Research On Wind Power Prediction And Power Generation Scheduling Considering Fuel Inventory In Electric Power Enterprises

Green energy development and power generation scheduling are the key problems in the management of energy saving and emission reduction in electric power enterprises.To solve the problem that wind power is forecasted in low accuracy and difficult to connect to the grid on a large scale,the wind power prediction problem is investigated in this thesis.The main task of the wind power prediction problem is to predict wind power in the future using historical wind power data and meteorological data.To coordinate between the power generation scheduling and the fuel inventory management,the integrated optimization problem of power generation scheduling and fuel inventory is investigated.The main task is to decide the generation level and the purchase of the fuel to maximize the profit of the power enterprise,including unit operation constraints and fuel inventory constraints under consideration.An accurate prediction of wind power and a reasonable power generation scheduling can improve the utilization of wind energy,promote the internal coordination and the overall optimization of the enterprise,and enhance the market competitiveness of the enterprise.For the wind power prediction problem,a dynamic space-time series model is proposed to predict the ultra-short and short term wind power and a dynamic STARMA and XGBoost hybrid model is proposed to predict the long-term wind power.For the integrated optimization problem of power generation scheduling and fuel inventory management,a stochastic mixed integer nonlinear programming model is established and a Benders decomposition algorithm is developed to solve the problem.The main contents are as follows:1)A dynamic STARMA model is proposed to improve the accuracy of the traditional ARIMA model in short-term wind power forecasting.The proposed dynamic STARMA model can reduce the influence of wind direction and other factors on forecasting accuracy.Experimental results verify the effectiveness of the model.2)To overcome the defect that the statistical forecasting method is greatly influenced by the forecasting time scale and shows low accuracy in long-term wind power forecasting,a hybrid model of XGBoost and dynamic STARMA is proposed to forecast wind power.By using the proposed model,the accuracy in long-term wind power prediction is improved and the sensitivity of the model to the forecasting time scale is weakened.3)In accordance with the operation characteristics of high emission and high cost in electric power enterprises and the stochastic volatility of electricity prices,fuel prices and electricity demands in the markets,the integrated optimization problem of power generation scheduling and fuel inventory management is studied.The volatility of prices and demands is described by the stochastic programming method and the considered problem is formulated as a mixed integer nonlinear programming model.4)According to the characteristics of the model,a Benders decomposition algorithm is developed to solve the problem.According to the separation characteristics of the structure of the sub problem,multiple cuts are generated in each iteration to accelerate the convergence of the algorithm.Numerical results show that the Benders decomposition method is superior to the commercial optimization software CPLEX and can obtain a satisfactory solution in a short time.