Interval Energy Disaggregation for Solar, HVAC, and EV Loads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Utility bills do not provide customers with detailed information on energy consumption by specific appliances or loads, and utility companies lack insights into customer behaviors and load types, leading to inefficiencies in managing energy usage and grid management.
Innovation Solution
Utilizing machine learning models to disaggregate energy consumption data from advanced metering infrastructure, including solar, HVAC, and EV consumption, by optimizing solar panel orientation and processing weather and aggregate data to provide detailed energy usage insights to customers and utilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional utility billing systems are used, then utility companies can collect aggregate energy consumption data, but customers cannot obtain detailed information about energy consumption by specific appliances or loads
Solution Approach 1:
The patent segments aggregate energy consumption data into individual appliance-level consumption data using machine learning models. The system divides the overall energy consumption signal into distinct components attributable to different appliances (HVAC, water heater, refrigerator, etc.), enabling customers to see detailed consumption information without requiring complex hardware modifications at the appliance level.
Solution Approach 2:
The patent introduces an intermediary data processing system that sits between the utility meter and the customer. This intermediary uses machine learning algorithms to analyze aggregate consumption data and infer individual appliance usage patterns, acting as a mediator that transforms raw aggregate data into actionable insights without requiring direct connections to each appliance.
2Loss of information
If utility companies implement detailed monitoring systems to obtain customer consumption information, then they can gain insights into customer behaviors and load types, but the system complexity and cost increase
Solution Approach 1:
The patent enables the data processing system to perform self-service by using machine learning models that automatically analyze and disaggregate consumption data without requiring manual intervention or complex configuration. The system trains models on historical data and autonomously identifies appliance-level patterns, reducing the need for utility company personnel to manually configure monitoring systems.
Solution Approach 2:
The patent changes the parameters of the data analysis process by using machine learning models that can dynamically adjust to different consumption patterns. The system transforms fixed-structure monitoring into a flexible parameter-based approach where the model learns optimal disaggregation parameters from the data itself, enabling detailed insights without proportionally increasing system complexity.
3Measurement precision
If machine learning models are used for energy disaggregation, then detailed appliance-level consumption data can be obtained from aggregate data, but computational resources and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on historical consumption data before actual disaggregation is needed. The system performs offline model training and validation using past utility billing data, so that when real-time disaggregation is required, the pre-trained models can quickly process new data with minimal computational overhead, reducing actual processing time while maintaining high precision.
4Measurement precision
If solar energy production is estimated and removed from aggregate data, then accurate non-solar consumption data can be obtained, but additional data processing steps are required
Solution Approach 1:
The patent merges solar production estimation with the overall energy disaggregation process by integrating it into the same machine learning framework. Instead of treating solar estimation as a separate preprocessing step, the system combines solar panel characteristics, weather data, and consumption data into a unified model that simultaneously estimates solar production and disaggregates non-solar consumption, simplifying the overall workflow while maintaining accuracy.
Data Source
AI summary
Systems and methods for interval energy disaggregation utilizing machine learning are provided. An example method includes obtaining and cleaning weather data and customer data, and based on the customer data, determining whether a structure utilizes solar energy. If the structure utilizes solar energy, the overall aggregate data is disaggregated by using and training a first machine learning model, to ultimately produce a predicted solar production. The predicted solar production is extracted from an overall aggregate data of the structure. Then, the method continues with disaggregating the non-solar aggregate data by utilizing at least a second ML model, to produce disaggregated data relating to at least one of AC energy consumption and EV (electric vehicle) consumption. The disaggregated data is provided to the customer or utility. The disaggregated data comprises at least one of the predicted solar production, the AC energy consumption, and the EV consumption of the structure.