Research On Short-term Load Forecasting Method Based On Hourly Weather Factors

Short-term load forecasting is one of the important parts of power market and its technical support system,it is the presupposition and basis of arranging dispatch plan and deal plan under power market circumstances. The forecast accuracy relates to the efficiency,benefit and security of the entire power system, therefore, how to raise the accuracy of load forecasting has been a hotspot also difficult issue for research workers.Research shows that power load is influenced by many factors such as economy,power-utilization construction,power price,weather, etc. Recently, civilian power consumption grows quickly along with economic growth and people's living standard upturn, the proportion of heating and cooling load in total load increases rapidly. Moreover, it is very sensitive to weather changes,load level vary evidently in case weather conditions changes. The relationship between weather factors and load has raised more and more concerns.In this thesis, firstly, the load and weather characteristics of area A are analyzed, and then a load decomposing model which is suitable to the area is built, the total load is decomposed into base load,weather sensitive load and random load, among which the weather sensitive load is influenced by weather factors greatly.Secondly, the relationship between total load,weather sensitive load and weather factors are studied, the study begins from daily characteristic and hourly weather factors, then daily characteristic weather factors such as daily average temperature,daily maximum humidity and hourly weather factors such as hourly temperature,hourly humidity are selected; As it is always a continuous hot weather for this area in summer, so temperature has an accumulation effect on load, in this thesis, the accumulative effect is also analyzed briefly; In addition, a method for similar days selecting based on load characteristics and weather conditions is proposed, it can avoid the inflexibility of traditional method, and has advantage over that under complex weather conditions of the forecasting day.Finally, on the basis of the studies above, a short-term load forecasting method based on load decomposing and hourly weather factors is proposed in this paper. The base load component is forecasted by gray system GM(1,1) model which is usually used in medium and long term load forecasting, the weather sensitive load is forecasted by a BP neural network which is optimized by Levernberg-Marquardt algorithm. The availability of the method is proved by forecasting simulation of this area in summer 2007, it can improve the forecasting speed and accuracy remarkably.