| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 101 === The microgrid system load forecasting can help to provide better planning, scheduling and control of the system. It can make the entire operation of the microgrid more economic and reliable. Therefore, by improving the accuracy of the prediction model, the microgrid operation stability and security will be enhanced. There are two model of load forecasting, very short term model and short term model. The very short term model forecast the load from several minutes ahead until 1 hour ahead. It can be used for security control, prevention and emergency condition. Short term model that forecast from several hours ahead until several days ahead can be used to plan the power dispatch. The number of load forecasting research in microgrid system is still much fewer than the research in the large scale power system. The total load changes in microgrid system is affected by the changes of each load element, so it is less likely to achieve an accurate prediction. Based on this, algorithm that is applicable to very short term and short term load forecasting in micro grid is, analizing the similarity of the electrical load curve in the historical data. It is followed by using Back Propagation Neural. Network (BPNN) to learn the similar historical load data that reflect the relationship between the input and the output of the relevant variables. The prediction model can be divided every five minutes, every hour and every twenty-four hours of load forecast. Finally, the actual load time series is used as validation, to determine the availability and accuracy of the proposed method. The simulation results show that the proposed method can provide the the microgrid system the online real-time control and scheduling planning reference.
|
|---|
