No abstract

Tidal power is an emerging field of renewable energy, harnessing the power of regular and predictable tidal currents. However, maintenance of submerged equipment is a major challenge. Routine visual inspections of equipment must be performed onshore, requiring the costly removal of turbines from the sea bed and resulting in long periods of downtime. The development of condition monitoring techniques providing automated fault detection can therefore be extremely beneficial to this industry, reducing the dependency on inspections and allowing maintenance to be planned efficiently.This paper investigates the use of deep learning approaches for fault detection within a tidal turbine’s generator from vibration data. Training and testing data were recorded over two deployment periods of operation from an accelerometer sensor placed within the nacelle of the turbine, representing ideal and faulty responses over a range of operating conditions. The paper evaluates a deep learning approach through a stacked autoencoder network in comparison to feature-based classification methods, where features have been extracted over varying rotation speeds through the Vold-Kalman filter.

Within the field of power generation, aging assets and a desire for improved maintenance decision-making tools have led to growing interest in asset prognostics. Valve failures can account for 7% or more of mechanical failures, and since a conventional power station will contain many hundreds of valves, this represents a significant asset base. This paper presents a prognostic approach for estimating the remaining useful life (RUL) of valves experiencing degradation, utilizing a similarity-based method. Case study data is generated through simulation of valves within a 400MW Combined Cycle Gas Turbine power station. High fidelity industrial simulators are often produced for operator training, to allow personnel to experience fault procedures and take corrective action in a safe, simulation environment, without endangering staff or equipment. This work repurposes such a high fidelity simulator to generate the type of condition monitoring data which would be produced in the presence of a fault. A first principles model of valve degradation was used to generate multiple run-to-failure events, at different degradation rates. The associated parameter data was collected to generate a library of failure cases. This set of cases was partitioned into training and test sets for prognostic modeling and the similarity based prognostic technique applied to calculate RUL. Results are presented of the technique’s accuracy, and conclusions are drawn about the applicability of the technique to this domain.

Harnessing the power of currents from the sea bed, tidal power has great potential to provide a means of renewable energy generation more predictable than similar technologies such as wind power. However, the nature of the operating environment provides challenges, with maintenance requiring a lift operation to gain access to the turbine above water. Failures of system components can therefore result in prolonged periods of downtime while repairs are completed on the surface, removing the system’s ability to produce electricity and damaging revenues. The utilization of effective condition monitoring systems can therefore prove particularly beneficial to this industry.This paper explores the use of the CRISP-DM data mining process model for identifying key trends within turbine sensor data, to define the expected response of a tidal turbine. Condition data from an operational 1 MW turbine, installed off the coast of Orkney, Scotland, was used for this study. The effectiveness of modeling techniques, including curve fitting, Gaussian mixture modeling, and density estimation are explored, using tidal turbine data in the absence of faults. The paper shows how these models can be used for anomaly detection of live turbine data, with anomalies indicating the possible onset of a fault within the system.