 -0.0006 -0.0010 -0.0006 0.0129 -0.0009; -0.0002 -0.0001 -0.0006 -0.0008 -0.0002 0.0150]; b2=(10^-2)*(10^-3)*[-0.3908 -0.1297 0.7047 0.0591 0.2161 -0.6635]; B00=(10^-2)*0.056; rpow=x; cpow=x'; B=rpow*b1*cpow; B01=b2*cpow; loss=B+B01+B00 error= ((x(1)+x(2)+x(3)+x(4)+x(5)+x(6))-(pd+loss))^2 cost=f1+f2+f3+f4+f5+f6

53

A.1.2.Economic Load Dispatch with Renewable Energy: A.1.2.1 No wind Case:

% Main function % Program to calculate ED with renewable sources % Problem without wind power clc clear fprintf('No wind case\n'); global demand pre PD = [575 725 675 750 760 850 1180 1210 1125 1050 940 900]; lb=[75 50 100 50]; ub=[400 225 500 300]; up_ramp=[70 90 100 60 ]; down_ramp=[95 110 135 75 ]; lbb=[75 50 100 50 0]; ubb=[400 225 500 300 150]; pre=[150 50 200 50]; % Intial state assumed for i=1:12 fprintf('-------------------------------------------------------------------------------------------------------------------------------\n'); demand=PD(i); % Check for reserve limit available=pre+up_ramp; tot_reserve=sum(available); percentage_reserve=((tot_reserve-demand)/demand)*100; if percentage_reserve<=25 sol=GODLIKE(@eldrwb_fcn_nowind, 200, lbb, ubb, 'DE'); Bcost=80*sol(5); else sol=GODLIKE(@eldr_fcn_nowind, 200, lb, ub, 'DE'); sol(5)=0; Bcost=80*sol(5); end fprintf('No. of hour %d \n',i); sol demand sum(sol) cost=sum(cost_gen(sol)); fprintf('Cost of Thermal Generation: %f \n',cost); % To compute reserve capacity res(1)=pre(1)+70-sol(1); res(2)=pre(2)+90-sol(2); res(3)=pre(3)+100-sol(3); res(4)=pre(4)+60-sol(4); rescap=(sum(res)/demand)*100; fprintf('Available reserve capacity: %f \n',rescap); if rescap <=25 fprintf('Biodiesel is added\n'); %up ramp limit %down ramp limit

54

else fprintf('Thermal Reserve is available\n'); end Rcost=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); Tcost=cost+Rcost+Bcost; fprintf('Cost of Thermal Generation: %f \n',cost); fprintf('Cost of Thermal Reserve: %f \n',Rcost); fprintf('Cost of Bio-diesel Generation: %f \n',Bcost); fprintf('Total Cost : %f \n',Tcost); pre=[sol(1) sol(2) sol(3) sol(4)]; end

55

% Calculation Function without Bio-Diesel function val=eldr_fcn_nowind(x) global demand pre %To check if generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %Cost of Generation cth=cost_gen(x); cost=sum(cth); %Cost of Reserve %To compute reserve capacity res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); %Total Cost Tcost=cost+cost_res; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

56

%Calculation Function with Bio-diesel function val=eldrwb_fcn_nowind(x) global demand pre %To check if generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %Cost of Reserve %To compute reserve capacity res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); %Total cost cth=cost_gen(x); Tcost=sum(cth)+(x(5)*80)+cost_res; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

57

%cost Calculation function ret=cost_gen(pg) % Cost calculation for if pg(1)<= 100 cth(1)=12*pg(1); elseif pg(1) <= 200 cth(1)=20*pg(1); elseif pg(1) <= 300 cth(1)=28*pg(1); else cth(1)=36*pg(1); end % Cost calculation for if pg(2)<= 75 cth(2)=25*pg(2); elseif pg(2) <= 125 cth(2)=29*pg(2); elseif pg(2) <= 175 cth(2)=33*pg(2); else cth(2)=35*pg(2); end % Cost calculation for if pg(3)<= 125 cth(3)=22*pg(3); elseif pg(3) <= 250 cth(3)=26*pg(3); elseif pg(3) <= 375 cth(3)=30*pg(3); else cth(3)=35*pg(3); end % Cost calculation for if pg(4)<= 75 cth(4)=16*pg(4); elseif pg(4) <= 175 cth(4)=25*pg(4); elseif pg(4) <= 225 cth(4)=31*pg(4); else cth(4)=40*pg(4); end ret=cth; P1

P2

P3

P4

58

A.1.2.2 Negative Load:

