Research Of Short-term Load Forecasting And Load Dispatch Optimization

Due to the non storage of electric energy and every step in power system has to been in balance,in order to make the system running economy and safe,it has to keep in balance of generation,transmission,distribution and power supply.Short-term load forecasting and the subsequent optimization of load dispatch is the main content of a economy and safe power system.In this thesis,the characteristic of short-term load forecasting and load dispatch optimization are in consideration,an improved support vector machine and a hybrid particle swarm sequential quadratic programming are proposed to solve the problems.The contents of this paper are as follow:This thesis,firstly,analyze the characteristic of power load,including the composition of load,and analyses the characteristics of the load,mainly analysis of the relation between load and temperature,characteristics of periodic load,and the influence of other factors on the load change.Then introduce the characteristic of short-term load forecasting.And it studied the load data preprocessing methods and the analysis of the prediction error.In this thesis,for the sample of short-term load forecasting is not so large,and the nonlinear characteristics of this problem are obvious,the method of support vector machine is adopted to solve this problem.When choosing the parameters of SVM,the commonly used grid search method based on cross validation for predicting result is not very ideal.In this thesis,genetic algorithm is used to search the parameters of SVM,so GA-SVM is proposed,and it can be used in the model of load forecasting.Finally compare the prediction result of the grid search method of SVM and GA-SVM,the GA-SVM prediction result is more accurate,has a smaller error,and the error distribution is more uniform.The purpose of short-term load forecasting is to optimize load dispatch.According to the characteristics of the load dispatch optimization,this thesis firstly study of optimal load dispatch model,mainly from the following two aspects:objective function and constraint conditions.Standard coal consumption function is chosen as objective function,in order to make the model played well in actually,the upper and lower boundary of power,power balance constraint,ramp rate constraint and prohibited operating zones are in consideration.This thesis adopts particle swarm optimization to solve the problem of load dispatch.When the scale of the problem is small,the particle swarm optimization has good result.But when the problem is very complicated,the convergence of the particle swarm algorithm is not so sufficient,especially in large-scale nonsmooth linear programming problem such as a system with 13 or 40 generators,the convergence rate of the particle swarm algorithm maybe slow.And as the SQP is recognized as a good way to solve nonlinear programming problem,so that the SQP is introduced into PSO method,the SQP-PSO method is proposed to solve the economic load dispatch problems,in order to improve the convergence of particle swarm optimization algorithm.And in the process of algorithm,this paper proposed a new strategy to deal with constraint conditions in view of the particle update after the processing of constraints.The results show that,when the problem is in smaller scale,the two algorithms almost have the same performance.But when the problem is very complicated,the SQP-PSO has a better performance than PSO,especially in the speed of convergence.