Please use this identifier to cite or link to this item: http://dlib.scu.ac.ir/handle/1842/36085
Title: Health monitoring of renewable energy systems
Authors: Shek, Jonathan;Oterkus, Erkan;Thies, Philipp
subject: reliability;offshore wind turbines;data mining;physics based models;failure diagnosis;failure prognosis;risk assessment;FMEA;wind turbine maintenance;wind turbine CBM
Publisher: The University of Edinburgh
Description: The offshore wind energy industry has grown exponentially; globally, there is 12GW of installed capacity of offshore wind, of which over 95% has been installed in the past ten years. Access and maintenance in offshore wind farms can be difficult and considerably more expensive than onshore wind farms. Additionally, with low availability levels and greater downtime due to failures, there is a growing interest in the optimisation of operation and maintenance (O&M) activities to maximise profitability. Traditionally, maintenance activities on critical components and subsystems have deployed two maintenance approaches; time-based preventative or corrective. Time-based preventative or scheduled maintenance approaches are based on intervening at fixed intervals, determined in advance for each component. Scheduling is based on failure statistics such as mean time between failures (MTBF), mean time to repair (MTTR) or mean time to failure (MTTF). These come either from publicly available databases or operational measurements. As part of preventive maintenance activities, there are annual services of the turbine to replace and maintain any component or assembly based on manufacturers’ indications. On the other hand, the corrective maintenance approach involves operating equipment until it fails and then restoring it, repairing it, or replacing it. Due to conservative estimates regarding the probability of failure, preventive and corrective maintenance approaches have financial implications associated with them. In the preventive approach, components are frequently replaced before they reach the end of their working life. In contrast, corrective maintenance guarantees that the serviceable life of a component is maximised, but it is subjected to long downtime, which is expensive regarding energy generation loss. Additionally, failure of the component may cause consequential damage to other parts of the wind turbine system, resulting in even greater repair costs, downtime and loss of revenue. A comprehensive literature review has been undertaken in the areas of maintenance, turbine reliability, turbine failure modes and causes, physics of failure, condition monitoring techniques, and costs. The limitations and disadvantages of current operation and maintenance practices are identified, and new approaches combining the knowledge of the condition of components and historical data are proposed and compared to achieve optimal turbine availability and maintenance cost reduction. A Failure Modes and Effects Analysis (FMEA) was performed for the functional modes of each system, subsystem, assembly and component following the British standard BS EN 60812:2006. Currently, the most common offshore wind turbine uses three blades, a 3-stage gearbox, induction generator and a fully rated power converter. The Siemens 3.6MW -120 turbine is selected for this project as an example of this configuration. The main objectives of undertaking this comprehensive FMEA are to identify critical components and their failures with significant impact on the wind turbine operation in terms of maintainability, safety and availability. The assessment identified 500 components and almost 1000 failure causes. The most critical assemblies identified in terms of severity, occurrence and undetectability of the failure are; the frequency converter, pitch system, yaw system and gearbox. The implementation of a condition-based maintenance philosophy, including the development of real predictive approaches which estimate the remaining useful life of degrading critical components has been analysed by the recent literature. However, developing such capabilities for the critical assemblies identified is a significant technical challenge. This study aims to develop and demonstrate the implementation of a methodology and appropriate algorithms to optimise O&M of offshore wind farms, by estimating the remaining useful life of critical components with greater accuracy using a combination of physics-based models, statistical-based models and data mining approaches. A register of trends and likely the main causes of failures of the power converter, gearbox, yaw system and pitch system was generated through a thorough literature search and participation in conferences and workshops during the project. The main sources of failure of the power converter and gearbox have been represented by algorithms and physics-based models developed in Python and proprietary software, respectively. These algorithms comprise two phases: diagnosis or learning phase using historical data (such as SCADA or digital information recorded by condition monitoring systems) and prognosis phase using simulated data (using as a basis the wind turbine aero-elastic software FASTv8). The pitch system failure mechanisms were explored using a combination of data mining approaches and subject matter expert knowledge. Examples of approaches investigated and implemented include: Support Vector Machine (SVM) to define normal behaviour and K Nearest Neighbour (KNN) to classify new observations regarding operation state (green for normal operation, amber for abnormal operation, red for failure). New observations with amber or red colours need to be analysed further, to diagnose potential failure modes using a decision tree algorithm with more variables related to the pitch system. The goals of developing a well-defined strategy for maintenance interventions and optimised management of wind farm logistics are required to effectively improve wind farm availability while reducing the cost of operations. Additionally, a clear identification of uncertainties inherent in stochastic processes, necessary for estimating access, failure prognosis and failure probabilities is required for operators to make informed decisions. The final output of this work is an O&M cost model which analyses and compares a conventional O&M strategy using a combination of preventive and reactive maintenance against an O&M strategy using the approaches described above for failure prognosis and diagnosis. The analysis is performed for a fictitious offshore wind farm with one-year operational data. The results include availability, downtime, the cost of repair, loss of production, revenue losses and the hidden CO2 emissions of the maintenance activities taking into account a combined probability level to account for the uncertainties.