% Main function % Program to calculate ED with renewable sources % Problem with wind power Negative load model clc clear global demand pre PD = [575 725 675 750 760 850 1180 1210 1125 1050 940 900]; Pwind=[68 40 60 118 169 150 70 41 48 85 159 142]; wind_res=[20 22.5 24 25 26 27 27 27 27 27 27 27]; lb=[75 50 100 50]; ub=[400 225 500 300]; up_ramp=[70 90 100 60]; down_ramp=[95 110 135 75]; lbb=[75 50 100 50 0]; ubb=[400 225 500 300 150]; pre=[150 50 200 50]; for i=1:12 % Intial state assumed %up ramp limit %down ramp limit

fprintf('-------------------------------------------------------------------------------------------------------------------------------\n'); demand=PD(i)-Pwind(i); sol(5)=0; % Check for reserve limit available=pre+up_ramp; tot_reserve=sum(available); percentage_reserve=((tot_reserve-demand)/demand)*100; if (percentage_reserve) <= (wind_res(i)+25) sol=GODLIKE(@eldrwb_fcn_wind, 200, lbb, ubb, 'DE'); Bcost=80*sol(5); else sol=GODLIKE(@eldr_fcn_wind, 200, lb, ub, 'DE'); sol(5)=0; Bcost=80*sol(5); end fprintf('No. of hour %d \n',i); sol demand sum(sol) cost=sum(cost_gen(sol)); cost_wind=3082+0.01*Pwind(i);

To compute reserve capacity res(1)=pre(1)+70-sol(1); res(2)=pre(2)+90-sol(2); res(3)=pre(3)+100-sol(3); res(4)=pre(4)+60-sol(4); rescap=(sum(res)/demand)*100; fprintf('Available reserve capacity: %f \n',rescap);

59

if (rescap) <=(25+wind_res(i)) fprintf('Biodiesel is added\n'); else fprintf('Thermal Reserve is available\n'); end Rcost=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); Tcost=cost+Rcost+Bcost; fprintf('Cost of Thermal Generation: %f \n',cost); fprintf('Cost of Thermal Reserve: %f \n',Rcost); fprintf('Cost of Bio-diesel Generation: %f \n',Bcost); fprintf('Total Cost : %f \n',Tcost); pre=[sol(1) sol(2) sol(3) sol(4)]; end

60

%Calculation Function without Bio-Diesel function val=eldr_fcn_wind(x) global demand pre %To check if generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %Cost of Generation cth=cost_gen(x); cost=sum(cth); %To compute reserve capacity and cost res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); %Total Cost Tcost=cost+cost_res; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

61

%Calculation function with Bio-diesel function val=eldrwb_fcn_wind(x) global demand pre %To check if generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %To compute reserve capacity res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+res(2)*2+res(3)*1.5+res(4)*1.75; %Total cost cth=cost_gen(x); Tcost=sum(cth)+(x(5)*80)+cost_res; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

62

A.1.2.3 Inclusive Approach:

% Main function % Program to calculate ED with renewable sources % Problem with wind power Inclusive model clc clear global demand pre PD = [575 725 675 750 760 850 1180 1210 1125 1050 940 900]; Pwind=[68 40 60 118 169 150 70 41 48 85 159 142]; wind_res=[20 22.5 24 25 26 27 27 27 27 27 27 27]; up_ramp=[70 90 100 60]; down_ramp=[95 110 135 75]; pre=[150 50 200 50]; for i=1:12 %up ramp limit %down ramp limit % Intial state assumed

fprintf('-------------------------------------------------------------------------------------------------------------------------------\n'); demand=PD(i); % Check for reserve limit available=pre+up_ramp; tot_reserve=sum(available); percentage_reserve=((tot_reserve-demand)/demand)*100; if (percentage_reserve) <= (wind_res(i)+25) fprintf('Biodiesel is added\n'); lbb=[75 50 100 50 0 0]; ubb=[400 225 500 300 150 Pwind(i)]; sol=GODLIKE(@eldrwb_fcn_wind, 200, lbb, ubb, 'GA'); Bcost=80*sol(5); Wcost=3082+(0.01*sol(6)); else fprintf('Biodiesel is not added\n'); lbb=[75 50 100 50 0]; ubb=[400 225 500 300 Pwind(i)]; sol=GODLIKE(@eldr_fcn_wind, 200, lbb, ubb, 'GA'); sol(6)=0; Bcost=80*sol(6); Wcost=3082+(0.01*sol(5)); end fprintf('No. of hour %d \n',i); sol demand sum(sol) cost=sum(cost_gen(sol));

To compute reserve capacity res(1)=pre(1)+70-sol(1); res(2)=pre(2)+90-sol(2); res(3)=pre(3)+100-sol(3); res(4)=pre(4)+60-sol(4); rescap=(sum(res)/demand)*100; fprintf('Available reserve capacity: %f \n',rescap);

