Short-term Load Forecasting On New Energy Micro-grid

Aimed at the difficulty of using electricity in China’s western parts and thesoutheast coast of the island, combined with the advantage of local resources, it hadbecome a trend to establish stand-alone micro-grid. This thesis using ZhuhaiDong’ao island as a object, launched a series of research work.This thesis introduced the system framework of the stand-alone micro-grid inZhuhai Dong’ao island, and showed all the possibilities of the energy flow,calculated the energy efficiency of different paths. We concluded that theestablishment of energy management system played an essential role in the control ofefficient and effective flow of energy, short-term load forecasting guaranteed therealization of the energy management.This thesis used BP algorithm for short-term load forecasting, analyzed the BPneural network, and applied a three-tier network architecture. We established thesimilarity calculation rules，selected the input and output sample data based on thesimilarity。As the BP algorithm used the gradient descent method to optimize theobjective function, the learning speed is slow, and the local search process was easy tofall into local minima. The thesis also proposed a method to use similar day method toinitialize the weights and the offset value. By analyzed the historical data in theZhuhai Dong’ao island in the MATLAB program, we verified the superiority of theimproved BP algorithm.This thesis also proposed a algorithm of Fruit Fly Optimization AlgorithmSupport Vector Machine (FOASVM), and used it for short-term load forecasting. Fruitfly searched the optimization according to the concentration of the food,automatically fly to the food, hence found the the optimal parameters of SVM and itskernel function. We also established the processes and procedures for this algorithmand analyzed it using MATLAB. We studied the effect that the initial location of thefruit fly, population size and the step value produced on the flight speed and algorithmprediction accuracy, then we were able to find a reasonable initial position, populationsize and step value. We used the historical data in the Zhuhai Dong’ao island toverify the superiority of FOASVM.The stand-alone micro-grid in the Zhuhai Dong’ao island applied three-levelenergy management platform, different level had different function and operating interface, they constituted the entire energy management system by coordinatedoperation. This thesis built the second-level energy management platform of thestand-alone micro-grid in the Zhuhai Dong’ao island, including the choise ofhardware and the implementation of sorfware. We realized the short-term loadforecasting in the second-level platform, including the establishment of the database,weather forecasting systems and load forecasting system. Finally, we sampled severalpredicting outcomes and analyzed them, which proved the practical value of theshort-term load forecasting.