A region-based commercial building load forecasting method based on electricity consumption behavior patterns is proposed to fully exploit the fine-grained load data collected by smart meters and to improve the accuracy of regional commercial building load forecasting. Firstly, the mean-variance normalization method is used to standardize the collected load data. Then, to extract different electricity consumption behavior patterns in regional commercial building loads, the elbow method is used to determine the number of clusters, followed by k-Shape clustering. Next, an improved Informer model is introduced to address the challenge of predicting large-scale commercial building loads within a region, which often requires significant memory resources while struggling to achieve high accuracy. This model uses clustering algorithms to identify commercial buildings with similar electricity consumption patterns and accounts for the impact of anomalous load data collected by smart meters on the training results. The proposed model effectively addresses the problem of low accuracy in predicting load for large commercial buildings. Finally, experiments are conducted using commercial building loads in California.The experimental results demonstrate the effectiveness of our proposed method in the improvement of the accuracy of regional commercial building load forecasting.