Research on Load Forecasting of Multi-scenario Building Air-Conditioning Based on TSO-LSTM Algorithm

This paper proposes a short-term air-conditioning load forecasting model of multi-scenario building based on Tuna Swarm Optimization (TSO) algorithm and long short-term memory network (LSTM). The model uses the TSO algorithm to optimize the hyperparameters of LSTM, which solves the problem that the manual selection of hyperparameters of LSTM neural network is difficult to obtain the optimal results. The TSO algorithm has the characteristics of fast global convergence and high searching optimal ability, which can optimize the number of neurons in the hidden layer, learning rate, and training times of the LSTM neural network. Through the analysis of different case data, the forecasting effect of the proposed algorithm is verified. Forecasting results of LSTM, PSO-LSTM, and TSO-LSTM algorithms are compared in different scenario building airconditioning load, which shows that the proposed TSO-LSTM algorithm has better accuracy and rapidity.