Research on Short-Term Air Conditioning Cooling Load Forecasting Based on Bidirectional LSTM

The air-conditioning energy consumption accounts for a large proportion of the building energy consumption. Accurate forecasting of the air-conditioning load at the hour level is the basis for studying the time-sharing control strategies for building energy conservation and the reasonable arrangement of the operation mode of the power grid system. However, there are two aspects that affect the accuracy of the air-conditioning load forecasting: the forecasting model and the inconsistent signal sampling time. This paper proposes a short-term air-conditioning cooling load forecasting model based on bidirectional long short-term memory network (Bi-LSTM). Firstly, the interpolation and completion processing are performed for the air-conditioning cooling load data with different sampling times. Then the processed load data is used to train in the forecasting model. Finally, the forecasting strategies are optimized to obtain a better forecasting model. The optimized forecasting model is used to forecast the air-conditioning cooling load every 2 hours in the next 2 days. The experiments results show that the forecasting model based on Bi-LSTM has higher forecasting accuracy compared with the three recurrent neural network models of RNN, GRU and LSTM. Besides, the optimization results of the forecasting strategies show that the historical load data similar to the maximum temperature of the forecasted target day can be selected as the training set to improve the load forecasting accuracy.