人工智能在电力系统优化中的应用

需积分: 8 49 浏览量更新于2024-07-17 1 收藏 7.75MB PDF 举报

"Artificial Intelligence in Power System Optimization" 是一本由Weerakorn Ongsakul和Dieu Ngoc Vo合著的书籍，主要探讨了人工智能在电力系统优化中的应用。书中涵盖了粒子群优化、进化编程、模糊逻辑和增强拉格朗日霍普菲尔德网络等AI技术，用于解决电力系统运行和电力市场中的优化问题。电力系统是现代社会的关键基础设施，其稳定高效运行对于经济和社会发展至关重要。随着科技的进步，人工智能已经成为电力系统优化的重要工具。在这本书中，作者深入浅出地介绍了如何利用这些先进的AI算法来处理电力系统的复杂问题，如发电调度、输电网络优化、负荷预测以及电力市场的交易策略等。粒子群优化（Particle Swarm Optimization, PSO）是一种基于群体智能的全局优化算法，模拟了鸟群或鱼群的行为，用于寻找复杂空间中的最优解。在电力系统中，PSO可以用于发电机的调度，以最小化运行成本或最大化发电效率。进化编程（Evolutionary Programming, EP）是遗传算法的一种变体，通过模拟生物进化过程来搜索问题的解决方案。在电力系统中，EP可用于优化电力网络的配置，提高系统的稳定性，并降低故障风险。模糊逻辑（Fuzzy Logic）允许处理不确定性和模糊性的数学工具，它在电力系统中的应用包括不精确数据的处理、控制策略的制定，以及对电网状态的评估。例如，模糊逻辑可以用于智能断路器的控制，以更精确地管理电网的开关操作。增强拉格朗日霍普菲尔德网络（Augmented Lagrange Hopfield Network）是一种结合了神经网络和优化理论的模型，特别适用于求解约束优化问题。在电力系统中，它可以用于解决如电压稳定性、潮流计算等问题，同时考虑系统的物理限制。这本书的内容不仅对电力系统管理的学生有教育价值，也为从事电力系统运行和规划的专业人员提供了宝贵的参考。通过学习这些AI技术，读者将能够更好地理解和应对电力系统中的挑战，提升电力系统的性能和效率。"Artificial Intelligence in Power System Optimization" 是电力行业专业人士理解和应用人工智能技术的宝贵资源，有助于推动电力系统的现代化和智能化进程。

Introduction 5

Taboo search [6] is also a meta-heuristic search for solving combinatorial

optimization problems in management science, industrial engineering, economics,

and computer science. This method belongs to the local search techniques but it

enhances the performance of local search methods using memory structures to

match them with local minima at the beginning. Once a potential solution has been

obtained, it is marked as taboo, thus the algorithm does not visit that possibility

again and again during the search process. Taboo search was developed in the 1970s

and recently has been widely used for its powerful search capabilities.

Ant colony optimization algorithm [22] is a probabilistic technique to solve

optimization problems. It can be reduced to the problem of ﬁ nding the shortest

paths through graphs based on the behavior of ants in ﬁ nding food for their colony

by marking their trails with pheromones. The shortest path is the trail with the most

pheromone marks which the ants will use to carry their food back home. The ﬁ rst

algorithm was developed in 1991 and since then, many variants of this principle

have been developed.

Genetic algorithm [2, 4] is a search technique used to ﬁ nd the exact or

approximately best solution for optimization problems. The genetic algorithm

belongs to evolutionary computation using the techniques inspired by evolutionary

biology such as inheritance, mutation, selection and crossover. The genetic algorithm

was developed part by part from the 1950s onward and is one of the most popular

methods applied to various optimization problems in bioinformatics, phylogenetics,

computer science, engineering, economics, chemistry, manufacturing, mathematics,

physics and other ﬁ elds. This method can take long computational times to ﬁ nd

the optimal solution.

Evolutionary programming [8] also follows the evolutionary computation

paradigms to ﬁ nd the globally optimal solution for an optimization problem.

