Comprehensive Evaluation of Electric Power Prediction Models Based on D-S Evidence Theory Combined with Multiple Accuracy Indicators

A comprehensive evaluation method of electric power prediction models using multiple accuracy indicators is proposed. To obtain the preferred models, this paper selects a number of accuracy indicators that can reflect the accuracy of single-point prediction and the correlation of predicted data, and carries out a comprehensive evaluation. First, according to Dempster-Shafer (D-S) evidence theory, a new accuracy indicator based on the relative error (RE) is proposed to solve the problem that RE is inconsistent with other indicators in the quantity of evaluation values and cannot be adopted at the same time. Next, a new dimensionless method is proposed, which combines the efficiency coefficient method with the extreme value method to unify the accuracy indicator into a dimensionless positive indicator, to avoid the conflict between pieces of evidence caused by the minimum value of zero. On this basis, the evidence fusion is used to obtain the comprehensive evaluation value of each model. Then, the principle and the process of consistency checking of the proposed method using the entropy method and the linear combination formula are described. Finally, the effectiveness and the superiority of the proposed method are validated by an illustrative instance.

Automated summary

A comprehensive evaluation method for electric power prediction models using multiple accuracy indicators is proposed in this paper. The method selects accuracy indicators that reflect both single-point prediction accuracy and correlation of predicted data, and conducts a comprehensive evaluation. The paper introduces a new accuracy indicator based on relative error and a dimensionless method to unify the accuracy indicator. The effectiveness and superiority of the proposed method are validated through an illustrative instance.