Enhanced model-free predictive control using a Cuckoo search algorithm

This paper proposes a model-free predictive control method for nonlinear systems on the basis of polynomial regression. In contrast to conventional model predictive control, model-free predictive control does not require mathematical models. Instead, it uses the previous recorded input/output datasets of the controlled system to predict an optimal control input so as to achieve the desired output. The novel point in this paper is the improvement of existing model-free predictive control by adopting polynomial regression, which is a generalization of the so-called Volterra series expansion of nonlinear functions.

In this paper, we introduce model-free predictive control based on a polynomial regression expression for nonlinear systems. In contrast to conventional methods, model-free predictive control does not explicitly require a mathematical model of the controlled systems. In stead of the model, it utilizes massive stored and observed input/output dataset to predict the optimal control input. To date, the model-free predictive control method is based on linear regression vectors of the input/output data. Here we extend the method to polynomial regression vectors. By numerical simulations, we illustrate the effectiveness of model-free predictive control.

In order to solve the problem of large leakage current and poor parameter robustness of three-phase H8 inverter, an auxiliary power supply based nine switch transformerless photovoltaic grid-tied (AP-H9) inverter is proposed. The inverter limits the common mode voltage of the system to 2/3 times of the DC side voltage without increasing the system loss and control difficulty, so as to effectively suppress the system leakage current. At the same time, a model-free predictive current control method based on the ultra-local model is proposed. The gain and dynamic part of the ultra-local model of the inverter are updated online using the voltage and current information of the past two times, and then the predictive voltage vector of the next time is obtained. This method does not require the accurate model of the inverter and any system parameters. Finally, an experimental prototype with TMS320F28335 as the control chip is developed. The simulation and experimental results show that the proposed inverter and its control strategy can effectively suppress the leakage current of the photovoltaic grid connected system and improve the robustness of the system.

To address the issues of excessive current ripple and poor dynamic response in conventional angle position control (APC) for high-speed switched reluctance generator (SRG), this paper proposes an online parameter identification-based model-free predictive control (MFPC) strategy. First, the system dynamics are represented as an ultra-local model (ULM), enabling the design of an extended state observer (ESO) for two-step current prediction to compensate for control delays. Second, an improved Recursive Least Squares (RLS) algorithm with covariance resetting and error clearance is implemented to accurately identify dynamic inductance online, thereby enhancing the prediction accuracy of the ESO. Third, a bus current estimation-based adaptive feedforward compensation (AFC) technique is introduced to accelerate DC-bus voltage regulation and system dynamic response. Finally, simulations conducted on a 250 kW SRG platform demonstrate that the proposed method achieves superior dynamic performance and significantly reduced current ripple compared to conventional APC method.

The model-free predictive control (MFPC) is a new trend in the studies of the predictive control theory. In the MFPC, the classic model of the plant is substituted by an ultra-local model (ULM) that does not depend on the parameters and the uncertain variables of the system. In this manner, the model-free scheme improves the robustness of the predictive control system. This paper presents an integral sliding mode observer (ISMO)based MFPC for the induction motor (IM) drive. The proposed ISMO is utilized to establish an ULM for the IM. The ISMO guarantees the complete robustness of the system throughout an entire period. This feature is studied by using the Lyapunov candidate function. The implementation of the proposed MFPC is performed by the finite-set predictive current control (FSPCC). The proposed ISMO-based MFPC is evaluated through simulations and experimental tests.

Improved cascaded model-free predictive speed control for PMSM speed ripple minimization based on ultra-local model

Conventional model predictive control (MPC) of power converter has been widely applied to power inverters achieving high performance, fast dynamic response, and accurate transient control of power converter. However, the MPC strategy is highly reliant on the accuracy of the inverter model used for the controlled system. Consequently, a parameter or model mismatch between the plant and the controller leads to a sub-optimal performance of MPC. In this paper, a new strategy called model-free predictive control (MF-PC) is proposed to improve such problems. The presented approach is based on a recursive least squares algorithm to identify the parameters of an auto-regressive with exogenous input (ARX) model. The proposed method provides an accurate prediction of the controlled variables without requiring detailed knowledge of the physical system. This new approach and is realized by employing a novel state space identification algorithm into the predictive control structure. The performance of the proposed model-free predictive control method is compared with conventional MPC. The simulation and experimental results show that the proposed method is totally robust against parameters and model changes compared with the conventional model based solutions.

Traditional model predictive current control (MPCC) method depends on motor model for predictive control, when the motor parameters change with the working conditions, the predictive performance of MPCC will be deteriorated. To improve the parameter robustness of MPCC, a model-free current predictive control method that combines ultra-local model and sliding mode observer is proposed. First, the prediction model of MPCC based on the mathematical model of surface-mounted permanent magnet synchronous motor (SPMSM) is replaced by the ultra-local model that does not use any motor parameters. Second, the sliding mode observer is adopted to observe the parameter of ultra-local model and compensate parameter disturbance. Finally, the stability of the sliding mode observer is proved by the Lyapunov stability criterion. The traditional MPCC method and the proposed model-free current predictive control method are comparatively analyzed, simulation and experimental results show that the proposed model-free current predictive control method can improve the parameter robustness of MPCC.

Model-based predictive control that highly depends on system parameters has been widely investigated. In contrast, model -free predictive current control (MFPCC) can be performed based on input or output measured data, rather than on any system model information, hence eliminating the influence of parameter uncertainties. In such a strategy, the current gradients due to each of the possible voltage vectors are stored and used to predict future currents. In this paper, a model free predictive control method based on discrete space vector modulation is used, which increases the effective vector control set by constructing a virtual vector which is composed of linear combination of real vectors. The THD of grid connected current is reduced. Simulation and experimental results confirmed the effectiveness of the proposed method.

