Anomaly Detection for Energy Storage Systems Using AI Temperature Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anomaly detection techniques for power supply devices are prone to misjudgments due to temperature changes in liquid-cooling systems being influenced by various factors, and the collection of high-power signals poses operational risks.
Innovation Solution
An anomaly detection method that involves receiving sensing data from energy storage system components, inputting this data into a temperature prediction model comprising model encoders, a reweighting model, and a model decoder, and determining abnormal operation by estimating the error between predicted and current temperatures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If temperature changes in liquid-cooling system are used to detect anomalies, then temperature monitoring is achieved, but misjudgment occurs due to environmental factors and fan influence
Solution Approach 1:
The patent segments the temperature monitoring function into two independent parts: (1) physical temperature sensing through sensors, and (2) anomaly detection through AI model analysis of operational parameters. This separation eliminates the direct coupling between environmental temperature fluctuations and anomaly detection, allowing each subsystem to perform its specialized function without interference from the other's limitations.
Solution Approach 2:
The patent introduces an AI model as an intermediary between raw temperature data and anomaly determination. The model processes temperature changes along with other operational parameters (loading amount, fan speed, environmental conditions) to distinguish between normal environmental fluctuations and actual anomalies, thereby filtering out false positives caused by environmental factors.
2Loss of information
If control system collects high-power signals directly, then sensing data is obtained, but operational risk increases due to bulky system structure and multiple modules
Solution Approach 1:
The patent extracts the high-power signal collection function from the main control system and implements it through dedicated sensing modules. This extraction allows the control system to receive processed sensing data without directly handling high-power signals, thereby maintaining operational safety while preserving complete sensing capabilities.
Solution Approach 2:
The patent uses sensing modules that create electrical copies or representations of the high-power signal states without requiring the control system to physically interface with high-power circuits. These copied signals are then transmitted to the control system for analysis, enabling safe remote monitoring of power supply device status.
3Reliability
If immediate shutdown is executed when sensing data exceeds tolerant ranges, then damage prevention is achieved, but false alarms cause unnecessary system interruptions
Solution Approach 1:
The patent performs preliminary analysis of temperature trends and operational parameters using an AI model before triggering protective shutdown actions. By evaluating multiple parameters and predicting future states, the system can distinguish between transient fluctuations that will self-correct and genuine anomalies requiring intervention, thereby preventing false shutdowns while maintaining protective functionality.
Solution Approach 2:
The patent implements continuous feedback loops where the AI model constantly monitors operational parameters, compares predicted versus actual states, and adjusts anomaly detection thresholds dynamically. This feedback mechanism allows the system to learn from operational patterns and improve its discrimination between normal variations and true anomalies, reducing false alarms while maintaining reliable damage prevention.
Data Source
AI summary
An anomaly detection method for an energy storage system includes receiving multiple sensing data retrieved from an electronic component of the energy storage system; inputting the multiple sensing data to a temperature prediction model; respectively receiving the multiple sensing data of same type and computing multiple time-series features related to the type based on the multiple sensing data of the type by each of a plurality of model encoders; outputting the multiple time-series features of each type to a reweighting model; computing a predicted temperature feature based on the multiple time-series features by the reweighting model and outputting the predicted temperature feature to a model decoder; reconstructing a predicted temperature feature by the model decoder to generate a predicted temperature of the electronic component; and determining whether the electronic component operates abnormally by estimating an error between the predicted temperature and a current temperature.