Building Energy Estimation Using Dilated CNN Time-Series Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing energy management systems struggle to accurately analyze and reduce energy consumption in large non-residential buildings due to complex energy consumption patterns and the lack of tools that do not require expert oversight.
Innovation Solution
A dilated convolutional neural network architecture is used to predict energy consumption patterns in buildings, incorporating time-series data from smart meters and environmental factors, allowing for real-time energy consumption estimation and identification of causal reductions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If smart meters are used to measure energy consumption, then energy load measurement becomes easier and more widely available, but the data remains at a single gross level making it difficult to identify causal factors for energy reduction
Solution Approach 1:
The patent segments the single gross energy consumption measurement into multiple detailed energy components by using machine learning models to disaggregate the data. This allows the system to break down total energy usage into individual appliance and system-level consumption patterns, enabling identification of specific causal factors for energy reduction while maintaining the ease of measurement provided by smart meters.
2Measurement precision
If machine learning systems are used to forecast energy load demand, then future energy patterns can be predicted, but the systems require complex analysis and expert oversight to implement effectively
Solution Approach 1:
The patent implements self-service by enabling building controllers to autonomously analyze energy consumption data and identify causal factors for energy reduction without requiring continuous expert oversight. The machine learning system automatically processes smart meter data, segments it into meaningful components, and provides actionable insights that building operators can implement independently, thereby maintaining high prediction accuracy while reducing implementation complexity.
3Loss of energy
If detailed analysis of energy consumption patterns is performed to identify reduction opportunities, then energy efficiency can be improved, but the analysis becomes increasingly complex and difficult to implement
Solution Approach 1:
The patent introduces an intermediary machine learning system that acts as a mediator between raw smart meter data and actionable energy reduction insights. This intermediary automatically performs the complex segmentation and analysis of energy consumption patterns, translating gross energy data into detailed component-level information without requiring users to directly handle the complexity of the analysis themselves.
Data Source
AI summary
Systems and methods for estimating energy consumption data for a building. A system for estimating an energy use of a building uses a dilated convolutional neural network architecture to receive time-series data for the building and to predict one or more time-series data points representing an estimated energy consumption for the building. A method for estimating an energy use of a building includes obtaining time-series data for the building, providing the time-series data as input to a dilated convolutional neural network architecture, and predicting one or more time-series data points representing an estimated energy consumption for the building using the dilated convolutional neural network architecture. The systems and methods may be used to help users and building controllers reduce energy use within a building.