Medium-Term Load Forecasting Based on Electricity Constraint

Accurate medium-term load forecasting provides data support for system operation maintenance and power purchase plan management, which has important practical and scientific significance. The traditional model for the multi-step load forecasting in the medium term is prone to the problem of error accumulation leading to the forecast out of control, and the power as the integral of the load can be used as an effective boundary constraint condition in the forecast. Therefore, this paper presents a step by step mid-term load forecasting algorithm based on LSTM and SVR. The algorithm takes the power-load step-by-step prediction as the forecasting idea, firstly realizes the medium-term forecast of daily power based on LSTM neural network, then constructs the weighted similar day data set through feature analysis, and finally realizes the power-load decomposition based on SVR model based on the similar day data set, and obtains the medium-term load prediction curve. This method effectively solves the problem of error accumulation in multi-step prediction, and the prediction result is more accurate and stable.