New Energy Power Forecasting Model Performance Evaluation Method Based on New Energy Station Resource Complexity Quantification

Gang Li, Nan Li, Ling Hao, Li He, Wenjing Yang, Haonan Dai, Fei Wang, Fei Xu, Dong Lv, Yuan Wang, Yuzhi Liu, Mengjiao Chen

Abstract

The accuracy of new energy power forecasting is vital for new energy power generation enterprise (NEPGE) in both avoiding the forecasting deviation penalty and improving the formulation of market trading strategy. Due to the different geographical and climatic environments in different stations that equipped with different NEPFSs, the resource complexity of station which affects the performance of NEPFS is also different. It leads to NEPGE's inability to directly evaluate the performance of different NEPFSs through the accuracy of forecasting results, because it is not clear whether the lower accuracy is caused by the poor performance of NEPFS or the high complexity of station resources. However, current research does not involve how to quantify the complexity of resources. The performance of the forecasting system is mostly directly quantified by the accuracy evaluation index such as root mean square error. To address these issues, this paper first introduces a dynamic framework for assessing resource complexity, featuring seven indicators spanning holistic and local perspectives, to facilitate scientifically quantified complexity assessments of new stations. Second, considering the real-time dynamic change of resource complexity, this paper puts forward an evaluation method of NEPFS performance, called complexity-adjusted performance assessment indicator (CAFA). Case studies based on actual operational data demonstrate that the proposed framework effectively quantifies the complexity of power station resources and accurately evaluates the performance, which can provide valuable guidance for NEPGE to choose the most outstanding NEPFS for new energy power stations.

Source: semanticscholar

Read the AI summary and key takeaways for traders on WOBR Quant Research.