Abstract:
With the steady advancement of the agent power purchase business of the grid company, the agent power purchase business system has been gradually improved, and the accurate prediction of power consumption of agent power purchase users lays the foundation for guaranteeing the safe and stable power supply. Therefore, this paper constructs an adaptive weight combination model to assign weights to the calibration results of different calibration methods, so as to improve the accuracy of the calibration results. First of all, the prediction business deviation calibration process framework is constructed to determine the agent power purchase prediction business calibration process. Then, the quartile mapping method, incremental change method and support vector regression (SVR) are selected to calibrate the prediction results, and the calibration results of different methods at the same latitude are obtained. Finally, a genetic algorithm- technique for order preference by similarity to ideal solution (GA-TOPSIS) model is established to optimize the accuracy and stability of the calibration results, and the optimal weight combinations of the different calibration methods are selected. The optimal weight combinations of different checking methods are selected. The test results show that the prediction accuracy and precision are significantly improved after the weight combination of the calibration methods, compared with the initial prediction value and the result after the single calibration method.