Research On Model And Strategy Of Micro-grid Ultra-short Term Load Forecasting

As the energy scarcity and environment pollution have been worsened, comprehensivedevelopment and effective use of new energy resources is imperative under the circumstance.Micro-grid construction is one of the efficient paths that can use new energy sufficiently andoptimize energy construction. In order to make sure the micro-grid run effectively, theimportant basis of micro-grid optimize running and energy management decision is accurateload forecasting. Therefore, this thesis studies on the question that how to improve theaccuracy of a middle or short exceed time load forecasting. It has important theoreticalsignificant and practical value for micro-grid system optimal operation.According to the problem of forecasting precision reduced by the ultra-short term loadforecasting model based on error judgment in Micro-grid, an improved method of ultra-shortterm load forecasting based on a difference degree of load sample is proposed in this paperafter exhaustive research of load sample. Based on improved operating strategy of theforecasting model and combined with regression analysis, a difference degree of load sampleis defined to adjust forecasting model in order to adapt to load changing automatically andenhance precision of forecasting.It selects the quarterly data from load of the micro-grid on island as samples. The datasamples are processed by identifying and completion and wavelet based de-noising to avoidthe interference of abnormal data. The characteristic of load and weather data samples isanalyzed to get a full grasp on relative demographic variables. The ultra-short term loadforecasting model proposed based on a difference degree of load sample and RBF neuralnetwork in micro-grid is created to observe the feasibility. The experimental results indicatethat compared to the forecasting model based on error judgment, the improved ultra-shortterm load forecasting model can enhance effectively the precision of forecasting.