Research On Micro Power Grid Group Of Short Term Prediction Based On DKF Algorithm

Due to the environmental pressure and energy crisis and the policy support of all countries, renewable clean energy has gradually become a hot topic in the world and the use of renewable clean energy power generation technology is more and more mature. Micro power grid is becoming a new energy supply mode at the end of power,whose investment main body also gradually tend to be more widely scattered,personal, and the community system formed by multiple adjacent micro grids which exists power interchange between each other is called micro power grid group.However in grid connected state, the distributed generation within each sub micro grid of micro grid groups exists random, intermittent,volatility and problems of smooth load on the user side, which makes great difficulties and challenges for the distribution network of micro grid groups in the power dispatching and operation.Therefore, the total power demand forecasting in the micro grid group is a very good solution.Therefore, the paper researches from the basic elements and characteristics of micro grid group, and the distributed prediction idea is added in the existing prediction algorithms.Gathering the foundation theory of network graph, graph theory model of the micro grid group is established. According to the graph theory model of micro power grid group, the consistency of multi-agent algorithm is introduced. Finally the consistency algorithm and the traditional Kalman filter prediction algorithm effectively are set up and a distributed Kalman filter power prediction algorithm(Distributed Kalman Filtering, DKF) is put forward which is used to predict the total short-term power grid demand of micropower grid group contained multiple child micro power grid.This kind of forecasting method can replace the high cost of traditional centralized forecasting method which needs high communication requirements, long online computation time, good quality communication devices.Finally through the experiment simulation,it can be found that the error accuracy of the distributed kalman filter prediction algorithm is better than that of kalman filtering which usually used to predict the power in a traditional centralized mode.