Research On Short-term Load Forecasting And Multi-objective Optimal Dispatching Of Microgrid

In recent years,due to the rapid development of society,fossil energy has become scarce and environmental pollution has become increasingly problem.In order to cope with this phenomenon and meet the rising power demand,clean energy power generation has become a research target.Using clean energy for power generation will not only alleviate the energy crisis,but also decline environmental pressure.Distributed power generation is a technical method to effectively use clean energy for power generation.The microgrid combines various distributed power sources for centralized power supply,which can not only fully utilize the advantages of distributed power generation and maximize the use of renewable energy,but also provide safe and reliable power supply.Optimal dispatch of microgrid is to use the EMS energy management system to control the output power of each power supply under the premise of ensuring safe and reliable power supply.Therefore,as to achieve the purpose of economic operation of the microgrid.Load forecasting is an indispensable part of micro-grid optimal dispatching.Only by obtaining future load data in advance,the output power of micro-power supply be effectively optimized dispatched.Although the traditional BP neural network can solve the problem of load forecasting,it has the disadvantage of poor accuracy and easy to fall into local extreme value.Thus,this paper use genetic algorithms to optimize the BP neural network parameters,establish a GA-BP load prediction model,using the actual data of the power grid to train the neural network and perform load prediction and compare it with the single BP neural network prediction results.Comparison of the results shows that the GA-BP model predicts the results more accurately than traditional ways.Considering the conditions such as power balance constraints and controllable unit output constraints,the micro-grid multi-objective optimal dispatch model is established with the micro-grid operating cost and the minimum load loss probability as the objective functions.A multi-objective particle swarm optimization algorithm isused to solve the scheduling model,and the Pareto frontier graph is obtained.Taking the lowest probability of load loss as the priority selection condition,the scheduling results of each distributed power source are obtained.From the scheduling results,the microgrid operates at a lower cost and the clean energy is fully utilized under the condition that the user's power consumption loss probability is low,which proves the economy and practicability of the scheduling model.