Research On Network Supply Of Abundant Small Hydropower Load Forecast Considering Meteorological Factors

Power system load forecasting is an important basis for dispatching plan,operation mode arrangement and ispatching operation,and also an significant guarantee for safe and stable operation.Load forecasting can enhance the ability of management in large customer and local power plants,which not only meets the requirements of lean management,but also ensures the safe,stable,quality and economic operation of the power grid.For municipal power departments with small capacity and large load fluctuation,accurate prediction of network supply load is of great importance for unit commitment,economic dispatch and optimal power flow calculation.The changeable meteorological factors have great influence on network supply load forecasting in the regions with abundent hydropower.Under certain weather conditions,such as the changing air temperature,the electricity load and consumption will rise sharply.When the weather becomes cold or hot,there will be a large number of heating or temperature-lowering load put into operation;When the average temperature continues to be too high or too low,the daily load will have a great change compared with the same date in previous years.What's more,other meteorological factors,such as rainfall,will directly affect the power output of small hydropower,and indirectly affect the network supply load.Based on the above idea,The two-phase reduction method of network supply load forecasting considering meteorological factors is proposed in this paper.In this method,the network supply load is decomposed into small hydro power output and regional load,using the high-adaptability forecasting method separately,and then reduct into network supply load.The BP neural network method considering fuzzy clustering is used to predict the power output of small hydropower plants with strong meteorological factors.Firstly,fuzzy clustering of training samples is carried out according to the historical operation data of small hydropower plants which have the stronge relativity in forecasting targets,with large correlation with prediction targets,and samples with the same characteristics are classified into same class,so as to enhance generalization ability of the model and improve the accuracy of the method.Then,the neural network model is established for each sample.The meteorological data and the historical operation data of small hydropower are taken as model input,and the output of small hydropower as the output of the model.As for the regional load which does not exist too much lag effect and cumulative effect,the grey Elman neural network method is used to forecast.In order to accelerate the convergence speed,the greyElman neural network is improved using genetic algorithm.A genetic grey Elman neural network forecasting method is proposed.Aiming at the meteorological factors such as temperature and rainfall,a correction model of regional load forecasting result is put forward.At last,based on practical data in two cities,the small hydro power output and regional load is forecasted separatly,and the network supply load is reducted using the two-phase reductiog method.The results reveal the applicability and progressiveness of this method.