Nonlinear adaptive normalized least mean absolute third algorithm for the control of five‐level distribution static compensator

Abstract

Summary This paper presents the design and control of single‐phase five‐level cascaded H‐bridge multilevel inverter (CHB‐MLI) as a shunt active power filter (SAPF) unit to mitigate the various power quality (PQ) problems such as reactive power burden, injection of harmonics in the supply, poor power factor, and so on. The control of CHB‐MLI has been done using the normalized least mean absolute third (NLMAT) algorithm, which overcomes the shortcoming of conventional least mean square (LMS), normalized LMS (NLMS), and least mean absolute third (LMAT) algorithm. The control algorithm is designed in MATLAB Simulink, and its scaled down prototype model is developed in the laboratory to find the fundamental active component of load current and to mitigate the aforementioned PQ problems. The proposed system is tested for single‐phase grid‐connection, and it works satisfactorily under steady state and dynamic load conditions. A dSpace1104 microcontroller is used to test the proposed algorithm and to generate firing pulses. Besides, the detailed comparison in the context of mean square error (MSE), weights convergence and intermediate error with conventional LMS, NLMS, and LMAT algorithm have also been discussed. Moreover, the harmonic distortion observed in the source current is within stipulated IEEE‐519 limits.

Design and development of leaky least mean fourth control algorithm for single‐phase grid‐connected multilevel inverter

“…For the calculation of the unit template, the grid voltage (v s ) along with its PS quadrature component is passed through peak estimation block as shown in Figure 6 to calculate v m as shown in Equations ( 20)- (22).…”

“…The DSTATCOM can be configured to solve PQ problems such as harmonics filtering, reactive power compensation, voltage support in low‐voltage distribution networks, load balancing, and lowering SPV power fluctuations using several control algorithms. These algorithms include synchronous reference frame theory (SRFT), 10 P‐Q theory, 11 salp swarm optimization algorithm (SSOA), 12 cascaded enhanced second‐order general integrator with a prefilter (CESOGI‐WPF) control technique, 13 LCL‐filter‐based DSTATCOM, 14 ANOVA Kernel Kalman filter (AKKF), 15 Hermite function‐based ANN, 16 optimized multi‐layer perceptron (MLP) neural network, 17 Bernoulli polynomial‐based control technique, 18 proportionate affine projection algorithm (PAPA), 19 Laguerre polynomial (LP)‐based algorithm, 20 instantaneous symmetrical components theory (ISCT), 21 normalized least mean absolute third (NLMAT) algorithm, 22 and immune feedback control algorithm 23 . The DC link of the DSTATCOM can also be used to interface solar photovoltaic (SPV) arrays to extract DC power to three‐phase system 24 .…”

Advanced adaptive algorithm controlled single‐phase DSTATCOM operation during weak grid conditions

“…Stochastic gradient descent‐based advanced adaptive controls, that is, least mean square (LMS) 19,20 and least mean fourth (LMF), 21 generalized integrator (GI), 22 and maximum correntropy criteria (MCC) 23 are among the noted adaptive controls have also evolved in their advanced derivatives to overcome their disadvantages. The LMS control offers less complexity and ease of implementation.…”

“…Adaptive neural network (ADALINE) control offers a better training procedure but suffers from the linear separability constraint of the perceptron. However, the hardware implementation of ADALINE requires a microcontroller with parallel processing 5–26 …”

“…17 The frequency domain controls, that is, fast Fourier transformation (FFT) and Kalman filter-based control algorithms, ensure better steady-state error at the cost of computational complexity and computational time. 5,18 Stochastic gradient descent-based advanced adaptive controls, that is, least mean square (LMS) 19,20 and least mean fourth (LMF), 21 generalized integrator (GI), 22 and maximum correntropy criteria (MCC) 23 are among the noted adaptive controls have also evolved in their advanced derivatives to overcome their disadvantages. The LMS control offers less complexity and ease of implementation.…”