Refined electricity forecasting is an important trend in the development of power grids. Medium and long-term electricity forecasting is generally based on months or years to predict the total electricity consumption in the next few years. This paper first introduces several commonly used forecasting methods, and then proposes a variable-weight combination forecasting model that combines the advantages of a single model, calculates the weight coefficients according to the error index to combine the prediction results of the sub-models, and proposes a method to optimize the weight distribution according to the time distance of prediction error. Aiming at the problem of poor adaptability of a single model to changes in electricity, the weights are calculated separately in different months for prediction. Finally, taking the electricity data of a certain city as an example, the prediction on the monthly time scale and the annual time scale is carried out, and the error index is calculated. The results show that the method in this paper can effectively improve the reliability of medium and long-term electricity forecasting.

medium and long-term electricity forecast,Grey theory,Support vector machine,LSTM neural network,combined forecast