A Novel Evolution Kalman Filter Algorithm for Short-Term Climate Prediction

With the increasing integration of grid-connected photovoltaic (PV) power generation system, the short-term climate prediction is becoming critical. In this paper, we propose a novel evolution Kalman filter (ELKF) based short-term climate prediction algorithm, which combines the advantages of statistical and dynamic methods. We first establish the Kalman forecast recursive model, and then apply the genetic algorithm (GA) to optimize the transfer matrix which reflects the interaction relationship of prediction factors in a Kalman filter. The experiment to predict average sunshine hours and daily temperature for a certain place is conducted. The simulation results demonstrate that, compared with the traditional Kalman filter, our approach enhances the prediction accuracy for average sunshine hours within 1 h by 16.5% and for average daily temperature within 1C by 5.8%.