Sensorless Control of PMSM Drives using Kalman Filter for Speed and Load Estimation
Authors
Toft, Mads Wowk ; Aldous, Franz Alexander Rose
Term
4. term
Education
Publication year
2023
Submitted on
2023-06-02
Pages
95
Abstract
Electric motors are used in many devices. A common AC type is the permanent magnet synchronous motor (PMSM), known for high reliability. Drives often use an encoder to measure rotor position and speed, but sensors add cost and potential reliability issues. Sensorless control avoids the physical sensor by estimating position and speed from motor currents and voltages. This thesis examines whether adding a Kalman filter—a mathematical method that blends a model with noisy measurements to estimate hidden states—can improve an existing sensorless PMSM drive. In a laboratory setup, a PMSM-based drive was coupled to a load motor. An encoder was installed only to validate the results, not for control. The motor was modeled (in several reference frames), and a speed controller was designed and verified. As a baseline, a sensorless estimation structure with two phase-locked loops (PLLs) and a load observer was designed and validated. The structure was then modified to include a Kalman filter. Both approaches were tested and compared. The results show that the Kalman-based design improved the response to sudden load changes (better transient behavior).
Elektriske motorer findes i mange produkter. En udbredt vekselstrømsmotor er permanentmagnet-synkronmotoren (PMSM), som er kendt for høj pålidelighed. Mange drev bruger en encoder til at måle rotorens position og hastighed, men sensorer øger omkostningerne og kan give pålidelighedsproblemer. Sensorløs styring undgår den fysiske sensor ved at beregne position og hastighed ud fra motorens strømme og spændinger. Denne afhandling undersøger, om et Kalman-filter – en matematisk metode, der kombinerer en model med støjfyldte målinger for at estimere skjulte tilstande – kan forbedre en eksisterende sensorløs PMSM-styring. I en laboratorieopstilling blev et PMSM-baseret drev koblet til en belastningsmotor. En encoder blev kun brugt til at validere resultaterne, ikke til selve styringen. Motoren blev modelleret (i flere referencerammer), og en hastighedsregulator blev designet og verificeret. Som udgangspunkt blev der udviklet en sensorløs estimationsstruktur med to fase-låsesløjfer (PLL'er) og en belastningsobservatør. Derefter blev strukturen ændret, så et Kalman-filter indgik. De to metoder blev testet og sammenlignet. Resultatet viser, at Kalman-filteret gav en forbedret respons ved pludselige ændringer i belastningen (bedre transientadfærd).
[This apstract has been rewritten with the help of AI based on the project's original abstract]