Energy Meter System for Appliance State Identification via Neural Network Disaggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing energy meter systems face challenges in determining the identity and operational state of individual electrical devices connected to a common power source, as exact inference of appliance states from aggregate power measurements is computationally intractable due to conditional dependencies.
Innovation Solution
The energy meter system employs a tractable auxiliary distribution parameterized by a deep neural network, combined with Variational Bayes and Factorial Hidden Markov Models, to perform energy disaggregation by decomposing power signals into component parts representing remote devices, leveraging high-frequency information and temporal patterns to overcome computational complexity and stitching issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exact inference methods are used to determine appliance states from aggregate power measurements, then measurement precision is improved, but computational complexity becomes intractable
Solution Approach 1:
The patent introduces an intermediary neural network model that acts as a mediator between the aggregate power measurements and the appliance state predictions. Instead of directly performing exact inference which is computationally intractable, the system uses this trained neural network intermediary to approximate the inference process, achieving both computational efficiency and accurate appliance state identification.
2Reliability
If Factorial Hidden Markov Models are used to model appliance states, then reliability of state inference is improved, but computational cost becomes prohibitively expensive
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network model offline using Factorial Hidden Markov Models and appliance operation data. This pre-training phase captures the complex relationships between power measurements and appliance states. Once trained, the model can be deployed for real-time inference with minimal computational energy requirements, as the heavy computational work has already been performed during the offline training phase.
3Measurement precision
If high frequency information is exploited for energy disaggregation, then measurement precision is improved, but computational complexity increases
Solution Approach 1:
The patent replaces traditional signal processing methods with a neural network-based approach. Instead of using complex mechanical or algorithmic signal decomposition techniques that struggle with high-frequency information, the system uses a neural network that can naturally process and learn from high-frequency power signal patterns, achieving both precision and computational efficiency.
Data Source
AI summary
An energy meter is configured to determine component waveforms that form a measured waveform. The meter inputs the waveform into one or more entries of a data structure, each entry of the one or more entries of the data structure storing a weight value that is determined based at least in part on values of the data signatures representing the plurality of remote devices, each entry being connected to one or more other entries of the data structure. The meter, for each of the one or more entries, generates an output value by performing an arithmetic operation on the waveform stored at that entry, the arithmetic operation comprising a function of the weight value. The meter identifies, from among the data signatures, one or more particular data signatures that are represented in the waveform. The meter determines, based on the particular data signatures, an operational state of another device.