Evolutionary programming was developed in 1960 placing emphasis on the

behavior of the linkage between parents and their offspring rather than trying to

emulate the speciﬁ c genetic operators as observed in nature. The main operators

of evolutionary programming consist of mutation, evaluation and selection. This

method is also widely used in different optimization techniques due to its powerful

search capabilities.

Particle swarm optimization [6] is one of the most heuristic algorithms

developed under emulation of the simpliﬁ ed social behavior of animals in swarms,

e.g., in ﬁ sh schools and bird ﬂ ocks. It is a population based evolutionary algorithm

found to be efﬁ cient in solving continuous non-linear optimization problems.

Particle swarm optimization provides a population-based search procedure, in which

individuals (particles) change their positions (states) over time. It uses a velocity

vector based on the social behavior of the individuals of the population to update

the current position of each particle in the swarm ﬂ ying in a multidimensional

search space of a problem. During the ﬂ ight, each particle with a certain velocity is

dynamically adjusted according to its ﬂ ight experience and that of its neighboring

particles to ﬁ nd the best position for itself among its neighbors. The particle swarm

6 Artiﬁ cial Intelligence in Power System Optimization

optimization technique can deliver a high-quality solution within shorter calculation

time with more stable convergence characteristics than other stochastic methods.

Developed since 1995, particle swarm optimization has been successfully applied in

many researches and application areas such as engineering, ﬁ nance and management

systems and is now one of the most widely used methods in optimizations.

Differential evolution [23], belonging to the class of evolution strategy

optimizers, is a method of mathematical optimization of multidimensional functions

to ﬁ nd the global minimum of a multidimensional and multimodal function fairly

fast and reasonably robust. Developed in the mid 1990s, the differential evolution

method is a simple population based and stochastic function minimizer and has

become one of the most popular methods used by researchers. The central idea of

this method is a scheme to generate trial parameter vectors by adding the weight

difference between two population vectors to a third one that makes the scheme

completely self-organizing. The trial vector is used for the next generation if it

yields a reduction in the value of an objective function.

In general, the methods based on artiﬁ cial intelligence are continuously

developed further for other application in different power system optimization

problems. Recently, hybrid systems combining the strengths of each single method

have been favored by researchers due to various advantages over the single methods

as presented above.

1.3 ARTIFICIAL INTELLIGENCE APPLICATIONS IN POWER

SYSTEMS

For the two recent decades, artiﬁ cial intelligence based methods have become

popular for solving different problems in power systems such as control, planning,

forecast, scheduling, etc. These methods can deal with the complex tasks faced

by applications in modern large power systems with ever more interconnections

installed to meet the increasing load demand. The application of these methods has

been successful in many areas of power system engineering.

Some major problems of power systems that artiﬁ cial intelligence methods

have been applied to, include planning, operation and modeling analysis.

• Power system expansion planning [24]: The structure of a typical electrical

power system is very large and complex including basic components as

generators, AC and DC transmission systems, a distribution system, load, and

FACTS devices. The main objective of least cost system expansion planning

is to optimize the components necessary to provide an adequate energy supply

at minimum cost. In power system expansion planning, many factors have

to be taken into consideration such as demand, line and FACTS placement,

environmental effects, etc. This includes the planning of any expansion of

generation, transmission and distribution. Further issues also considered in

this task are reactive power planning, reliability analysis and network design

among others.

Introduction 7

• Power system operation [6]: In real-time power system operation, the online

power generation should meet the total load requirement plus transmission

losses in a reliable and secure manner. In this approach, operating power

plants and the transmission system are considered. The problems involved

include generation scheduling, economic dispatch, optimal power ﬂ ow, unit

commitment, hydrothermal scheduling, reactive power dispatch, voltage

control, load frequency control, static and dynamic security assessment,

maintenance scheduling, fuel scheduling, contract management, equipment

monitoring and dynamic generator rescheduling. For supervisory purposes,

data acquisition and the energy management system (SCADA/EMS),

load forecasting, load management, alarm processing, fault analysis and

diagnosis, service restoration, network/substation switching, contingency

analysis, optimal power ﬂ ow and state estimation functions are included in

the design.