63

if (rescap) <=(25+wind_res(i)) fprintf('Biodiesel is added\n'); else fprintf('Thermal Reserve is available\n'); end Rcost=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); Tcost=cost+Rcost+Bcost+Wcost; fprintf('Cost of Thermal Generation: %f \n',cost); fprintf('Cost of Thermal Reserve: %f \n',Rcost); fprintf('Cost of Bio-diesel Generation: %f \n',Bcost); fprintf('Cost of Wind Generation: %f \n',Wcost); fprintf('Total Cost : %f \n',Tcost); pre=[sol(1) sol(2) sol(3) sol(4)]; end

64

%calculation function without bio-diesel function val=eldr_fcn_inclusive(x) global demand pre %To check generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %Cost of Generation cth=cost_gen(x); cost=sum(cth); %To compute reserve capacity and cost res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); wind_cost=(0.01*x(6)); %Total Cost Tcost=cost+cost_res+wind_cost; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

65

%calculation function with Bio-diesel function val=eldrwb_fcn_inclusive(x) global demand pre %To check if generation is withih ramp limits if((pre(1)-95)<=x(1)) && ((pre(1)+70)>=x(1)) if((pre(2)-110)<=x(2)) && ((pre(2)+90)>=x(2)) if((pre(3)-135)<=x(3)) && ((pre(3)+100)>=x(3)) if((pre(4)-75)<=x(4)) && ((pre(4)+60)>=x(4)) fail=0; else fail=1; end else fail=1; end else fail=1; end else fail=1; end if fail == 0 %To compute reserve capacity res(1)=pre(1)+70-x(1); res(2)=pre(2)+90-x(2); res(3)=pre(3)+100-x(3); res(4)=pre(4)+60-x(4); cost_res=res(1)+(res(2)*2)+(res(3)*1.5)+(res(4)*1.75); %Total cost cth=cost_gen(x); wind_cost=(0.01*x(6)); Tcost=sum(cth)+(x(5)*80)+cost_res+wind_cost; %Power Balance equation error= (sum(x)-demand)^2; %Fitness evaluation val=((error*10000))+Tcost; else val=10^10; end

66

A.2 Problem Data IEEE DATA FOR 6 UNIT TEST SYSTEM The system contains six thermal units. 26 buses, and 46 trimmission lines. The load demand is 1263 MW. The cost coefficients of 6 unit test system are given in Tables I. The ramp rate limits of corresponding generating units are given in Table II. The generalized B loss coefficients for the system are shown Table III. The system data is on 100 MVA base.
Table I. Generating Unit Capacity and Coefficient Pimin
(MW)

Table II. Ramp Rate Limits and Prohibited Operating Zones Pio
(MW)

Units

Pimax
(MW)

ai
($/MWh2)

bi
($/MWh)

ci 240 200 220 200 220 190

Uri
(MW/h)

Dri
(MW/h)

Prohibited zones (MW) [210 240] [350 380] [90 110] [140 160] [150 170] [210 240] [80 90] [110 120] [90 110] [140 150] [75 85] [100 105]

1 2 3 4 5 6

100 50 80 50 50 50

500 200 300 150 200 120

0.0070 0.0095 0.0090 0.0090 0.0080 0.0075

7.0 10.0 8.5 11.0 10.5 12.0

440 170 200 150 190 110

80 50 65 50 50 50

120 90 100 90 90 90

Table III. Generalized loss coefficient Bij= [0.0017 0.0012 0.0007 -0.0001 -0.0005 -0.0002 0.0012 0.0014 0.0009 0.0001 -0.0006 -0.0001 0.0007 0.0009 0.0031 0.0000 -0.0010 -0.0006 -0.0001 0.0001 0.0000 0.0024 -0.0006 -0.0008 -0.0005 -0.0006 -0.0010 -0.0006 0.0129 -0.0009 -0.0002 -0.0001 -0.0006 -0.0008 -0.0002 0.0150 ] B01=(10-3)[-0.3908 -0.1297 0.7047 0.0591 0.2161 -0.6635] B00=0.056;

67

References [1]. Power Generation, Operation, And Control. Allen J. Wood, Bruce F, Wollenberg. [2]. Power System Analysis. Hadi Saadat [3]. Economic Load Dispatch Optimization of Renewable Energy in Power System using Genetic Algorithm. Warsono, D. J. King, C. S. zveren and D.A. Bradley. [4]. A novel string structure for economic dispatch problems with practical constraints. Cheng-Chien Kuo. [5]. Using Evolutionary Computation to Solve the Economic Load Dispatch Problem. Samir SAYAH, Khaled ZEHAR.

68