Motor Current Command Learning via Reinforcement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining optimal current control parameters for synchronous motors is challenging due to changes in inductance with rotation speed and current values, as well as magnetic saturation, making it difficult to achieve fast torque response under varying temperature, torque, and voltage conditions.
Innovation Solution
A machine learning device that observes state variables such as motor rotation speed, torque, and voltage, and learns optimal current commands using reinforcement learning techniques, including reward calculation and value function updates, to determine d-axis and q-axis current commands for three-phase alternating-current synchronous motors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional current control parameters are used for synchronous motors, then the control system is simple, but the torque response is slow and performance deteriorates under varying temperature, torque, and voltage conditions
Solution Approach 1:
The motor controller performs self-learning by autonomously observing its own operational states (current, voltage, speed, torque) and automatically updating control parameters through reinforcement learning, eliminating the need for external manual tuning and complex pre-configuration while achieving optimal torque response under varying conditions
Solution Approach 2:
The controller pre-learns optimal control policies by simulating various operating conditions during idle periods or low-load states, storing learned parameters that can be quickly applied when actual torque response is needed, thus preparing optimal control strategies in advance without affecting real-time performance
2Measurement precision
If current control parameters are determined according to motor characteristics, then control accuracy is maintained, but the determination process requires a large number of steps and is difficult due to inductance changes and magnetic saturation
Solution Approach 1:
The reinforcement learning mechanism continuously monitors feedback from motor performance (torque output, current consumption, voltage levels) and automatically adjusts control parameters in real-time, replacing lengthy manual determination processes with rapid iterative learning that adapts to inductance changes and magnetic saturation effects
Solution Approach 2:
The system dynamically changes control parameters (d-axis and q-axis current commands) based on learned relationships between motor states and optimal performance, automatically adapting to varying inductance and magnetic saturation conditions without requiring time-consuming manual recalibration for each operating point
3Adaptability or versatility
If manual tuning of current control parameters is performed, then some optimization is achieved, but it is difficult to obtain optimal parameters under varying temperature, torque, and voltage conditions
Solution Approach 1:
The motor controller autonomously adapts to varying temperature, torque, and voltage conditions by continuously learning from operational feedback, eliminating the need for manual intervention to adjust parameters under different environmental and load conditions while maintaining optimal performance across the entire operating range
Data Source
AI summary
A machine learning device which learns a current command for a motor, the machine learning device including a state observation unit which observes a state variable including a motor rotation speed or a motor torque command of the motor and at least one of a motor torque, a motor current, and a motor voltage of the motor; and a learning unit which learns the current command for the motor based on the state variable.