Adaptive Testing of Analog Prediction Algorithms Without Fault Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Power plant operators face challenges in monitoring and predicting anomalies in over 500 analog measurement points due to a lack of fault samples and effective testing methods for intelligent prediction algorithms, which are costly and incomplete in covering trend characteristics under fault conditions.
Innovation Solution
A self-adaptive test method for intelligent prediction algorithms that reads event records, calculates simulated measured values based on historical operating conditions and alarm thresholds, and adjusts sensitivity automatically, providing a standardized and economic means to obtain and test fault samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If intelligent prediction algorithms are implemented for monitoring analog measurement points, then prediction accuracy and anomaly detection capability are improved, but the requirement for fault samples increases making the system harder to deploy
Solution Approach 1:
The patent creates simulated fault samples by copying and modifying normal operational data. Virtual fault conditions are generated by injecting artificial anomalies into historical operational data, allowing the algorithm to be trained and tested without requiring actual fault occurrences. This copying approach resolves the contradiction by providing sufficient training data while maintaining system simplicity.
Solution Approach 2:
The patent performs preliminary generation of fault samples before algorithm deployment. By pre-generating simulated fault scenarios and storing them as training data, the system prepares all necessary samples in advance, eliminating the need to collect actual fault samples during operation and simplifying the deployment process.
2Quantity of substance
If actual fault samples are collected from debugging and troubleshooting, then training data availability is improved, but the cost and time consumption increase significantly
Solution Approach 1:
Instead of collecting actual fault samples through debugging and troubleshooting, the system copies normal operational data and synthetically generates fault conditions. This virtual sample generation approach provides sufficient training data quantity without the time-consuming process of actual fault collection and analysis.
Solution Approach 2:
The system performs preliminary generation of fault samples by processing historical operational data before algorithm training. This advance preparation creates a comprehensive dataset of simulated faults, eliminating the need for time-consuming actual fault collection during system deployment and operation.
3Ease of manufacture
If simulated measured values are calculated based on historical statistics, then the need for actual fault samples is reduced, but the accuracy of trend characteristics may be compromised
Solution Approach 1:
The patent applies parameter changes to normal operational data to generate simulated fault conditions. By systematically modifying parameters such as adding noise, changing trends, and adjusting statistical properties, the system creates diverse fault scenarios that maintain statistical validity while capturing various fault characteristics, thus preserving accuracy while enabling easy generation.
Solution Approach 2:
The system applies different simulation techniques to different portions of the data based on local characteristics. Normal operational phases use one simulation approach while transition phases use another, ensuring that each segment's specific characteristics are preserved in the simulated fault samples, thereby maintaining overall accuracy while enabling comprehensive coverage.
4Device complexity
If manual tracking of change trends is performed by attendants, then system complexity is reduced, but the ability to detect abnormalities in advance is lost
Solution Approach 1:
The patent uses simulated fault samples copied from normal operations to train the prediction algorithm, allowing the system to achieve advanced abnormality detection without requiring complex manual tracking procedures. The trained algorithm automatically detects patterns that would be difficult for human operators to identify manually.
Solution Approach 2:
The system replaces manual tracking with an automated intelligent algorithm. The mechanical process of attendants manually monitoring and analyzing trends is substituted with an automated computational system that processes data and predicts abnormalities, maintaining simplicity while dramatically improving prediction capability.
Data Source
AI summary
The present disclosure provides a self-adaptive test method for an intelligent prediction algorithm of analog measured values. Firstly, an event recording sequence, an analog measurement point ID and an analog measurement point alarm value are read from a time sequence event record table, an analog measurement point table and an alarm threshold table. Next, operation records of a normal operation state of a unit within a statistical cycle are acquired to form historical statistics of measured values of the analog measurement point based on switching value signals. Then, simulated measured values of the analog measurement point with time scales are calculated based on the historical statistics, the analog measurement point alarm value and an analog measurement point current measured value. Finally, sensitivity is calculated; and an alarm is sent to remind a technician to adjust the algorithm when the sensitivity is greater than a threshold.