Power System Short-term Loading Forecasting by Genetic Neutral Network Based on Principal Components Analysis
Yi Mao · Hu'nan Shifan Daxue xuebao. Ziran kexue ban · 2011
The principal components analysis was used in order to solve two main weaknesses of BP neural network such as the speed of network's training is slow,sensitive to the initial weight and threshold,and easy to fall into the partial minimal point.The main component of load data can be obtained without losing initial load data of the main information so that the input quantity of the prediction model can be decreased.At the same time,the local convergence and other problems of BP algorithm can be effectively overcome based on the combination of genetic algorithm(GA) with BP neural network,and using GA's global searching function to optimize BP network's structure parameters.This model is applied in the power system short-term load forecasting simulation process.The results show that it can improve the performance of network and the accuracy of predictions.