Sensorless current prediction in single phase inverter circuits with machine learning algorithms

Abstract

Inverter circuits are widely used in power electronics applications such as electric motor control, induction heating or different Alternating Current (AC) loads. The control signal applied to the switching elements can affect the quality of the sinusoidal signal that occurs at the output of the inverter circuit by the means of voltage and current values. The inverter circuit topologies are generally designed as closed loop. However, these cause complexity of the circuit topology, increase the production costs and give difficulties in designing the control signals. In the present work, simulations of a single-phase inverter circuit with the Sinusoidal Pulse Width Modulation (SPWM) control signal are performed. Thus, the effect of the SPWM signals on the output of the inverter has been observed by changing the Modulation rate (M), the carrier signal frequency ( ) and the reference signal frequency ( ). Through the data obtained, the output current of a phase inverter without a sensor is estimated by Machine Learning Algorithms (MLA) such as Multi-Layer Perceptron (MLP), Support Vector Machine (SVM) and K-Nearest Neighbors (KNN).

FPGA implementation of KNN Algorithm-Based Prediction for DPWM Methods

“…The rotor bearings in the wind turbines are checked for faults using the KNN algorithm to increase energy production and decrease the maintenance cost [9]. The KNN algorithm is used to detect the output current of the inverter without a sensor by varying the Modulation rate, carrier frequency, and signal frequency of the Power Inverter [10].…”