EV Charging Detection Using Motif Discovery and ML Classifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing adoption of electric vehicles poses a challenge for electric utilities as they struggle to accurately identify EV charging loads, leading to potential transformer overloading and equipment failure, as existing techniques are either inaccurate or unable to detect EVs effectively.
Innovation Solution
A system and method using motif discovery and machine learning classifiers to identify electric vehicle charge events by analyzing time series data from utility customer accounts, encoding kWh values into patterns, and training classifiers to detect EV charging motifs and features, enabling proactive measures to manage grid load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If EV charging load is not accurately identified, then utility equipment may be overloaded and fail, but implementing accurate detection increases system complexity
Solution Approach 1:
The patent introduces an intermediary detection system that analyzes utility meter data to identify EV charging events. This intermediary layer between the EV charging infrastructure and utility equipment enables accurate identification of EV loads without requiring direct modification of transformers or charging stations, thus improving reliability while limiting complexity increase to the detection system alone.
Solution Approach 2:
The patent replaces physical modifications to electrical equipment with a data-driven detection approach. Instead of installing additional sensors or modifying transformers physically, the system uses machine learning algorithms to analyze existing utility meter data patterns, substituting mechanical/electrical complexity with computational analysis.
2Measurement precision
If EV charging events are detected using traditional methods, then detection may be achieved, but accuracy is insufficient leading to false identification
Solution Approach 1:
The patent applies preliminary action by training machine learning classifiers on historical EV charging data before deployment. The system pre-processes utility meter data, identifies characteristic EV charging patterns, and stores trained models that can accurately detect EV events. This preliminary training phase enables high detection accuracy when the system is deployed in production.
Solution Approach 2:
The patent implements dynamics by using adaptive machine learning classifiers that can learn and adapt to different EV charging patterns. The detection system dynamically adjusts to various charging behaviors, vehicle types, and usage patterns, enabling accurate identification across diverse scenarios rather than relying on fixed, static detection rules.
3Adaptability or versatility
If EV penetration increases, then more vehicles can be charged, but transformer overload risk increases
Solution Approach 1:
The patent implements feedback by continuously monitoring utility meter data to identify EV charging events and their impact on load patterns. This feedback information enables utilities to proactively manage transformer loads, adjust charging strategies, and prevent overload conditions before they occur, allowing higher EV penetration while maintaining transformer reliability.
Solution Approach 2:
The detection system performs preliminary identification of EV charging events, enabling utilities to take proactive measures before transformer overload occurs. By detecting EV loads in advance, utilities can implement load management strategies, redistribute charging demand, or alert customers to avoid peak charging times, thus preventing equipment failure.
Data Source
AI summary
Systems, methods, and other embodiments are associated with detecting an electric vehicle charging event. The system receives unknown time series data of usage values of electricity consumption, wherein the unknown time series data is unknown to have electric vehicle (EV) charge events. For a given account, the unknown time series data is converted into time intervals with corresponding usage values. Each time interval is encoded with a symbol from a series of symbols representing a level of electricity consumption, wherein the encoding generates an encoded consumption pattern of symbols. The system detects whether the encoded consumption pattern includes a string of high usage symbols that are similar to a known EV charge motif that represents a known EV charging event. Based on the detecting, the given account is marked as having an electric vehicle charge event or as not having an electric vehicle charge event.