Research On Short Time Electric Load Forecasting In Power System

Electric load forecasting is the foundation of power system operation and control and also it's the foundation of power market. Accuracy real time reliability intelligence are the new request of electric load forecasting in power market. Study higher accuracy and higher calculation speed of short time electric load forecasting models are the mainly research in this thesis.Electric load character is deeply discussed in this thesis. Reaffirming the regularity of electric load and the indispensable factor, which play an effecting role.These bad date can be sussed out and eliminated by adjusting the maximum of modulus and by processing the wavelet decomposed detail signals by soft threshold based on wavelet analysis and singularity theory. In doing so,we can get the real historical information reflecting the regularity of the forecast and testify the effectiveness of the methods by simulating examples.The selection of forecasting method is adopted in this thesis:The electric load is decomposed into high frequence electric load and low frequence electric load .Artificial neural networks model is established to forecast low frequence electric load in favor of import weather -, date type that relationship with low frequence electric load. Forecasting high frequence load by time sequence method.so that ,forecasting precision and modeling efficiency can be unified.Three short term electric load forecasting models are put forward based on this thinking: one hour electric load forecasting model one day electric load forecasting model and local model .The feasibility of models are testified by simulation examples.Time sequence model is established to forecast high frequence electric load decomposed by wavelet. Time sequence model identifing methods are detailedly discussed in this paper. The fact that, High frequence electric load date are more stable after eliminating bad date by wavelet is also testified by the simulation example.Import of the newest information to improve forecasting precision adopting nolinear programming method in restricted condition is advanced in this thesis. The feasibility of model is testified by simulation example.