Research On Optimal Dispatching Of Community Microgrid Based On Electricity Price Incentive

In the power system,smart grid uses advanced Internet of things technology and communication technology to achieve flexible,efficient,safe and reliable goals,and provide users with high-quality services.As an important part of smart grid,community microgrid is faced with the challenge of intermittent and uncertain renewable energy generation in the process of maintaining the power balance of microgrid.The key to ensure the efficient and economic operation of community microgrid is to develop an effective optimal scheduling strategy.In view of this,this paper studies the optimal scheduling problem of single community microgrid and multi microgrid.The main contents are listed follows:For the problem of power system user load forecasting.First,analyze the regularity of user load changes,and give error evaluation standards.Then an adaptive BP neural network load forecasting model is proposed.Finally,the validity of the prediction model is verified by numerical experiments and error analysis.For the problem of optimal dispatching of a single community microgrid,a community microgrid optimization model based on electricity price incentives is established.Established and analyzed the Markov decision process for optimal scheduling problems,which provides a theoretical basis for the reinforcement learning algorithm to solve such problems.For the problem that the single agent Q-learning algorithm converges slowly and is easy to fall into local optimal solutions,an asynchronous Q-learning algorithm with multiple agents parallel computing is proposed,and the asynchronous Q-learning algorithm is verified through three microgrids numerical experiments on the electricity consumption behavior of three different users.The effectiveness of the Q-learning algorithm.For the joint optimization problem of multiple micro-grids,an optimization model of micro-grid group based on electricity price incentives is established to improve the traditional electricity market transaction model,and propose a transaction model for microgrids peer-to-peer transactions.For the dimensional disaster of multi-threaded Qlearning algorithm when facing complex problems,an ant colony-reinforcement learning algorithm combining ant colony algorithm and asynchronous Q-learning algorithm is proposed.Based on the weibull distribution and β distribution,the wind and light intensity are fitted and predicted.Finally,the effectiveness of the algorithm is verified through numerical simulation experiments of four adjacent micro-grids forming a micro-grid group.