Robust fault estimation for wind turbine energy via hybrid systems

Abstract

The rapid development of modern wind turbine technology has led to increase demand for improving system reliability and practical concern for robust fault monitoring scheme. This paper presents the investigation of a 5MW Dynamic Wind Turbine Energy System that was designed to sustain condition monitoring and fault diagnosis with the goal of improving the reliability operations of universal practical control systems. A hybrid stochastic technique is proposed based on an augmented observer combined with eigenstructure assignment for the parameterisation and the genetic algorithm (GA) optimisation to address the attenuation of uncertainty mostly generated by disturbances. Scenarios-based are employed to explore sensor and actuator faults that have direct and indirect impacts on modern wind turbine system, based on monitoring components that are prone to malfunction. The analysis is aimed to determine the effect of concerned simulated faults from uncertainty in respect to environmental disturbances mostly challenged in real-world operations. The efficiency of the proposed approach will improve the reliability performance of wind turbine system states and diagnose uncertain faults simultaneously. The simulation outcomes illustrate the robustness of the dynamic turbine systems with a diagnostic performance to advance the practical solutions for improving reliable systems.

Diagnosis Methods for Mechatronic Systems

“…It gives a design of the gain observer by using the left and right eigenstructure methods and aims to cancel directly the disturbance (for more details, see (Xu & Tseng 2007)). Recently, diagnostic observers are used in many applications in order to diagnosis damage (see, e.g., (Odofin et al 2016), (Sen & Bhattacharya 2017) and (Odofin et al 2018)). Residual generation using unknown input observer is a important diagnostic method and applied specially to isolate the fault.…”

Event‐triggered scheme for fault detection and isolation of non‐linear system with time‐varying delay

“…Recent research carried out in this field identified two main approaches of FDI: signal‐based FDI and model‐based FDI. With the advancement in the field of digital computers and system identification, model‐based FDI has emerged as a powerful approach [10–12]. The model‐based approach depends on the mathematical model, which could cause false alarm between the actual system and model, which makes FDI dormant in most cases.…”

Development of an SVR Model for the Fault Diagnosis of Large-Scale Doubly-Fed Wind Turbines Using SCADA Data

“…The literature review also shows some successful studies in predicting faults caused by complex nonlinearity reasons using operational SCADA data combined with diverse approaches such as adaptive neuron-fuzzy interference systems (ANFIS), Bayesian Networks and Deep Learning Networks [7][8][9]. A hybrid stochastic technique is proposed in reference [10], which is based on an augmented observer combined with Eigen structure assignment for the parameterization and the genetic algorithm (GA) optimization to address the attenuation of uncertainty mostly generated by disturbances. Data mining techniques are also applied to this data combination to model the power curve [11].…”