Jingtao Huang

Accurate flux observation is crucial for the high-performance control of induction motors (IMs). Implementing a full-order flux observer algorithm in digital controllers requires discretizing the continuous-domain full-order flux observer. However, the errors introduced by discretization increase with rising rotor speed. In the field-weakening region, inappropriate discretization methods can lead to significant flux estimation errors, severely affecting the performance of model predictive control-based induction motors and potentially causing system instability. To enhance the convergence speed and stability of the observer and reduce discretization errors in the field-weakening region, this paper designs a feedback gain matrix suitable for high-speed field-weakening regions and conducts a study and summary of commonly used discretization methods. Discrete full-order flux observer models based on the forward Euler method, improved Euler method, and third-order Runge–Kutta method are designed. The discretization error, stability, and model complexity of the observers using these three discretization methods in the field-weakening region are analyzed. The experimental results demonstrate that the improved Euler method can achieve high discretization accuracy with relatively low computational complexity, making it a suitable discretization approach for full-order flux observers.

The intermittent and random nature of wind brings great challenges to the accurate prediction of wind power; a single model is insufficient to meet the requirements of ultra-short-term wind power prediction. Although ensemble empirical mode decomposition (EEMD) can be used to extract the time series features of the original wind power data, the number of its modes will increase with the complexity of the original data. Too many modes are unnecessary, making the prediction model constructed based on the sub-models too complex. An entropy ensemble empirical mode decomposition (eEEMD) method based on information entropy is proposed in this work. Fewer components with significant feature differences are obtained using information entropy to reconstruct sub-sequences. The long short-term memory (LSTM) model is suitable for prediction after the decomposition of time series. All the modes are trained with the same deep learning framework LSTM. In view of the different features of each mode, models should be trained differentially for each mode; a rule is designed to determine the training error of each mode according to its average value. In this way, the model prediction accuracy and efficiency can make better tradeoffs. The predictions of different modes are reconstructed to obtain the final prediction results. The test results from a wind power unit show that the proposed eEEMD-LSTM has higher prediction accuracy compared with single LSTM and EEMD-LSTM, and the results based on Bayesian ridge regression (BR) and support vector regression (SVR) are the same; eEEMD-LSTM exhibits better performance.

In a weak grid, the line impedance variation will cause the resonant frequency of the LCL filter to shift towards lower frequencies, thus reducing the quality of the grid-connected current and affecting the power grid stability. To solve this problem, a hybrid active damping strategy with feedforward compensation is proposed for the neutral point clamped (NPC) LCL grid-connected inverter system. In order to reshape the output conductance of the grid-connected system, suppress the resonance spikes of the LCL filter and improve the adaptability of the grid-connected system to the weak grid. A first-order low-pass filter is designed in the grid-connected current loop, and an active damping control of grid-connected current based on a first-order high-pass filter is also proposed. Compared with the conventional capacitive current active damping, no additional sensors are required, and the use of a differential is avoided, which reduces the high-frequency noise. The use of passive resistors is reduced, which reduces the power loss of the grid-connected system. In addition, a point of common coupling (PCC) voltage feedforward strategy based on a low-pass filter is designed to suppress the background higher harmonics at PCC and improve the quality of grid-connected current. In this work, the robustness of the system is analyzed when the parameters of the LCL filter change. Finally, the virtual space vector modulation strategy is used to balance the neutral voltage of the DC bus. Simulation and experimental results show that the control strategy can effectively improve the adaptability of the system to the weak power grid, improve the quality of grid-connected current, and demonstrate strong stability. The THD can be decreased by 0.2% at least, and the improvements are more significant with larger line impedance; the THD is only 2.33% even at 10 mH line impedance.

The accurate detection of wind power outliers plays a crucial role in wind power forecasting, while the inherited strong randomness and high fluctuations bring great challenges to this issue. This work investigates the way to improve the outlier detection accuracy based on support vector machine (SVM). Although SVM can achieve good results for outlier detection in theory, its performance is heavily dependent on the hyper-parameters. Parameter optimization is not an easy task due to its complex nonlinear multi-optimum nature; an improved Harris hawk optimization (IHHO) is proposed to optimize the parameters of SVM for more accurate outlier detection. HHO takes the cooperative behavior and chasing style of Harris’ hawks in nature called surprise pounce and can effectively search the optimal one in large parameter space, but it tends to fall into local optimum. To solve this issue, an improved Harris hawk optimization algorithm (IHHO) was proposed to obtain the optimal parameters of SVM. First, Hammersley sequence initialization is carried out to acquire good initial solutions. Then, a nonlinear factor control mode and an adaptive Gaussian–Cauchy mutation perturbation strategy are proposed to avoid getting trapped in local optima. In this way, a novel wind power outlier detection method named IHHO-SVM was constructed. The results on several wind power data with outliers show that IHHO-SVM outperforms SVM and HHO-SVM, which achieves the highest average F1 score of 96.63% and exhibits the smallest standard deviation. Compared to commonly used models for detecting outliers in wind power, such as isolation forest (IF), local outlier factor (LOF), SVM with grey wolf optimization (GWO-SVM), and SVM with particle swarm optimization (PSO-SVM), the proposed IHHO-SVM model shows the best overall performance with precision, recall, and F1 scores of 95.76%, 96.94%, and 96.35%, respectively.

