Short-term Power Load Data Forecasting Method Based On SSA-CNN-BILSTM

In recent years,with the expansion of power grid scale and the use of smart meters,the scale of power data is getting larger and larger.By processing the massive power load data to predict the future load,great economic benefits can be obtained,and it plays an important role in assisting the decision-making of power companies,improving the power grid benefits and assisting the power grid planning.At present,deep learning and artificial intelligence methods are popular in the research of short-term power load forecasting.However,at present,the accuracy of load forecasting is low and the selection of network parameters is locally optimal.Therefore,this paper has done the following work in short-term power load forecasting:1.Complete data analysis and processing.At present,there is a problem that data from different sources are not considered comprehensively in load forecasting.Firstly,a preliminary analysis of influencing factors based on experience is made.Aiming at the problems that some text data in engineering data and competition data samples can not be directly used,and there are missing values,abnormal values,nonstandard data,etc.,the quantitative processing and preprocessing methods are studied.Finally,based on Pearson correlation system analysis method,the influencing factors of two kinds of samples are determined.2.Establish the bidirectional long-term and short-term memory network(CNN-Bi LSTM)model with convolutional neural network.The bi-directional long-term and short-term memory network(Bi LSTM)model is established,and the reverse LSTM layer is added on the basis of LSTM to optimize the prediction results.Then add a layer of convolutional neural network(CNN)on the basis of Bi LSTM to improve the utilization rate of feature data.Finally,the comparison test proves that the prediction accuracy of this model is improved to some extent compared with other models.3.Establish CNN-Bi LSTM model based on sparrow search algorithm optimization.Aiming at the problem of local optimization in the selection of model network parameters,the sparrow search algorithm is used to train the model parameters to improve the prediction accuracy of the model.Comparative experiments show that the model optimized by sparrow search algorithm has improved the prediction accuracy.