Developments And Improvements To Load Forecasting Subsystem Of Computer Aided Decision-making System For Urban Power Distribution Network Planning

Load forecasting is the basis of urban power distribution network planning.Based on Load Forecasting Subsystem of Computer Aided Decision-Making Systemfor Urban Power Distribution Network Planning (CNP4.0) developed by TianjinUniversity, this paper makes a series of developments and improvements.First, a methodology of analysis and prediction on urban electric load saturationcharacteristic based on growth curve is proposed. With the approach of growth curvemodel to urban electric load, the saturation value, the saturated year, the fasted loadgrowth year and the load development level of the current year are achieved. Bycomparison of several growth curve models and parameter-solving methods, aconclusion is made that the four-parameter model is the most suitable growth curvemodel for electric saturation analysis.Second, an integrated prediction method based on Wavelet Analysis-GrayModel-Artificial Neural Network (WGG method) is proposed according to the changeregularity of monthly load. WGG separates the long-term trend component and periodcomponent by wavelet analysis to monthly load series. Gray Model and ArtificialNeural Network model are used respectively to forecast different components. Thismethod forecasts the monthly load development very well by comprehensiveconsideration of load long-term trend and period increasing trend.Third, a practical spatial load forecasting method based on main feeder isestablished by analyzing the current spatial load forecasting methods. This methodovercomes the disadvantages of current spatial load forecasting methods which arecomplicated and lowly automated. By dividing power supply areas into sub areas andusing main feeders as elementary units, this method forecasts the feeder load ofobjective year with analysis to regulation of normal trend and special events of load.Through low-to-up load superposition of substations and power supply areas, the finalforecasting procedure is completed. Besides, a discussion of improved feeder loadforecasting models and the features of them is carried out, with a preliminaryconclusion on the applicable situations of each model.Finally, a design of functions, user interfaces, data interfaces and databases of thealgorithms above is performed successively. These algorithms are integrated into Load Forecasting Subsystem of CNP4.0. The improved functions of that subsystemtake on very good performance in practical planning projects.