Joaquim Blesa

Fault Diagnosis of Wind Turbines using a Set-membership Approach

Abstract In this paper, the problem of fault diagnosis in wind turbines will be addressed applying a set-membership approach. Fault detection is based on the use of parity equations and unknown but bound description of the noise and modelling errors. Fault detection test is based on checking consistency between measurements and the model by finding if there is parameter in the feasible parameter set (approximated by a zonotope) that is compatible with modelling and noise bounds. The fault isolation algorithm is based on analyzing the observed fault signatures on-line and matching them with the theoretical ones obtained using structural analysis. Finally, the proposed approach will be tested using the wind turbine benchmark proposed in the context of the WT FDI competition.

An Interval NLPV Parity Equations Approach for Fault Detection and Isolation of a Wind Farm

In this paper, the problem of fault diagnosis of a wind farm is addressed using interval nonlinear parameter-varying (NLPV) parity equations. Fault detection is based on the use of parity equations assuming unknown but bounded description of the noise and modeling errors. The fault detection test is based on checking the consistency between the measurements and the model, by finding if the formers are inside the interval prediction bounds. The fault isolation algorithm is based on analyzing the observed fault signatures online and matching them with the theoretical ones obtained using structural analysis. Finally, the proposed approach is tested using the wind farm benchmark proposed in the context of the wind farm fault-detection-and-isolation/fault-tolerant-control competition.

Sensor placement for leak detection and location in water distribution networks

The performance of a leak detection and location algorithm depends on the set of measurements that are available in the network. This works presents and optimization strategy that maximizes the leak diagnosability performance of the network. The goal is to characterize and determine a sensor configuration that guarantees a maximum degree of disnosability while the sensor configuration cost satifies a budgetary constraint. To efficiently handle the complexity of the distribuion networl an efficient branch and bound search strategy based on a strucutrual model is used. However, in order to reduce even more the size and the complexity of the problem the present work proposes to combine this methodology with clustering techniques. The strategy developed in this work is successfully applied to determine the optimal set of pressure sensors that should be installed to a District Metered Area in the Baarcelona Water Distribution Network.

Fault tolerant control of the wind turbine benchmark using virtual sensors/actuators

Abstract In this paper, the problem of Fault Tolerant Control (FTC) of the wind turbine benchmark is addressed. The paper proposes the use of virtual sensor/actuator approaches to deal with sensor and actuator faults, respectively. The paper suggests the reformulation of these FTC schemes that have been already proposed in state space form in input/output form. The FTC module will use the information from the Fault Detection and Isolation (FDI) module previously designed using set-membership techniques. A fault estimation scheme is also proposed based on batch least squares approach. The performance of the proposed FTC schemes will be assessed using the proposed fault scenarios considered in the FTC benchmark.

Gain-Scheduled Smith Predictor PID-Based LPV Controller for Open-Flow Canal Control

In this paper, a gain-scheduled Smith Predictor PID controller is proposed for the control of an open-flow canal system that allows for dealing with large variation in operating conditions. A linear parameter varying (LPV) control-oriented model for open-flow canal systems based on a second-order delay Hayami model is proposed. Exploiting the second-order structure of this model, an LPV PID controller is designed using H∞ and linear matrix inequalities pole placement. The controller structure includes a Smith Predictor, real time estimated parameters from measurements (including the known part of the delay) that schedule the controller and predictor and unstructured dynamic uncertainty, which covers the unknown portion of the delay. Finally, the proposed controller is validated in a case study based on a single real reach canal: the Lunax Gallery at Gascogne (France).

Linear parameter varying modeling and identification for real-time control of open-flow irrigation canals

