Precision and Accuracy Co-Optimization-Based Demand Response Baseline Load Estimation Using Bidirectional Data

Accurate baseline load estimation is critical for the compensation settlement of incentive-based demand response (DR). Baseline load estimation is different from load forecasting in evaluation metric and time scale. In terms of evaluation metric, Precision (i.e., the closeness between the estimated and the actual value) and Accuracy (i.e., lack of bias) are two aspects for assessing the performance of an estimation. Load forecasting usually focuses more on Precision and ignores Accuracy, while these two aspects are both important for baseline load estimation. In terms of time scale, load forecasting must be conducted in advance while baseline load estimation is an ex-post problem. However, baseline load estimation is usually formulated as a load forecasting problem in most existing studies, which can be further improved using the load data before and after the DR event. This paper proposes a Precision and Accuracy co-optimization based baseline load estimation framework to improve its performance. The proposed framework uses the load data not only before but also after the DR event day as the input features. An optimal weighted ensemble method is proposed to combine these input features, in which both Precision and Accuracy are considered into the objective function. The effectiveness and superiority of the proposed framework have been verified on a real smart meter dataset.