Short-Term Load Forecasting And Distributed Dynamic Economic Dispatch In Microgrids

With more and more renewable distributed generators and energy storage units penetrating in microgrids,the operating environment of microgrids is becoming more and more complex,which brings greater challenges to the safe and economic operation of microgrids.As a research hotspot,the dynamic economic dispatch in microgrids directly affects the reliability and economy during the operation of microgrids.At the same time,the dynamic economic dispatch is inseparable from the data support of the short-term load forecasting.Therefore,the short-term load forecasting and the dynamic economic dispatch problems in microgrids have extremely important research significance.Accordingly,the short-term load forecasting and the dynamic economic dispatch problems in microgrids are studied in this paper.The specific contents are described as follows:For the short-term load forecasting in microgrids,the regularized kernel extreme learning machine is built as a basic predictive model,and a novel improved firefly algorithm is proposed to optimize the parameters of the regularized kernel extreme learning machine model.Accordingly,a short-term load forecasting model based on the improved firefly algorithm and the regularized kernel extreme learning machine is proposed.Finally,simulation studies demonstrate the effectiveness of the prediction model proposed in this paper.Based on the data support of the short-term load forecasting,the dynamic economic dispatch problem is described as a convex optimization model in this paper.In our model,not only the conventional equality and inequality constraints,energy storage units’ constraints and power flow constraints are considered,but also operational costs of energy storage units and the power price of the main grid are considered in the total costs.To solve this problem,a fully distributed algorithm based on the alternating direction method of multipliers,the projected gradient method and the average consensus algorithm is proposed.The proposed algorithm can obtain the optimal output power settings of the energy storage units,distributed generators and the main grid for different demand loads with different initial states.In the proposed algorithm,firstly,the dispatch problem is decomposed into multiple subproblems on the basis of the alternating direction method of multipliers;then the subproblems can be solved according to the projected gradient method and the average consensus algorithm.In order to apply the projected gradient method,every bus is equipped with a bus agent which is responsible for the information communication and the output power optimization calculation.Simulation studies demonstrate the effectiveness of the proposed algorithm.