In this paper, a novel data-driven fault diagnosis method by combining deep canonical variate analysis and Fisher discriminant analysis (DCVA-FDA) is proposed for complex industrial processes. Inspired by the recently developed deep canonical correlation analysis, a new nonlinear canonical variate analysis called deep canonical variate analysis (DCVA) is first developed by incorporating deep neural networks into canonical variate analysis. Based on DCVA, a residual generator is designed for the fault diagnosis process. Fisher discriminant analysis is applied in the feature space spanned by residual vectors. Then, a Bayesian inference classifier is performed in the reduced dimensional space of FDA to label the class of process data. A continuous stirred-tank reactor and an industrial benchmark of the Tennessee Eastman process are carried out to test the performance of DCVA-FDA fault diagnosis. The experimental results demonstrate that the proposed DCVA-FDA fault diagnosis is able to significantly improve the fault diagnosis performance when compared to other methods also examined in this study.

Incipient fault detection plays a crucial role in preventing the occurrence of serious faults or failures in industrial processes. In most industrial processes, linear, and nonlinear relationships coexist. To improve fault detection performance, both linear and nonlinear features should be considered simultaneously. In this article, a novel hybrid linear-nonlinear statistical modeling approach for data-driven incipient fault detection is proposed by closely integrating recently developed canonical variate dissimilarity analysis and mixed kernel principal component analysis (MKPCA) using a serial model structure. Specifically, canonical variate analysis (CVA) is first applied to estimate the canonical variables (CVs) from the collected process data. Linear features are extracted from the estimated CVs. Then, the canonical variate dissimilarity (CVD) which quantifies model residuals in the CVA state-subspace is calculated using the estimated CVs. To explore the nonlinear features, the nonlinear principal components are extracted as nonlinear features through performing MKPCA on CVD. Fault detection indices are formed based on Hotelling's T ² as well as Q statistics from the extracted linear and nonlinear features. Moreover, kernel density estimation is utilized to determine the control limits. The effectiveness of the proposed method is demonstrated by the comparisons with other relevant methods via simulations based on a closed-loop continuous stirred-tank reactor process.

The paper introduces a set of modelling methods to predict the failure probability of offshore wind turbines in the China coastal waters. In detail, a series of full-set three-dimensional meteorology simulations are conducted using the Weather Research and Forecast (WRF) model to provide the extreme wind fields while the extreme wave fields are predicted according to the conditional probability model. After the extreme wind and wave fields are prepared, the ultimate loads on the three critical parts of an OC4-DeepCwind semisubmersible wind turbine are simulated using the Fatigue, Aerodynamics, Structures, and Turbulence (FAST) code, which in turn leads to the prediction of the failure probability. It has been found that the OC4-DeepCwind semisubmersible constructed in the Taiwan Strait has higher chance to fail under typhoon conditions than in other parts of the China coastal waters, showing a maximum failure probability of 0.6. In addition, the failure probabilities in the South and East China Sea are in the range of 0.3-0.45. The probability densities of the ultimate loads show, in some cases, a bimodal shape, which indicates that the loads as well as the failure probability are not in a simple linear relationship with the extreme wind speeds and wave heights/periods.

The Immersion and Invariance (I&I) wind speed estimator is a powerful and widely-used technique to estimate the rotor effective wind speed on horizontal axis wind turbines. Anyway, its global convergence proof is rather cumbersome, which hinders the extension of the method and proof to time-delayed and/or uncertain systems. In this letter, we illustrate that the circle criterion can be used as an alternative method to prove the global convergence of the I&I estimator. This also opens up the inclusion of time-delays and uncertainties. First, we demonstrate that the I&I wind speed estimator is equivalent to a torque balance estimator with a proportional correction term. As the nonlinearity in the estimator is sector bounded, the well-known circle criterion is applied to the estimator to guarantee its global convergence for time-delayed systems. By looking at the theoretical framework from this new perspective, this letter further proposes the addition of an integrator to the correction term to improve the estimator performance. Case studies show that the proposed estimator with an additional integral correction term is effective at wind speed estimation. Furthermore, its global convergence can be guaranteed by the circle criterion for time-delayed systems.

The combined wind speed estimator and tip-speed ratio (WSE–TSR) tracking wind turbine control scheme has seen recent and increased traction from the wind industry. The modern control scheme provides a flexible trade-off between power and load objectives. On the other hand, the Kω² controller is often used based on its simplicity and steady-state optimality and is taken as a baseline here. This paper investigates the potential benefits of the WSE–TSR tracking controller compared to the baseline by analysis through a frequency-domain framework and by optimal calibration through a systematic procedure. A multi-objective optimisation problem is formulated for calibration with the conflicting objectives of power maximisation and torque fluctuation minimisation. The optimisation problem is solved by approximating the Pareto front based on the set of optimal solutions found by an explorative search. The Pareto fronts were obtained by mid-fidelity simulations with the National Renewable Energy Laboratory (NREL) 5 MW turbine under turbulent wind conditions for calibration of the baseline and for increasing fidelities of the WSE–TSR tracking controller. Optimisation results show that the WSE–TSR tracking controller does not provide further benefits in energy capture compared to the baseline Kω² controller. There is, however, a trade-off in torque control variance and power capture with control bandwidth. By lowering the bandwidth at the expense of generated power of 2 %, the torque actuation effort reduces by 80 % with respect to the optimal calibration corresponding to the highest control bandwidth.

In the next decade, further digitalisation of the entire wind energy project lifecycle is expected to be a major driver for reducing project costs and risks. In this paper, a literature review on the challenges related to implementation of digitalisation in the wind energy industry is first carried out, showing that there is a strong need for new solutions that enable co-innovation within and between organisations. Therefore, a new collaboration method based on a digital ecosystem is developed and demonstrated. The method is centred around specific “challenges”, which are defined by “challenge providers” within a topical “space” and made available to participants via a digital platform. The data required in order to solve a particular “challenge” are provided by the “challenge providers” under the confidentiality conditions they specify. The method is demonstrated via a case study, the EDP Wind Turbine Fault Detection Challenge. Six submitted solutions using diverse approaches are evaluated. Two of the solutions perform significantly better than EDP’s existing solution in terms of Total Prediction Costs (saving up to €120,000). The digital ecosystem is found to be a promising solution for enabling co-innovation in wind energy in general, providing a number of tangible benefits for both challenge and solution providers.