Short-term Power Load Forecasting Method Based On EEMD And Its Applications

Load forecasting is an important module in power system operation management.It aims to provide support for the safe and economic operation of power systems.By effectively forecasting of power system loads and balancing the power demand,the development and production plans are formulated to maintain the stability and safety of power systems and improve economic efficiency and energy efficiency.In fact,short-term load forecasting is of great significance for demand-side management,and it is the basis for system power allocation,preventive control,and emergency handling.In recent years,artificial intelligence-based load forecasting has flourished.This paper belongs to the multi-step prediction problem of non-stationary and nonlinear time series.It uses signal decomposition,deep time-series neural network theory,combined learning methods and other methods to conduct research on ultra-short-term industrial power demand prediction,and proposes a combined learning prediction model based on EEMD decomposition and deep time-series neural network.Because the load data has a certain time series correlation,in order to effectively utilize the time series characteristics,the classic long short-term memory network(LSTM)is used as the basic neural network unit to establish a neural network model.The signal non-stationary and non-linear original time series are converted into several sub-sequences by means of signal decomposition,and the high-frequency sub-sequences decomposed are re-decomposed to obtain the deep features implicit in the data,which effectively improves the accuracy of load prediction.Finally,the AEMO public data set and the real load data of a large industrial user were used for experimental verification.The experimental results were analyzed and evaluated with a variety of classic time-series prediction models.The comparison shows that the model in this paper can achieve obvious stable and optimal results on three commonly used predictive performance evaluation indicators of MAPE,MAE,and NRMSE.The model proposed in this paper is applied to the demand load forecasting of a large industrial power user.The error control is at a lower value and the acceptable range on site,which can assist the realization of demand control strategies with different emergency levels.