Sensor Data Scoping for Early Component Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electrical network congestion management systems lack proactive forecasting capabilities due to computational inefficiencies, reliance on inaccurate models, and the integration of renewable energy sources, leading to reactive and costly mitigation strategies.
Innovation Solution
A system utilizing machine learning algorithms to analyze scoped historical sensor data from renewable energy assets, training classification models to predict component failures with lead time, and generating alerts for proactive congestion management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AC power flow with Monte-Carlo simulation is used for congestion forecasting, then comprehensive scenario analysis is achieved, but computational efficiency deteriorates and real-time forecasting becomes infeasible
Solution Approach 1:
The patent segments the complex AC power flow model into simplified DC power flow models for different network regions. This segmentation allows parallel processing of multiple scenarios without the computational burden of full AC simulations, enabling real-time congestion forecasting while maintaining acceptable accuracy for congestion detection purposes.
Solution Approach 2:
The patent employs simplified DC power flow models as computationally inexpensive approximations instead of expensive AC models. These simplified models can be executed rapidly and discarded after each forecasting cycle, enabling extensive Monte-Carlo scenario analysis in real-time without the prohibitive computational cost of using full AC models for each scenario.
2Measurement precision
If detailed internal and external network models are created for accurate simulation, then forecasting accuracy is improved, but model creation cost and complexity increase significantly
Solution Approach 1:
The patent applies different levels of modeling detail to different parts of the network based on their congestion risk and importance. Critical transmission corridors use more detailed models, while less critical areas use simplified representations. This local differentiation maintains forecasting accuracy for congestion-prone areas while reducing overall model complexity and data requirements.
3Quantity of substance
If historical sensor data including failure period data is used for training, then more training samples are available, but model accuracy deteriorates due to noisy failure data
Solution Approach 1:
The patent extracts and removes sensor data from known failure periods from the training dataset. By excluding data collected during actual failure conditions, the training model learns to recognize precursors to failure without being confused by the noisy, anomalous data present during failure itself, thereby improving classification accuracy while maintaining sufficient training sample size from normal operating conditions.
Data Source
AI summary
An example method comprises receiving first historical data of a first time period and failure data, identifying at least some sensor data that was or potentially was generated during a first failure, removing the at least some sensor data to create filtered historical data, training a classification model using the filtered historical data, the classification model indicating at least one first classified state at a second period of time prior to the first failure indicated by the failure data, applying the classification model to second sensor data to identify a first potential failure state based on the at least one first classified state, the second sensor data being from a subsequent time period, generating an alert if the first potential failure state is identified based on at least a first subset of sensor signals generated during the subsequent time period, and providing the alert.