Energy Usage Detection Models for Device Presence Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Disaggregating energy usage from various devices within a source location, such as households, is challenging due to the variety of devices and limited availability of training data, leading to inefficiencies in grid planning and energy management.
Innovation Solution
Utilizing trained machine learning models, particularly deep learning schemes, to predict the presence of target device energy usage by processing total energy usage data from metering infrastructure, leveraging limited training data and improving generalization across different locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional metering devices are used to monitor energy usage, then energy consumption data can be collected, but device-specific energy usage cannot be reliably discovered due to the variety of household devices
Solution Approach 1:
The patent segments the overall energy usage signal into individual device-specific components by analyzing temporal patterns, power consumption signatures, and usage behaviors. The machine learning model divides the complex disaggregation problem into detectable features for each device type, enabling precise identification of device-specific energy usage from aggregated metering data.
Solution Approach 2:
The patent transforms the energy usage detection problem by changing parameters such as time resolution, power consumption thresholds, and usage pattern characteristics. By analyzing energy data at multiple granularities and transforming temporal patterns into distinctive signatures, the system adapts to detect diverse device types without requiring device-specific hardware.
2Reliability
If machine learning models are trained with limited training data, then model development is feasible, but prediction accuracy decreases due to insufficient learning samples
Solution Approach 1:
The patent creates synthetic copies of training data by generating artificial energy usage patterns that mimic real device behaviors. Through data augmentation techniques, the system replicates and transforms existing training samples to create diverse synthetic datasets, effectively increasing the training data volume without requiring additional physical measurement campaigns.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction to prepare training data before model training. By pre-processing energy usage data to extract distinctive device signatures and temporal patterns, the system maximizes the information content of limited training samples, enabling more effective learning with fewer data points.
3Measurement precision
If complex deep learning schemes are used to improve prediction accuracy, then target device energy usage can be discovered more accurately, but computational requirements and resource consumption increase
Solution Approach 1:
The patent extracts and focuses on the most discriminative features from energy usage data, such as temporal patterns, power consumption signatures, and usage behaviors. By taking out only the essential features needed for device identification rather than processing complete raw data streams, the system reduces computational requirements while maintaining prediction accuracy.
Solution Approach 2:
The patent applies partial processing to energy data by analyzing only the most informative time windows and frequency components. Rather than processing entire datasets with full deep learning models, the system selectively applies computational resources to the most critical analysis stages, reducing overall resource consumption while achieving sufficient prediction accuracy.
Data Source
AI summary
Embodiments generate machine learning predictions to discover target device energy usage. One or more trained machine learning models configured to discover target device energy usage from source location energy usage can be stored. Multiple instances of source location energy usage over a period of time can be received for a given source location. Using the trained machine learning model, multiple discovery predictions for the received instances of source location energy usage can be generated, the discovery predictions comprising a prediction about a presence of target device energy usage within the instances of source location energy usage. And based on the multiple discovery predictions, an overall prediction about a presence of target device energy usage within the given source location's energy usage over the period of time can be generated.