Three-level NPC inverters have been widely used in grid-connected systems due to their superior performance compared with two-level inverters, but more switches lead to high fault probability. Meanwhile, the neutral point potential (NPP) fluctuation of the DC link is an inherent problem of three-level NPC inverters. To keep the three-level NPC inverter running stably after single-arm failure, a fault-tolerant control strategy based on an optimised space vector pulse width modulation (SVPWM) is proposed in this paper. Firstly, the common-mode voltage (CMV) of the postfault three-level NPC inverter is analysed and then the preliminary synthesis principles of the reference voltage vector are determined. Then, in order to ensure the NPP balance and the quality of the grid-connected currents, the reference voltage vector synthesis rules are optimised, a low-pass filter (LPF) and a hysteresis comparator are designed, respectively, to ensure the quality of grid-connected currents and effectively decrease the DC link NPP deviation. Finally, the simulation results show that the proposed fault-tolerant control strategy can realize the stable and reliable operation of the grid-connected three-level NPC inverter after single-arm failure, and the CMV can be reduced significantly, the quality of grid-connected currents is also improved. The proposed fault-tolerant strategy also shows good performance when the grid-connected currents change.

In the field-weakening region, the traditional field-weakening method for induction motor drives based on model predictive control (MPC) is to take a no-load operation as the premise and adjust the flux reference in the cost function proportional to the inverse of the rotor speed, which leads to poor torque output. This paper presents a novel field-weakening method for IM drives based on MPC. Considering the induction motor field-weakening limiting conditions and according to the speed adaptive field-weakening strategy with a voltage closed-loop, the speed adaptive field-weakening controllers were designed to optimize the references of the excitation current and torque current. In the rotor field-orientation d–q coordinate system, the stator flux amplitude and torque reference values were optimized by the optimal distribution current. Then, according to the dead-beat control principle, they were converted into an equivalent stator flux vector reference. Moreover, the stator voltage vector reference can be obtained. For an induction motor fed by a three-level neutral point clamped (3L-NPC) inverter, the cost function was constructed by combining all the constraints, including the voltage vector, the neutral potential balance, and the switching frequency. In this way, the high-performance field-weakening operation for the induction motor based on a model predictive control can be realized. The simulation and experiment results show that the proposed method can increase the torque output by 22% in the field-weakening region; at the same time, the steady characteristics and the dynamic response performance can be maintained well.

Traditional model predictive control (MPC) for the induction motor fed by a three-level inverter needs to explore 27 voltage vectors to obtain the optimal one, which leads to high switching frequency and requires too much computation. To solve this issue, a low switching frequency model predictive control with partition optimization is proposed. First, the reference voltage vector can be gained from the prediction model at the next time, and the space voltage vector plane is divided into 12 sectors for further vector choice. Furthermore, considering inverter constraints, the candidate voltage vectors are determined according to the sector location of the reference voltage vector. In this way, the candidate vectors can be reduced to 3 at most. Then, a boundary circle limit is designed to avoid unnecessary switch changes. If the reference voltage vector is within the boundary limit, the switches do not act, which can reduce the system switching frequency without introducing the extra weight coefficient into the cost function. These selected voltage vectors are substituted into the cost function to determine the optimal one. Finally, the neutral point voltage deviation is controlled by the positive and negative redundant small vectors to realize the multi-objective constraint without weighting coefficients. The simulation results show that the proposed control method can significantly reduce the switching frequency; at the same time, both the dynamic and steady performances can be maintained well, and the cost function has no weight coefficients.

With the rapid increase in wind power, its strong randomness has brought great challenges to power system operation. Accurate and timely ultra-short-term wind power prediction is essential for the stable operation of power systems. In this paper, an LsAdam–LSTM model is proposed for ultra-short-term wind power prediction, which is obtained by accelerating the long short-term memory (LSTM) network using an improved Adam optimizer with loss shrinkage (LsAdam). For a specific network topology, training progress heavily depends on the learning rate. To make the training loss of LSTM shrink faster with standard Adam, we use the past training loss-changing information to finely tune the next learning rate. Therefore, we design a gain coefficient according to the loss change to adjust the global learning rate in every epoch. In this way, the loss change in the training process can be incorporated into the learning progress and a closed-loop adaptive learning rate tuning mechanism can be constructed. Drastic changes in network parameters will deteriorate learning progress and even make the model non-converging, so the gain coefficient is designed based on the arctangent function with self-limiting properties. Because the learning rate is iteratively tuned with past loss-changing information, the trained model will have better performance. The test results on a wind turbine show that the LsAdam–LSTM model can obtain higher prediction accuracy with much fewer training epochs compared with Adam–LSTM, and the prediction accuracy has significant improvements compared with BP and SVR models.