The application of Model-Free Predictive Control (MFPC) in power electronic systems has garnered increasing attention. In this paper, MFPC control based on replacing the classical factory model with an ultra-local model (ULM) is studied. Generally, the Integral Sliding Mode Observer (ISMO) is used to estimate the unknown part in the ULM where the non-physical factor in the ULM is selected with approximate values ​​ranging from the nominal value of the system which is contrary to the concept of MFPC control. In this research, an improved adaptive integral sliding mode observer (AISMO) based MFPC (AISMO-MFPC ) is proposed to estimate this factor with the unknown function in the ULM equation. The new observer design enables the estimation of this factor based on the current error, which allows for independent prposedcontrol of the system parameters.To obtain the lowest current ripple, the concept of active vector execution time (AVET ) has been incorporated into the proposed control where two vectors are selected in the sampling period to minimize the cost function instead of selecting a single vector. ULM is also used to calculate AVET which facilitates the implementation of the imposed control. The combination of the proposed AISMO-MFPC and AVET gives faster system response and reduces the current ripple and lower harmonics, especially in case of mismatch parameters. Finally, the effectiveness of the proposed control is confirmed under various conditions by the presented simulation and experimental results.

Supplying the open-end winding induction motor (OEWIM) with two typical voltage source inverters is an efficient mechanism to achieve a multi-level voltage on the stator. The asymmetric switching states of these two inverters generate a zero-sequence voltage (ZSV). The ZSV must be controlled in a way to prevent the production of zero-sequence current (ZSC) in the motor windings. This paper presents a finite-set model-free predictive control (FS-MFPC) to eliminate the ZSC. The proposed FS-MFPC utilizes an ultra-local model, which is independent of the system parameters. So, the robustness of the control scheme is improved against the uncertainties of the OEWIM. The ultra-local model of the proposed FS-MFPC is constructed by an extended state observer (ESO), which estimates the unknown function of the system. To avoid the complexity of applying ZSV in the modulation techniques, the proposed scheme is implemented in the finite-set predictive voltage control approach. The simulation results confirm that the proposed ZSV control has a good steady-state and dynamic performance.

Recently, recursive least square (RLS) based model free predictive control (MF-PC) has been widely applied to model estimation for superior performance to model-based method. In most previous researches the forgetting factor which is a tradeoff between the convergence speed and the steady-state tracking error is usually neglected, thus selected as a constant to ensure robustness. In this paper, a dynamic forgetting factor based bias compensated RLS (DFF-BCRLS) is proposed for model free predictive current control of a voltage source inverter. The dynamic forgetting factor is calculated by local optimal forgetting factor (LOFF) algorithm to improve data convergency and steady-state robustness. A bias compensation (BC) scheme is presented to further increase the accuracy of parameter estimation and immune input noise. The simulation results show that the proposed strategy has smaller tracking error in steady-state and better performance against model or parameter mismatches compared to the conventional RLS.

Model predictive power control (MPPC) is a powerful and popular control method for the control of three-phase AC/DC converters due to its merits of quick response and simple principle. However, conventional MPPC suffers from the high steady power ripples due to the application of only one voltage vector during one control period. Furthermore, the performance will deteriorate when there are parameter variations due to saturation, temperature, and so on. Recently, model-free predictive current control (MFPCC) based on current difference detection has been proposed to solve the problem of parameter robustness for the control of ac/dc converters. However, it cannot be directly applied to the power control of ac/dc converters due to the complicated relationship between the complex power and converter voltage. This paper proposes an improved model-free predictive power control (MFPPC) for three-phase ac/dc converters. Instead of using accurate mathematical model, the proposed method uses an ultra-local model to predict the complex power. Different from the conventional ultra-local model, both the gain of the input voltage and the other parts of the ultralocal model are estimated and updated online based on the input and output information of system. Hence, the proposed MFPPC is not affected by the parameter variations and exhibits strong parameter robustness. The effectiveness of the proposed method is confirmed by the presented simulation and experimental results.

The commutation torque ripple of the joint drive motor in the humanoid flexible arm provides significance to the stability of the arm joints. Here, we established a model of the joint drive system to explain the influence of driving motor output torque on joint motion state and analyzed the mechanism of commutation torque pulsation. This general model of the drive system was established based on a model-free predictive control method. By deducing the universal model equation of the arm joint drive motor control system and using it as the prediction model, the optimal switching state is selected by the value function to maintain the constant phase current value of the non-commutation winding of the drive motor. Then, by using MATLAB/Simulink, a simulation model was built to verify the effectiveness of the proposed method.

The emerging theory of model-free control (MFC) has presented an ultra-local modeling approach for systems, which is independent of the system parameters. This paper proposes a model-free based predictive voltage control for the induction motor (IM) drive. In this method, replacing the classic mathematical model of IM with an ultra-local model provides a good possibility of achieving a robust predictive control. Generally, MFC requires an online identification technique to design the parameters of the ultra-local model. In this paper, to avoid the complex designing procedure of MFC, a linear extended state observer (LESO) is utilized to construct the ultra-local model of IM. Moreover, the proposed scheme is implemented in finite-set predictive voltage control, which doesn’t require a modulation technique. The evaluation of the proposed method is made through the simulation.