Research On Short-Term Power Load Forecasting Based On Combination Model Of IWOA And VMD

With the continuous development of power system and the connection of diversified energy sources to the grid,the operation of power system becomes more and more complex.In order to ensure the safe,reliable,stable and efficient operation of the power system,it is necessary to carry out the power load prediction for each link of the power system.Short-term power load prediction can provide scientific and reasonable plans for the operation and dispatching of the power system.Therefore,it is important to improve the accuracy of shortterm power load forecasting.Firstly,based on the network structure of convolutional network,long short-term memory network and attention mechanism,this paper constructs a combination model of short-term power load prediction.The combination model extracts the feature information of input data through the convolutional layer,and adds the attention layer after the pooling layer to further extract important features.The short-term load prediction output is carried out by combining the bidirectional long short-term memory network with the attention layer network.In order to solve the nonlinear and non-stationary problem of power load data,a combination model of variational mode decomposition(VMD)is proposed.VMD is used to decompose the power load data modal,and the combination model is used to predict each modal component one by one and reconstruct the predicted load value.Experiments show that the combined model based on VMD can effectively improve the accuracy of short-term power load prediction.Secondly,aiming at the difficulty of setting network super parameters of VMD combination model,improved whale optimization Algorithm(IWOA)was adopted to optimize the parameter selection of VMD combination model.The nonlinear strategy and random difference variation strategy are used to improve the convergence rate and optimization accuracy of WOA.The combined model based on IWOA and VMD takes the minimum mean square error of prediction as the target to optimize the selection of learning rate,training times and the number of nodes in the hidden layer of bidirectional long short-term memory network.Compared with different network models on different data sets,the experiment proves that the combination model based on IWOA and VMD has better prediction accuracy and adaptability.Finally,in order to further improve the practical application of the combined model of IWOA and VMD,the algorithm of the model is embedded in the Raspberry PI platform,and an intelligent short-term power load forecasting device is designed.By means of interface interaction,various functions such as user login,data import,model selection,model training and prediction are developed and designed.The actual power load data is used to test the shortterm power load prediction of the developed functions,which proves that the power load prediction platform developed based on Raspberry PI can complete the task of short-term power load prediction.