Multivariable Composite Prediction Based on Kalman Filtering and Charging and Discharging Scheduling Strategy of Energy Storage System

By estimating and predicting state variables of the distributed generation system and that of electric load in real time, the real-time state of energy storage system can be obtained, which will provide an important basis for the dispatching department of the power system to formulate the power generation plan and the transmission plan. In distributed PV generation and multi-load systems, both the power supply system and the power consumption system are multi-source and non-stationary. Firstly, based on the polynomial neural network framework, the model of state variables changing with time series is constructed. Further non-stationary of the model above is taken into consideration and is transformed into Kalman filter frame. Secondly, the block state model and prediction model of the related variables are established, which contributes to real-time block estimation and prediction of composite variables and is expected to overcome the shortcomings of existing methods. Finally, the performance of the new method is tested and verified by simulation.