Research On Power Load Forecasting Of Neural Network Based On Improved Whale Algorithm

In the context of smart grids,load forecasting is increasingly important.Neural network is a common method in load forecasting,but the prediction error increases due to the low accuracy of hyperparameter selection.Therefore,this paper proposes a method of selecting neural network hyperparameters based on the improved whale algorithm,and uses the optimized hyperparameters to establish a load forecasting model.The main work done in this paper is as follows:(1)Firstly,the data used in power load forecasting is introduced,and the data types,characteristics and influencing factors are analyzed.The predicted data is then processed for missing data,abnormal data detection and normalization.Finally,the dataset is divided for neural network modeling.(2)First of all,aiming at the problem of ignoring the load reversal feature in the long short-term memory neural network,the bidirectional long and short-term memory neural network is used to introduce the load reversal feature,and the prediction process of the bidirectional long and short-term memory neural network is given.Then,two methods of hyperparameter selection in the prediction process are introduced: the classical selection method and the hyperparameter selection method based on the whale optimization algorithm.Finally,in view of the problem that the above methods are difficult to select the optimal hyperparameters,a hyperparameter selection method based on the improved whale algorithm is proposed,and a bidirectional long short-term memory neural network prediction model is established by using the optimal hyperparameters.(3)Firstly,13 kinds of test functions and 4 kinds of prediction model evaluation criteria are described.Secondly,the comparison of test functions and algorithms proves that the hyperparameters selected based on the improved whale algorithm are more accurate,reliable and robust.Then,by comparing the impact of various improved methods on load forecasting,it is proved that the method used can further improve the load forecasting accuracy.Finally,the prediction model based on the improved whale algorithm is used to forecast the load in three scenarios.By comparing with the pre-improved prediction model,the prediction model based on the whale optimization algorithm,the long short-term memory neural network prediction model and the BP prediction model,it is proved that the improved whale algorithm The algorithm further improves the load forecasting accuracy.