A short-term gas load forecasting using a hybrid model based on fuzzy code genetic algorithm and improved LSTM-BPNN residual modification method
The accurate forecasting of the natural gas daily load plays a very important role in the reasonable supply and dispatch of energy in the city. Due to the complex characteristics of gas load data itself, which is periodic and random, and the limitations of the single-stage single forecasting model, this paper proposes a Multi-stage hybrid model of fuzzy coding genetic algorithm and improved LSTM-BPNN residual correction model. First, in the first stage, the LSTM model make a preliminary forecasting of the gas load, and then calculating the residual value of the gas load. In the second stage, The BPNN model predict the residual value, and then using the Adam adaptive learning rate algorithm to automatically adjust the learning rate of LSTM-BPNN during the learning process, wich accelerates the fitting speed, and then the fuzzy coding genetic algorithm optimize the initial weights and thresholds of the BPNN in order to find the global optimal solution. Finally, the sum of the forecasting values of the two stages is used as the final forecasting gas load. Through comparative experiments, it is concluded that the model in this paper has higher prediction accuracy than the single model and the original two-stage forecasting model.