Abstract: In smart cities, accurate residential load forecasting is the key to achieving a balance between power supply and demand and reducing resource waste. In order to improve the accuracy of load forecasting of urban residential districts, a short-term load forecasting model LSTM-AdaBoost, which combines long and short-term memory network (LSTM) and ensemble learning, is proposed. The model uses dew point (the temperature at which water vapor in the air condenses into water droplets), historical load, week type and other characteristics as data input; then the LSTM network with timing memory function is used as the base learner for integrated learning; finally, the ensemble AdaBoost  algorithm performs a weighted combination of the base learner to obtain a strong learner. Experimental results show that the integrated LSTM-AdaBoost model has higher forecast accuracy than single forecasting methods such as LSTM network, support vector machine (SVM) and CART decision tree.

Key words: residential load forecasting, long and short-term memory network, ensemble learning, AdaBoost