URI: https://www.era.lib.ed.ac.uk/handle/1842/36085
More Information: http://hdl.handle.net/1842/36085
Appears in Collections:Engineering, School of

Files in This Item:
Click on the URI links for accessing contents.
Title: Health monitoring of renewable energy systems
Authors: Shek, Jonathan;Oterkus, Erkan;Thies, Philipp
subject: reliability;offshore wind turbines;data mining;physics based models;failure diagnosis;failure prognosis;risk assessment;FMEA;wind turbine maintenance;wind turbine CBM
Publisher: The University of Edinburgh
Description: The offshore wind energy industry has grown exponentially; globally, there is 12GW of installed capacity of offshore wind, of which over 95% has been installed in the past ten years. Access and maintenance in offshore wind farms can be difficult and considerably more expensive than onshore wind farms. Additionally, with low availability levels and greater downtime due to failures, there is a growing interest in the optimisation of operation and maintenance (O&M) activities to maximise profitability. Traditionally, maintenance activities on critical components and subsystems have deployed two maintenance approaches; time-based preventative or corrective. Time-based preventative or scheduled maintenance approaches are based on intervening at fixed intervals, determined in advance for each component. Scheduling is based on failure statistics such as mean time between failures (MTBF), mean time to repair (MTTR) or mean time to failure (MTTF). These come either from publicly available databases or operational measurements. As part of preventive maintenance activities, there are annual services of the turbine to replace and maintain any component or assembly based on manufacturers’ indications. On the other hand, the corrective maintenance approach involves operating equipment until it fails and then restoring it, repairing it, or replacing it. Due to conservative estimates regarding the probability of failure, preventive and corrective maintenance approaches have financial implications associated with them. In the preventive approach, components are frequently replaced before they reach the end of their working life. In contrast, corrective maintenance guarantees that the serviceable life of a component is maximised, but it is subjected to long downtime, which is expensive regarding energy generation loss. Additionally, failure of the component may cause consequential damage to other parts of the wind turbine system, resulting in even greater repair costs, downtime and loss of revenue. A comprehensive literature review has been undertaken in the areas of maintenance, turbine reliability, turbine failure modes and causes, physics of failure, condition monitoring techniques, and costs. The limitations and disadvantages of current operation and maintenance practices are identified, and new approaches combining the knowledge of the condition of components and historical data are proposed and compared to achieve optimal turbine availability and maintenance cost reduction. A Failure Modes and Effects Analysis (FMEA) was performed for the functional modes of each system, subsystem, assembly and component following the British standard BS EN 60812:2006. Currently, the most common offshore wind turbine uses three blades, a 3-stage gearbox, induction generator and a fully rated power converter. The Siemens 3.6MW -120 turbine is selected for this project as an example of this configuration. The main objectives of undertaking this comprehensive FMEA are to identify critical components and their failures with significant impact on the wind turbine operation in terms of maintainability, safety and availability. The assessment identified 500 components and almost 1000 failure causes. The most critical assemblies identified in terms of severity, occurrence and undetectability of the failure are; the frequency converter, pitch system, yaw system and gearbox. The implementation of a condition-based maintenance philosophy, including the development of real predictive approaches which estimate the remaining useful life of degrading critical components has been analysed by the recent literature. However, developing such capabilities for the critical assemblies identified is a significant technical challenge. This study aims to develop and demonstrate the implementation of a methodology and appropriate algorithms to optimise O&M of offshore wind farms, by estimating the remaining useful life of critical components with greater accuracy using a combination of physics-based models, statistical-based models and data mining approaches. A register of trends and likely the main causes of failures of the power converter, gearbox, yaw