Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Equipment Health Assessment: Time Series Analysis for Wind Turbine Performance

Jana Backhus, Aniruddha Rajendra Rao, Chandrasekar Venkatraman, Abhishek Padmanabhan, A. Vinoth Kumar, Chetan Gupta

TL;DR

This work tackles wind turbine health and performance forecasting using SCADA data from 13 turbines, applying LSTM and Functional Neural Networks (FNN) to predict power output. An equal-weight ensemble of LSTM and FNN delivers stable, high accuracy on good-performance timelines and clearer deterioration signals on bad timelines, enabling proactive maintenance. The deterioration-detection framework uses per-turbine RMSE and RMSPE cutoffs derived from validation data, with RMSPE and a mixed RMSE/RMSPE metric yielding the strongest per-turbine F1-scores (0.634–0.892). The study reveals pronounced turbine-specific differences that necessitate customized models and demonstrates the approach’s potential to extend to other machinery for health monitoring and performance optimization.

Abstract

In this study, we leverage SCADA data from diverse wind turbines to predict power output, employing advanced time series methods, specifically Functional Neural Networks (FNN) and Long Short-Term Memory (LSTM) networks. A key innovation lies in the ensemble of FNN and LSTM models, capitalizing on their collective learning. This ensemble approach outperforms individual models, ensuring stable and accurate power output predictions. Additionally, machine learning techniques are applied to detect wind turbine performance deterioration, enabling proactive maintenance strategies and health assessment. Crucially, our analysis reveals the uniqueness of each wind turbine, necessitating tailored models for optimal predictions. These insight underscores the importance of providing automatized customization for different turbines to keep human modeling effort low. Importantly, the methodologies developed in this analysis are not limited to wind turbines; they can be extended to predict and optimize performance in various machinery, highlighting the versatility and applicability of our research across diverse industrial contexts.

Equipment Health Assessment: Time Series Analysis for Wind Turbine Performance

TL;DR

This work tackles wind turbine health and performance forecasting using SCADA data from 13 turbines, applying LSTM and Functional Neural Networks (FNN) to predict power output. An equal-weight ensemble of LSTM and FNN delivers stable, high accuracy on good-performance timelines and clearer deterioration signals on bad timelines, enabling proactive maintenance. The deterioration-detection framework uses per-turbine RMSE and RMSPE cutoffs derived from validation data, with RMSPE and a mixed RMSE/RMSPE metric yielding the strongest per-turbine F1-scores (0.634–0.892). The study reveals pronounced turbine-specific differences that necessitate customized models and demonstrates the approach’s potential to extend to other machinery for health monitoring and performance optimization.

Abstract

In this study, we leverage SCADA data from diverse wind turbines to predict power output, employing advanced time series methods, specifically Functional Neural Networks (FNN) and Long Short-Term Memory (LSTM) networks. A key innovation lies in the ensemble of FNN and LSTM models, capitalizing on their collective learning. This ensemble approach outperforms individual models, ensuring stable and accurate power output predictions. Additionally, machine learning techniques are applied to detect wind turbine performance deterioration, enabling proactive maintenance strategies and health assessment. Crucially, our analysis reveals the uniqueness of each wind turbine, necessitating tailored models for optimal predictions. These insight underscores the importance of providing automatized customization for different turbines to keep human modeling effort low. Importantly, the methodologies developed in this analysis are not limited to wind turbines; they can be extended to predict and optimize performance in various machinery, highlighting the versatility and applicability of our research across diverse industrial contexts.
Paper Structure (15 sections, 3 equations, 17 figures, 3 tables)

This paper contains 15 sections, 3 equations, 17 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Data and Methods
  3. Data
  4. Methods
  5. Long Short-Term Memory (LSTM)
  6. Functional Neural Networks (FNN)
  7. Model settings
  8. Hyperparameter Search
  9. Ensemble Method
  10. Deterioration Detection Approach
  11. Results
  12. Prediction Models
  13. Deterioration detection
  14. Discussion
  15. Conclusions

Figures (17)

  • Figure 1: Correlation Heatmap between Features of Interest
  • Figure 2: Actual power output of the 13 wind turbines
  • Figure 3: LSTM Cell (https://towardsdatascience.com/lstm-recurrent-neural-networks-how-to-teach-a-network-to-remember-the-past-55e54c2ff22e)
  • Figure 4: The general architecture of Functional Neural Network (FNN)
  • Figure 5: Deterioration Detection Method Overview
  • ...and 12 more figures