12 June 2023 in IEEE Transactions on Industry Applications
Strong nonlinearity and high volatility are the main features of wind turbine systems. Under the conditions of random wind speed changes, how to quickly control the wind turbine output power within the rated range is a major challenge for wind power system control. This paper proposed an iterative neuro-fuzzy Hammerstein model based model predictive control for wind turbines. Initially, an iterative neuro-fuzzy Hammerstein model is used to characterize the feature of wind turbines. The nonlinear static component's aerodynamic property is estimated using neuro-fuzzy networks, and the linear dynamic component is identified using the AutoRegressive with Exogenous Input (ARX) model. Following this, a generalized wind turbine controlled object with linear properties is constructed by computing the inverse of the nonlinear part, converting the wind turbine's nonlinear control challenge into a linear model control problem. This is done by the unique structure of the Hammerstein model, which allows the linear and nonlinear parts to be separated. Furthermore, a globally convergent parameter learning method is proposed and applied to identify the nonlinear parameters of the Hammerstein model. Eventually, the implementation of the Hammerstein-MPC is compared with the MPC, fuzzy MPC, and PI controllers by FAST simulation. These results demonstrate the superiority of the control effect based on the generalized wind turbine system.
1 April 2025 in IEEE Internet of Things Journal
Prevailing interest and advancement of fifth-generation (5G) related technology has inspired the study of the Internet of Vehicles (IoV). As per the request for privacy-preserving and collaborative learning, the federated learning paradigm has emerged as a significant supporting component, but it also requires facilitating the explainability of the incentive mechanism. Inspired by this, an asynchronous federated learning paradigm with graph-based reputation incentives has been proposed in the present paper for trajectory prediction tasks in IoV cases. According to prominent characteristics of IoV scenarios, where inter-relations of clients impose great effects on trajectory prediction, a feature extraction strategy based on Graph Neural Network(GNN) is proposed to extract latent features of IoV data and establish a reputation-based incentive mechanism. Integrated with this incentive scheme, an asynchronous federated learning framework is proposed to carry out collaborative trajectory forecasting tasks. As illustrated by the testing results on real-world datasets, the proposed scheme is capable of promoting overall federated learning performance greatly. Additionally, more experiments on proposed reputation incentives have been conducted and show an advance over loss function-based reputation incentives.
1 November 2022 in Journal of Renewable and Sustainable Energy
The pitch adjustment is critical to maintain the stable operation of the wind energy conversion system (WECS) under high wind speed, which can reduce mechanical load and provide safe and reliable power input for the power grid. It makes the design of a pitch controller arduous because the WECS is the high order system with nonlinear, multi-variable, and strong coupling. Based on this, a novel pitch controller is presented by employing dynamic just-in-time learning (JITL) based model predictive control to investigate wind turbine performance in a high wind speed operating region. This method captures the wind power process nonlinearities by the set of local models available online via JITL technology. Moreover, in order to meet the requirements of the real-time operation of the wind power process, a three-dimensional space data classification algorithm is employed to classify the data by calculating the local density and the minimum distance. Furthermore, considering the multi-objective and strongly nonlinear optimal control characteristics of model predictive control, the proposed dynamic linear model is introduced into the design of the nonlinear wind power model predictive control system. The optimal control rate of the controller is determined by solving the convex optimization problem within the design framework. Finally, the feasibility and superiority of the strategy are demonstrated by a typical wind power process.
23 August 2026 in Transactions of the Institute of Measurement and Control
The wind power generation process exhibits strong nonlinearity and multiple constraints, making it difficult to establish an accurate global model for model predictive control. In practical applications, model mismatch often leads to a decline in control performance. To address this, a tube-based model predictive control for wind power system using adaptive just-in-time learning modeling methodology is proposed. There are two core innovations: (1) A pre-clustering adaptive just-in-time learning method is adopted to construct a local dynamic model online as the nominal system, which ensures modeling accuracy while significantly reducing computational burden and (2) without explicitly distinguishing between the maximum power point tracking region and the pitch control region, a tube-based model predictive control strategy is developed so that the power tracking error is constrained within a Tube invariant set centered on the nominal system, effectively suppressing the effects of wind speed randomness and model mismatch. Simulation results on a 5-MW wind turbine demonstrate that the proposed strategy can smooth power fluctuations, improve tracking accuracy, and achieve superior robustness and computational efficiency.
13 May 2026 in IEEE Internet of Things Journal
As a prominent alternative for fossil energy, solar photovoltaic (PV) resource plays a key role during the migration process of energy supply structure. However, the prediction of its electric power production (EPP) data possesses randomness and instability due to environmental changes. Therefore, a novel framework is proposed which consists of a two-stage hierarchical clustering and optimal prediction model selection. According to the stability characteristics of EPP data, initial rough clusters are attained with Fuzzy C-means clustering through the first stage. The following second stage clustering based on hierarchical clustering is applied to finally generate three clusters of stable, weakly fluctuated and strongly fluctuated EPP data. Afterwards, three properly constructed prediction models have been implemented accordingly. A CNN-LSTM quantile regression model is proposed for the prediction of stable EPP data with consistent distributions. Yet for weakly fluctuated type, D-Vine copula quantile regression is applied to explore correlations between PV characteristics, which then helps to describe the specific distributions underneath. Since strongly fluctuated type of data tends to exhibit extreme variations and randomness, probability generation model combined with regression form of D-Vine copula could precisely establish correlations of different characteristics and hence weight the predictions to yield point prediction. Our framework addresses the PV power data with different random fluctuation characteristics by applying tailored prediction models to each type. Comparative experimental results show that the proposed framework has achieved higher prediction interval coverage, lower average interval width and comprehensive evaluation index in various scenarios, verifying its effectiveness and superiority.