Clustering Power Storage Systems for Anomaly Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In power storage systems, individual differences in operation and environment lead to a broadened threshold value range for anomaly detection, degrading the performance of anomaly detection models.

Innovation Solution

Clustering power storage systems into groups based on similar operational characteristics and generating detection models for each cluster to reduce individual differences and improve anomaly detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a detection model is generated using data from a large number of power storage systems to account for individual differences, then the model becomes more generalizable, but the threshold value range expands and detection performance degrades

Engineering Contradiction:

Improvegeneralizability of detection modelVSAvoidanomaly detection performance

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the power storage systems into multiple clusters based on operational characteristics and environmental factors. Each cluster has its own detection model and threshold values, allowing the system to maintain high detection performance while accounting for individual differences through group-specific parameters rather than a single broad model

Inventive Principle:

Principle #1Segmentation

2Device complexity

If a single detection model is used for all power storage systems, then the system complexity is reduced, but individual differences cause threshold value expansion and detection accuracy loss

Engineering Contradiction:

Improvedetection model structureVSAvoidanomaly detection accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

Instead of using a single detection model for all systems, the patent divides the system into multiple clusters with separate detection models. This segmentation maintains reasonable complexity while significantly improving detection accuracy by accounting for individual differences through cluster-specific parameters

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the parameters of the detection model by creating cluster-specific threshold values and detection parameters. Each cluster has optimized parameters tailored to its operational characteristics, allowing the system to maintain high accuracy without excessive complexity

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11293989B2Anomaly detection device, anomaly detection method, and recording medium

Publication Date: 2022.04.05 NEC CORP

AI summary

An anomaly detection device includes: at least one memory configured to store instructions; and at least one processor configured to execute the instructions to: cluster a plurality of power storage systems into a plurality of clusters; generate a detection model for detecting an anomaly in the power storage systems for each of the clusters; and detect an anomaly in the power storage system for each of the clusters by using the detection model associated with the cluster.