Microgrid Based On Short-Term Load Forecasting Research On Optimal Dispatching

The main content of this paper is the optimal scheduling of the microgrid.By establishing the model of the microgrid and its components,a prediction method of STI-CNN is proposed.The input data is accurately clustered using the two-branch deep network model,and the STI-CNN method is used for load forecasting;A general model for multi-objective dynamic optimal scheduling of microgrid is constructed to realize the dual optimization objectives of economy and environment of microgrid system.The NSGA-Ⅱ algorithm is improved,which effectively improves the convergence performance of the algorithm and improves the distribution characteristics of Pareto frontier.Finally,taking the grid-connected mode and island mode in the microgrid as examples,the research on the multi-objective optimal scheduling of the microgrid is carried out,and the corresponding constraints and objective functions of the two modes are given to verify the applicability of the improved NSGA-Ⅱ algorithm in the microgrid scheduling.(1)The knowledge points related to short-term load forecasting and the theoretical concepts related to convolutional neural networks are introduced,then the load forecasting influence factors are analysed and the evaluation indexes for forecasting are introduced.Finally,the load sequences are converted into several load image data and the STI-CNN method is applied to short-term forecasting of customer-side load of microgrid and compared with other forecasting models and SVM methods,the experimental results show that the STI-CNN method has excellent performance in different forecasting indexes,the forecasting time used is shorter and has higher accuracy.(2)A generic model for multi-objective dynamic optimisation of dispatch is established with an independent system simulation module and an operation optimisation module as the core.The simulation module uses the energy model to evaluate the economic and environmental indicators of the system dispatch scheme,and the operation optimisation module uses the multi-objective genetic algorithm NSGA-Ⅱ and combines the evaluation results of the dispatch of the simulation module to carry out optimisation.In the improved NSGA-Ⅱ algorithm,initial point bootstrap techniques and de-duplication operations are introduced,which greatly improve the convergence performance of the algorithm and the distribution characteristics of the Pareto front.The proposed model and method are applied to the daily optimal scheduling of a typical wind power microgrid system,and the effectiveness of the proposed model and improved algorithm is verified.(3)A multi-objective optimal dispatching model corresponding to the grid-connected mode and the islanding mode in the microgrid system is developed,and the working strategies of the two models are presented separately.By using the experimental results data from the load forecasting in Chapter 3 and considering the time-sharing tariff mechanism,the improved NSGA-Ⅱ algorithm proposed in Chapter4 is also used in the multi-objective optimal scheduling of the microgrid,compared with the convergence curve of the unimproved NSGA-Ⅱ algorithm,and finally a compromise solution is used to further analyse the output of each microsource in the microgrid system.The experimental results show that the improved NSGA-Ⅱ algorithm provides significant improvements in environmental and economic aspects.