system and pitch system was generated through a thorough literature search and participation in conferences and workshops during the project. The main sources of failure of the power converter and gearbox have been represented by algorithms and physics-based models developed in Python and proprietary software, respectively. These algorithms comprise two phases: diagnosis or learning phase using historical data (such as SCADA or digital information recorded by condition monitoring systems) and prognosis phase using simulated data (using as a basis the wind turbine aero-elastic software FASTv8). The pitch system failure mechanisms were explored using a combination of data mining approaches and subject matter expert knowledge. Examples of approaches investigated and implemented include: Support Vector Machine (SVM) to define normal behaviour and K Nearest Neighbour (KNN) to classify new observations regarding operation state (green for normal operation, amber for abnormal operation, red for failure). New observations with amber or red colours need to be analysed further, to diagnose potential failure modes using a decision tree algorithm with more variables related to the pitch system. The goals of developing a well-defined strategy for maintenance interventions and optimised management of wind farm logistics are required to effectively improve wind farm availability while reducing the cost of operations. Additionally, a clear identification of uncertainties inherent in stochastic processes, necessary for estimating access, failure prognosis and failure probabilities is required for operators to make informed decisions. The final output of this work is an O&M cost model which analyses and compares a conventional O&M strategy using a combination of preventive and reactive maintenance against an O&M strategy using the approaches described above for failure prognosis and diagnosis. The analysis is performed for a fictitious offshore wind farm with one-year operational data. The results include availability, downtime, the cost of repair, loss of production, revenue losses and the hidden CO2 emissions of the maintenance activities taking into account a combined probability level to account for the uncertainties.
URI: https://www.era.lib.ed.ac.uk/handle/1842/36085
More Information: http://hdl.handle.net/1842/36085
Appears in Collections:Engineering, School of

Files in This Item:
Click on the URI links for accessing contents.
Title: Health monitoring of renewable energy systems
Authors: Shek, Jonathan;Oterkus, Erkan;Thies, Philipp
subject: reliability;offshore wind turbines;data mining;physics based models;failure diagnosis;failure prognosis;risk assessment;FMEA;wind turbine maintenance;wind turbine CBM
Publisher: The University of Edinburgh
Description: The offshore wind energy industry has grown exponentially; globally, there is 12GW of installed capacity of offshore wind, of which over 95% has been installed in the past ten years. Access and maintenance in offshore wind farms can be difficult and considerably more expensive than onshore wind farms. Additionally, with low availability levels and greater downtime due to failures, there is a growing interest in the optimisation of operation and maintenance (O&M) activities to maximise profitability. Traditionally, maintenance activities on critical components and subsystems have deployed two maintenance approaches; time-based preventative or corrective. Time-based preventative or scheduled maintenance approaches are based on intervening at fixed intervals, determined in advance for each component. Scheduling is based on failure statistics such as mean time between failures (MTBF), mean time to repair (MTTR) or mean time to failure (MTTF). These come either from publicly available databases or operational measurements. As part of preventive maintenance activities, there are annual services of the turbine to replace and maintain any component or assembly based on manufacturers’ indications. On the other hand, the corrective maintenance approach involves operating equipment until it fails and then restoring it, repairing it, or replacing it. Due to conservative estimates regarding the probability of failure, preventive and corrective maintenance approaches have financial implications associated with them. In the preventive approach, components are frequently replaced before they reach the end of their working life. In contrast, corrective maintenance guarantees that the serviceable life of a component is maximised, but it is subjected to long downtime, which is expensive regarding energy generation loss. Additionally, failure of the component may cause consequential damage to other parts