Wind Turbine Failure Forecasting With Multi-Window Model Scoring

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current systems for predicting failures in renewable energy assets, such as wind turbines and solar panels, face challenges in accuracy and lead time due to the complexity of data processing and scalability issues, leading to reactive maintenance approaches rather than proactive warnings.

Innovation Solution

A centralized system for model selection, training, and scoring is implemented, utilizing machine learning algorithms to generate multiple failure prediction models with different observation and lead time windows, allowing for improved accuracy and scalability by centralizing data processing and reducing computational burdens.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple failure prediction models with different observation and lead time windows are generated, then prediction accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the model generation process into distinct pipelines: a model selection pipeline that creates multiple candidate models with different observation and lead time windows, and a model scoring pipeline that evaluates and selects the best performing models. This segmentation allows systematic exploration of multiple model configurations without overwhelming computational burden, as each pipeline handles specific tasks independently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The model selection pipeline performs preliminary action by generating multiple candidate failure prediction models with varying parameters before the scoring phase. This preliminary generation of diverse models with different observation windows and lead time configurations enables subsequent selection of optimal models based on performance metrics, rather than relying on a single pre-configured model.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If centralized data processing is implemented, then scalability is improved, but system complexity increases

Engineering Contradiction:
ImprovescalabilityVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements universal pipelines that can process data from multiple wind turbines and generate multiple types of failure prediction models using the same computational framework. The model selection and scoring pipelines are designed to handle diverse data inputs and generate various model configurations, making the system scalable across different datasets and turbine fleets without requiring separate processing systems for each case.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Loss of time

If lead time window is increased, then proactive maintenance capability is improved, but prediction accuracy may deteriorate

Engineering Contradiction:
Improvelead timeVSAvoidprediction accuracy
Core Design Contradiction:
Loss of timeVSMeasurement precision

Solution Approach 1:

The patent applies dynamics by creating multiple failure prediction models with different lead time window configurations rather than using a fixed window. The model selection pipeline generates models varying in observation periods and lead time extensions, allowing the system to dynamically select the appropriate lead time based on specific prediction needs and data characteristics, thereby balancing proactive maintenance capability with prediction accuracy.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes parameters by systematically varying observation window durations and lead time extensions across multiple candidate models. The model selection pipeline explores different parameter combinations (e.g., 30-day vs. 60-day observation windows, 7-day vs. 14-day lead times), and the scoring pipeline identifies which parameter configurations optimize both lead time and accuracy for specific failure modes, allowing flexible adjustment of these parameters based on operational requirements.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3902992B1Scalable system and engine for forecasting wind turbine failure
Publication Date: 2024.01.31 UTOPUS INSIGHTS INC
  • EP3902992B1 patent drawingFigure 1
  • EP3902992B1 patent drawingFigure 2
  • EP3902992B1 patent drawingFigure 3

AI summary

An example method utilizing different pipelines of a prediction system, comprises receiving event and alarm data from event logs, failure data, and asset data from SCADA system(s), retrieve patterns of events, receiving historical sensor data from sensors of components of wind turbines, training a set of models to predict faults for each component using the patterns of events and historical sensor data, each model of a set having different observation time windows and lead time windows, evaluating each model of a set using standardized metrics, comparing evaluations of each model of a set to select a model with preferred lead time and accuracy, receive current sensor data from the sensors of the components, apply the selected model(s) to the current sensor data to generate a component failure prediction, compare the component failure prediction to a threshold, and generate an alert and report based on the comparison to the threshold.