With the rapid growth of data in the energy and public service industries, traditional billing and metering systems face challenges in processing massive amounts of user usage, ensuring accurate itemized billing, and identifying anomalies. These challenges require extensive manual intervention and pose inadequate risk control. To address these challenges, this paper proposes an intelligent billing solution based on the integration of artificial intelligence and the Internet of Things (IoT). This solution uses a Weighted Long Short-Term Memory Network (Weighted-LSTM) model, combined with user profile features, rolling statistics, and temporal features, to accurately predict itemized bills. It also utilizes a clustering algorithm and a Gradient Boosted Decision Tree (GBDT) + Deep Neural Networks (DNNs) classifier to intelligently identify fraudulent or erroneous bills. Based on these predictions, a reinforcement learning environment is constructed to automatically schedule bill generation and resource allocation to cope with peak loads and unexpected anomalies. The system collects metering data in real time through the IoT and feeds prediction and detection results back to the scheduling module, achieving closed-loop optimization. The results of offline simulation and small-scale grayscale experiments show that in the face of peak bursts, the system triggers scheduling ratio as high as 34.0%, the average peak suppression rate reaches 28.5%, and the price adjustment strategy is activated 112 times, effectively alleviating the load pressure and bringing a slight positive improvement in user satisfaction (+0.08 points). At the same time, resource costs decrease by 6.2%, indicating that the scheduling mechanism is both efficient and economical in peak scenarios.
Tan et al. (2026) studied this question.