of the wind turbine system, resulting in even greater repair costs, downtime and loss of revenue. A comprehensive literature review has been undertaken in the areas of maintenance, turbine reliability, turbine failure modes and causes, physics of failure, condition monitoring techniques, and costs. The limitations and disadvantages of current operation and maintenance practices are identified, and new approaches combining the knowledge of the condition of components and historical data are proposed and compared to achieve optimal turbine availability and maintenance cost reduction. A Failure Modes and Effects Analysis (FMEA) was performed for the functional modes of each system, subsystem, assembly and component following the British standard BS EN 60812:2006. Currently, the most common offshore wind turbine uses three blades, a 3-stage gearbox, induction generator and a fully rated power converter. The Siemens 3.6MW -120 turbine is selected for this project as an example of this configuration. The main objectives of undertaking this comprehensive FMEA are to identify critical components and their failures with significant impact on the wind turbine operation in terms of maintainability, safety and availability. The assessment identified 500 components and almost 1000 failure causes. The most critical assemblies identified in terms of severity, occurrence and undetectability of the failure are; the frequency converter, pitch system, yaw system and gearbox. The implementation of a condition-based maintenance philosophy, including the development of real predictive approaches which estimate the remaining useful life of degrading critical components has been analysed by the recent literature. However, developing such capabilities for the critical assemblies identified is a significant technical challenge. This study aims to develop and demonstrate the implementation of a methodology and appropriate algorithms to optimise O&M of offshore wind farms, by estimating the remaining useful life of critical components with greater accuracy using a combination of physics-based models, statistical-based models and data mining approaches. A register of trends and likely the main causes of failures of the power converter, gearbox, yaw system and pitch system was generated through a thorough literature search and participation in conferences and workshops during the project. The main sources of failure of the power converter and gearbox have been represented by algorithms and physics-based models developed in Python and proprietary software, respectively. These algorithms comprise two phases: diagnosis or learning phase using historical data (such as SCADA or digital information recorded by condition monitoring systems) and prognosis phase using simulated data (using as a basis the wind turbine aero-elastic software FASTv8). The pitch system failure mechanisms were explored using a combination of data mining approaches and subject matter expert knowledge. Examples of approaches investigated and implemented include: Support Vector Machine (SVM) to define normal behaviour and K Nearest Neighbour (KNN) to classify new observations regarding operation state (green for normal operation, amber for abnormal operation, red for failure). New observations with amber or red colours need to be analysed further, to diagnose potential failure modes using a decision tree algorithm with more variables related to the pitch system. The goals of developing a well-defined strategy for maintenance interventions and optimised management of wind farm logistics are required to effectively improve wind farm availability while reducing the cost of operations. Additionally, a clear identification of uncertainties inherent in stochastic processes, necessary for estimating access, failure prognosis and failure probabilities is required for operators to make informed decisions. The final output of this work is an O&M cost model which analyses and compares a conventional O&M strategy using a combination of preventive and reactive maintenance against an O&M strategy using the approaches described above for failure prognosis and diagnosis. The analysis is performed for a fictitious offshore wind farm with one-year operational data. The results include availability, downtime, the cost of repair, loss of production, revenue losses and the hidden CO2 emissions of the maintenance activities taking into account a combined probability level to account for the uncertainties.
URI: https://www.era.lib.ed.ac.uk/handle/1842/36085
More Information: http://hdl.handle.net/1842/36085
Appears in Collections:Engineering, School of

Files in This Item:
Click on the URI links for accessing contents.