Cloud Energy Metering With ML Anomaly Detection

The patent introduces a cloud-based platform as an intermediary between the power distribution network and end-users. This platform includes data reception modules, machine learning algorithms, and user interface modules that collectively handle data processing, anomaly detection, and visualization. By placing the complex processing logic in the cloud rather than at the network edge or user premises, the system achieves high data processing efficiency and sophisticated anomaly detection while keeping the local infrastructure simple.

The patent replaces conventional mechanical data analysis methods with machine learning algorithms. Specifically, it employs neural network-based anomaly detection algorithms that can automatically learn complex patterns in energy usage data without requiring explicit programming of detection rules. This substitution enables high-precision anomaly detection by leveraging the pattern recognition capabilities of machine learning models, while the cloud-based deployment abstracts the algorithmic complexity from the user perspective.

The patent extracts the complex machine learning processing functions into a separate cloud-based module, isolating it from the user interface layer. The user interface module provides simplified visualizations, alerts, and energy data access without exposing the underlying algorithmic complexity. This separation allows the system to implement sophisticated anomaly detection algorithms while maintaining an intuitive and easy-to-use interface for end-users.

The patent implements preliminary action by pre-training machine learning models on historical energy usage data before deployment. The anomaly detection algorithms are pre-configured with learned patterns of normal and abnormal energy consumption. When real-time data arrives, the pre-trained models can quickly compare incoming data against established patterns and generate anomaly scores without requiring extensive real-time computation, thus achieving real-time detection capability with reduced processing time.