Irrigation canals are open-flow water hydraulic systems, whose objective is mainly to convey water from its source down to its final users. They are large distributed systems characterized by non-linearity and dynamic behavior that depends on the operating point. Moreover, in canals with multiple reaches dynamic behavior is highly affected by the coupling among them. The physical model for those systems leads to a distributed-parameter model whose description usually requires partial differential equations (PDEs). However, the solution and parameter estimation of those PDE equations can only be obtained numerically and imply quite time-consuming computations that make them not suitable for real-time control purposes. Alternatively, in this paper, it will be shown that open-flow canal systems can be suitably represented for control purposes by using linear parameter-varying (LPV) models. The advantage of this approach compared to the use of PDE equation is that allows simpler models which are suitable for control design and whose parameters can be easily identified from input-output data by means of classical identification techniques. In this paper, the well-known control-oriented, model named integral delay zero (IDZ), that is able to represent the canal dynamics around a given operating point by means of a linear time-invariant (LTI) model is extended to multiple operating points by means of an LPV model. The derivation of this LPV model for single-reach open-flow canal systems as well as its extension to multiple-reach open-flow canals is proposed. In particular, the proposed methodology allows deriving the model structure and estimating model parameters using data by means of identification techniques. Thus, a gray-box control model is obtained whose validation is carried out using single-pool and two-pool test canals obtaining satisfactory results. Irrigation canals are open-flow water hydraulic systems that are described by a distributed parameter model.However, those PDE equations can only be obtained numerically and are not suitable for real-time control.This paper shows that open-flow canal systems can be suitably represented by using LPV models for control.In this paper, the IDZ model is extended to multiple operating points by means of an LPV model.Results on two cases studies illustrate the validity of the proposed approach.

Clustering techniques applied to sensor placement for leak detection and location in water distribution networks

This work presents an optimization strategy that maximizes the leak locatability performance of water distribution networks (WDN). The goal is to characterize and determine a sensor configuration that guarantees a maximum degree of locatability while the sensor configuration cost satisfies a budgetary constraint. The method is based on pressure sensitivity matrix analysis and an exhaustive search strategy. In order to reduce the size and the complexity of the problem the present work proposes to combine this methodology with clustering techniques. The strategy developed in this work is successfully applied to determine the optimal set of pressure sensors that should be installed in a district metered area (DMA) in the Barcelona WDN.

Nonlinear set-membership identification and fault detection using a Bayesian framework: Application to the wind turbine benchmark

This paper deals with the problem of nonlinear set-membership identification and fault detection using a Bayesian framework. The paper presents how the set-membership model estimation can be reformulated from a Bayesian viewpoint in order to determine the feasible parameter set and, in a posterior fault detection stage, to check the consistency between the model and the data. The paper shows that the Bayesian approach, assuming uniform distributed measurement noise and flat model prior probability distribution, leads to the same feasible parameter set as the set-membership technique. To illustrate this point a comparison with the subpavings approach is included. Finally, by means of the application to the wind turbine benchmark problem, it is shown how the Bayesian fault detection test works successfully.

Adaptive and predictive control architecture of inland navigation networks in a global change context: application to the Cuinchy-Fontinettes reach

In this paper, an adaptive and predictive control architecture is proposed to improve the management of inland navigation networks in a global change context. This architecture aims at ensuring the seaworthiness conditions of inland navigation networks, and to improve the efficiency of the water resource management. It is based on supervision and prognosis modules which allow the estimation of the current state of the network, and the forecasting of the extreme event occurrence. According to these indicators and to the management constraints and objectives, control strategies of the inland navigation networks will be adapted to limit the impacts of the extreme events. To achieve this aim, three challenges are considered and discussed in this paper. The first one consists in proposing an accurate modeling approach of navigation reaches which are characterized by large scale, nonlinearities, time delays, unknown inputs and outputs, etc. The second one is to increase the knowledge about potentiality of extreme events, consequences of the climate change. The prediction of these events is rather complex due to their rarity, the spacio-temporal scale of the networks, etc. Finally, the third one is the pooling of the two first contributions, i.e. the model of the system and the knowledge about extreme events. Thus, the resilience of the system and the adaptation of the management strategies could be realized

Leak Detection, Isolation and Estimation in Pressurized Water Pipe Networks using LPV Models and Zonotopes

In this paper, a leak detection, isolation and estimation methodology in pressurized water pipe networks is proposed. The methodology is based on computing residuals which are obtained comparing measured pressures (heads) in selected points of the network with their estimated values by means of a Linear Parameter Varying (LPV) model and zonotopes. The structure of the LPV model is obtained from the non-linear mathematical model of the network. The proposed detection method takes into account modelling uncertainty using zonotopes. The isolation and estimation task employs an algorithm based on the residual fault sensitivity analysis. Finally, a typical water pipe network is employed to validate the proposed methodology. This network is simulated using EPANET software. Parameters of LPV model and their uncertainty bounded by zonotopes are estimated from data coming from this simulator. A leak scenario allows to assess the effectiveness of the proposed approach.