• Power system modeling/analysis [11]: The issues of power system modeling/

analysis include power ﬂ ow analysis, transient stability, harmonics, dynamic

stability, control design, simulation, protection and bad data detection.

Most recently, artiﬁ cial intelligence based methods have been applied to

problems in deregulated environments aiding congestion management, optimization

of bidding strategies and generation scheduling [25].

However, there are still many challenges ahead. Therefore, these methods

are continuously developed and improved to deal with ever larger complex power

systems with an increasing number of constraints in both the traditional electric

power supply industry and the competitive electricity market environment.

1.4 OVERVIEW OF THE BOOK

This book includes seven chapters solving optimization problems in power

systems before and after deregulation addressed by both conventional and artiﬁ cial

intelligence methods. In each chapter, the problem formulation and the related

solution methods are given. At the end of every chapter, practical problems are

included to enable readers to practice further with the insight gained. The main

contents of the chapters are brieﬂ y given as follows:

Chapter 1 gives an introduction to the signiﬁ cance of optimization problems

in power systems and reviews the newest trends in applying artiﬁ cial intelligence.

These applications are also mentioned in this chapter.

The economic dispatch problem with different objective functions and

constraints is comprehensively covered in Chapter 2. Economic dispatch is one

of the attempts in power system operation to optimize the online units’ schedule in

the least expensive manner. The problems with economic dispatch and its various

constraints such as ramp rate, transmission and emissions are considered ﬁ rst.

Economic dispatch problems like non-smooth cost functions including prohibited

8 Artiﬁ cial Intelligence in Power System Optimization

operating zones and multiple fuels are introduced. Other economic dispatch

problems of multi-objective, combined heat and power and hydrothermal systems

are also presented. An optimal dispatch solution for a competitive electricity

market is also covered. Complementary, in this chapter, solution methods based

on artiﬁ cial intelligence including particle swarm optimization and augmented

Hopﬁ eld Lagrange network are described. The particle swarm optimization method

developed through simulation of simpliﬁ ed social models is one of the most modern

heuristic algorithms. It can deal with non-convex optimization problems similar

to other evolutionary algorithms. The Augmented Lagrange Hopﬁ eld network

is a continuous Hopﬁ eld neural network having its energy function based on an

augmented Lagrange function. This network is simpler than a Hopﬁ eld network

and more efﬁ cient in ﬁ nding the optimal solution within a shorter computational

time. In addition, fuzzy linear programming, a method based on the vagueness in

linguistics, is also used for solving economic dispatch problems in traditional and

competitive markets. The mathematical models and explanations for these methods

are also provided so that the reader can easily understand them.

In Chapter 3, unit commitment is used as the criteria for operation planning

for one day up to one week ahead based on load forecast considering generating

unit statuses as discrete variables and time dependent constraints. The solution

methods have to handle both continuous and discrete variables. In this problem,

many constraints are considered such as ramp rate, transmission line, environmental

emissions, fuel limitations etc. Additionally, the unit commitment approach in

competitive markets based on the maximization of total revenue is also considered.

The methods suggested in this chapter are a new adaptive Lagrangian relaxation

and hybrid systems based on generating a unit merit order, Hopﬁ eld network

and Lagrangian relaxation. In the adaptive Lagrangian relaxation method, the

Lagrangian multipliers are adaptively adjusted to enhance the convergence rate.

Moreover, introducing a new on/off criterion also improves the convergence in this

method. The augmented Lagrange-augmented Hopﬁ eld network is also an efﬁ cient

method to solve the unit commitment problem. This method is a combination of

both continuous and discrete Hopﬁ eld networks with its energy function based

on the augmented Lagrange function. Another method based on a merit order of

generating units is also efﬁ cient in resolving unit commitment issues. The merit

order of generating units based on their average production cost is effective in

determining the thermal unit scheduling. In the methods based on this, heuristic

search is needed to enhance the results or repair constraint violations. Besides this,

the sub-procedures of these methods for certain tasks are illustrated.

