Research On Power System Short Term Load Forecasting Method Based On Deep Learning Combination Model

With the continuous progress of global industrialization,the demand for electricity is increasing,so short-term load forecasting is particularly important.At present,many deep learning models have been applied to short-term load forecasting,but most of them ignore the error accumulation in the iterative training process.At the same time,the error sequence generated in the iterative process has high volatility and instability,which makes it difficult to predict the error sequence.In order to further improve the prediction accuracy,this paper proposes a new combination forecasting idea.In this paper,the basic principles of the three traditional deep learning models are described in detail,and the experimental comparison and analysis of the three models are carried out from the aspects of training time consumption and prediction accuracy.Secondly,in order to verify the effectiveness of the combined forecasting method,this paper proposes two specific combination algorithm models for short-term power system load forecasting.The first model combines the stacked bi-directional gated recurrent unit(SBi GRU)and the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and error correction.In the first stage of the model,SBi GRU model is established to learn the main characteristics of load series,and the error series generated in the prediction process of SBi GRU model reflects the error characteristics of load series;then in the second stage,CEEMDAN algorithm is used to decompose the error series into several relative stationary intrinsic mode function(IMF)component and residual component,the SBi GRU model is established again for each component to learn and predict,and the predicted values of all components are reconstructed in sequence to get the error prediction results;finally,the prediction results of the two stages are summed to achieve the purpose of correcting the error.Thirdly,in order to further verify the feasibility and effectiveness of the combination algorithm,this paper continues to propose the second combination algorithm model: SAE-CEEMDAN-Bi LSTM combined forecasting model.In this model,CEEMDAN combined with bidirectional long short-term memory(Bi LSTM)is used to correct the errors in the prediction process of stacked autoencoder(SAE).Finally,two different data sets are used to verify the performance of the two combination models.The two models have achieved good prediction results in their respective data sets,which proves the superiority of the combination forecasting idea.