Forecasting Customer Baseline Load Based on Similar Electricity Usage Patterns

In incentive-based demand response programs, compensation is calculated and paid to participants who adjust their electricity usage as requested. This is based on the difference between the Customer Baseline Load (CBL) the estimated electricity usage without participating in the program and the actual load during the event. Accurate CBL calculation ensures fair compensation, incentivizing participation. This research develops a model to forecast CBL using K-Means clustering, incorporating Dynamic Time Warping (DTW) as a distance metric. The centroid from each cluster represents the CBL for individual customers within the cluster. This approach is compared to the traditional "Max 4/5" method. Results show the centroid-based CBL calculation aligns more closely with actual electricity usage, with a relative root mean square error (RRMSE) of 0.175 compared to 0.312 for the Max 4/5 method.