One of the most complex problems in generation scheduling optimization,

hydrothermal scheduling, is presented in chapter 4. The role of hydro power

plants has become more important in power systems due to their comparably low

environmental impact and—over lifetime—marginal cost. However, the capacity

of hydro power plants depends on reservoir capacity and weather. Therefore, the

optimal scheduling of hydro-thermal systems is of utmost importance, especially

Introduction 9

in systems with dominant hydro power plants. This problem includes hydraulic

constraints in addition to the constraints of thermal systems in unit commitment as

described in chapter 3. In the methods including adaptive Lagrangian relaxation and

hybrid systems based on generating merit order, Hopﬁ eld network and Lagrangian

relaxation are applied. The methods in this chapter are similar to the ones applied

to the unit commitment problem.

Optimal power ﬂ ow is addressed in chapter 5. The optimal power ﬂ ow

approach has been researched for a long time and has drawn more attention

recently due to advancing numerical optimization techniques and computer and

communication technology. With this approach, many constraints can be handled

such as transmission capacity, real and reactive power limits of generators, bus

voltages, security margins, transformer taps etc. The solution methods for this

problem include conventional and artiﬁ cial intelligence methods such as linear

programming, quadratic programming and augmented Lagrange Hopﬁ eld network.

The conventional methods are applicable only to convex power ﬂ ow optimization

problems while artiﬁ cial intelligence based methods can easily resolve a non-convex

non-differentiable OPF by both augmented Lagrange and sigmoid function of the

Hopﬁ eld network.

Chapter 6 introduces the optimal reactive power dispatch in a power system.

The reactive power performs no real work but, in the contrary, consumes resources.

In fact, it is consumed throughout network elements and loads and has to be supplied

by reactive power sources in the system. This chapter provides various models

for optimal reactive power dispatch for passive and active elements consuming

reactive power. In addition, the optimal reactive power dispatch before and after

deregulation in power systems is also covered.

The problem of available transfer capability (ATC) is presented in chapter 7.

With the movement towards a competitive market, open access to the transmission

system plays an important role in interconnected transmission networks. Electric

power transfers will increase and the reliable operation of the transmission networks

becomes a difﬁ cult task. In this chapter, the concept, deﬁ nition, principles and

calculation methodologies of available transfer capacity are introduced. Besides,

the conventional methods applied to calculate the available transfer capability

such as linear approximation, continuation power ﬂ ow and repetitive power ﬂ ow,

stability-constrained ATC, and optimal power ﬂ ow based methods, meta-heuristic

based methods including evolutionary programming and hybrid evolutionary

algorithms are also proposed for ATC calculation.

1.5 REFERENCES

1. A. J. Wood and B. F. Wollenberg, Power generation, operation and control. 2nd edn., Wiley

and Sons, New York, 1996.

2. J. A. Momoh, Electric power system applications of optimization, Marcel Dekker, Inc., New

York, 2001.

3. E. El-Hawary and G. S. Christensen, Optimal economic operation of electric power systems,

剩余517页未读，继续阅读

quillmm

粉丝: 1
资源: 4

人工智能在电力系统优化中的应用

Artificial Intelligence in Medicine

Artificial Intelligence A Modern Approach (3rd Edition)

string1 = “Artificial Intelligence”, string2 = “Artificial Intelligence”,判断它们的id是否相同？

The current research topic in the field of artificial intelligence

aigc 是 AI-Generated Content ，你为什么说是 Artificial Intelligence Game Competition ？

Feng, L. (2022). Application Analysis of Artificial Intelligence Algorithms in Image Processing. Mathematical Problems in Engineering, 2022. https://doi.org/10.1155/2022/7382938这是什么

INTERNATIONAL SUMMIT ON ROBOTICS AND ARTIFICIAL INTELLIGENCE

英语辩论Should We Rapidly Develop the Artificial Intelligence?正方一辩怎么写稿

What thinking does artificial intelligence give us? How should we respond

最新资源