Machine Learning for Wind Turbine Fault Prediction through the Combination of Datasets from Same Type Turbines

Early fault detection in wind turbines is key to reduce both costs and uncertainty in the generation of energy and operation of these structures. The isolation of many wind farms, especially those offshore, makes scheduled maintenance very costly and on many occasions inefficient. In addition, the downtime of these structures is typically long and a predictive solution is much needed to 1) help prepare for the maintenance procedure beforehand, for instance to avoid delays when waiting for the required resources and components for maintenance to be available and, 2) avoid the possibility of more destructive system failures. Predicting failures in such complex systems requires modeling of multiple components in isolation and as a whole. Physics-based and data-based models are used for this purpose, which have been proven useful in this regard. Specifically, Machine Learning algorithms are proven to be a valuable resource in a wide range of problems in this industry, however a solution capable of accurately predicting the range of faults of a particular type of wind turbine is still a challenge. In this paper, we will introduce the capabilities of machine learning for wind turbine fault prediction, as well as a technique to predict different types of faults. We will compare the performance of two well established machine learning algorithms (namely K-Nearest Neighbour and Random Forest classifiers) on real wind turbine data which have produced great levels of prediction accuracy. We also propose data augmentation methods to help enhance the training of ML models when wind turbine data is scarce by merging data from turbines of the same type.

Modeling the Business Value of an Early Fault Detection System using Monte Carlo Simulation

The automotive industry is undergoing a period of rapid advancement, as OEMs race to develop the next generation of electric and autonomous vehicles. Many manufacturers are investing in prognostics technology, which has made advancements mainly in the aerospace industry over the past couple decades. Unlike aerospace applications, which have relatively more safety-critical systems, it can be more challenging to identify a business case for developing a prognostics or early fault detection system for an automotive application. In the retail setting, early fault detection systems may increase warranty costs, and the benefits to customer satisfaction may not be worth this additional cost. For fleet managers who own and operate many vehicles, however, a business case can be made based on the value of preventing unexpected downtime and unnecessary maintenance. Developing a reliable early fault detection algorithm for a complex system can be an expensive undertaking, requiring many parts, months of data collection, and possibly years of effort, so it is important to understand the possible return on investment for the effort. In this paper, we present a method to model the business value of an early fault detection system. The method is generic and may be applied to any system where the failure modes are purely fatigue based (i.e. abuse modes are excluded), and the failure rate of each part in the system can be independently modelled using a time-to-failure probability density function. The model is based on Monte Carlo simulation, and the assumptions and limitations are explored. The model can be used to estimate the expected savings from implementing an early fault detection system and derive requirements on the true positive and false positive rates required for the fault detection system to meet its business objectives. An example is presented with application to a two-stage gearbox, such as one that may be found in an electric vehicle powertrain. The example shows how to estimate the parameters for each component, how to estimate the costs associated with failure, and ultimately how to interpret the model outputs and drive business decisions.

An Adaptive Early Fault Detection Model of Induced Draft Fans Based on Multivariate State Estimation Technique

The induced draft (ID) fan is an important piece of auxiliary equipment in coal-fired power plants. Early fault detection of the ID fan can provide predictive maintenance and reduce unscheduled shutdowns, thus improving the reliability of the power generation. In this study, an adaptive model was developed to achieve the early fault detection of ID fans. First, a non-parametric monitoring model was constructed to describe the normal operating characteristics with the multivariate state estimation technique (MSET). A similarity index representing operation status was defined according to the prediction deviations to produce warnings of early faults. To deal with the model accuracy degradation because of variant condition operation of the ID fan, an adaptive strategy was proposed by using the samples with a high data quality index (DQI) to manage the memory matrix and update the MSET model, thereby improving the fault detection results. The proposed method was applied to a 300 MW coal-fired power plant to achieve the early fault detection of an ID fan. In addition, fault detection by using the model without an update was also compared. Results show that the update strategy can greatly improve the MSET model accuracy when predicting normal operations of the ID fan; accordingly, the fault can be detected more than 4 h earlier by using the strategy with the adaptive update when compared to the model without an update.

A Fault Detection System for The Yaw Control of A HAWT Based on Neural Networking

The objective of the yaw control system in a horizontal axis wind turbine (HAWT) is to follow the wind direction with a minimum error. In this paper, a data driven fault detection approach of a HAWT is applied. Three simulation programs were utilized in order to model a 1.5 MW HAWT. These programs are Fatigue, Aerodynamics, Structures, and Turbulence(FAST), TurbSim, and MATLAB. The approach is implemented under normal operating scenarios while considering different wind velocities. Different kinds of faults were applied to the system for a nacelle-yaw angle error ranging from -10° to +20°. The simulation results of the Tower Top Deflection (TTD) in the time domain were transferred into frequency domain by Fast Fourier Transform (FFT). The output variables were used in order to build a Neural Networking, which will monitor the performance of the wind turbine. The built Neural Networking will also provide an early fault detection to avoid the operating conditions that lead to sudden turbine breakdown. The present work provides initial results that are useful for remote condition monitoring and assessment of a 1.5MW HAWT. The simulation results indicate that the implemented Neural Networking can achieve improvement of the wind turbine operation and maintenance level.