Research And Application In The Electric Load Forecasting Of The Combined Models Based On PSO Optimization Weights

With the continuous renewal of algorithms and forecasting theories,all kinds of single,hybrid,combined forecasting models are emerging endlessly.In order to achieve good prediction effect,more and more combined models are put forward to improve the prediction accuracy by optimizing the weights of single models.Based on this theory,this paper puts forward a new combined method for forecasting analysis on the basis of previous studies,which combines the BP neural network model with the GA,CS,and PSO heuristic intelligent algorithms.The electric load data is taken for example to be applied in the model.Electric load is power which are consumed by all the electric equipment in the power system,besides electric load and power generated by the generator are unified whole,so the power system must adjust the power of the generator to balance the supply and provide the electricity which is reliable economic and up to standard for all types of users.Therefore,in order to reduce the effects of electric load fluctuation on the electricity grid,accurate and reliable electric load forecasting is of great importance.In this paper,we propose a new forecasting method by using this electric data of 3 years in the same month from the New South Wales in Australia to make predictions.We firstly make the forecasting to the electric data by using the original BP neural network and wavelet neural network.Secondly,we make further forecasting by using the BP neural network optimized by genetic algorithm to the electric load data;at the same time,we forecast the electric data by using BP neural network optimized by the cuckoo algorithm.At last,we optimize the weights of genetic algorithm and the cuckoo algorithm by using PSO algorithm and we get the value of the weights,then we get the finally forecasting results by weighted calculation.The research results show that compared with other methods,such as BP neural network,wavelet neural network,the BP neural network optimized by genetic algorithm,the neural network optimized by cuckoo algorithm;the combined model can effectively improve the predictive accuracy.Generally,the portfolio model has higher prediction precision and wide applicable scope compared with other individual model.