Synchronous Motor Torque Estimation Using Temperature-Interpolated Flux
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the torque of synchronous electric machines are imprecise and resource-intensive, particularly in the automotive field, where they require complex adjustments and many devices, and are challenged by non-linear relationships between torque, supply currents, and magnetic flux components.
Innovation Solution
A method that estimates torque using a simplified approach involving measurements of stator and rotor currents, temperature, and interpolation of global flux values, reducing the need for complex mappings and device resources, and allowing precise torque estimation without extensive neural network training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single neural network is used to estimate torque from current measurements, then the method is simple to implement, but the estimation error is significant due to non-linear relationships
Solution Approach 1:
The patent divides the torque estimation problem into two separate neural networks: one for estimating direct-axis flux and another for quadrature-axis flux. This segmentation allows each network to specialize in capturing specific non-linear relationships, improving overall estimation accuracy while maintaining implementation simplicity.
Solution Approach 2:
The patent transforms the single-network approach into a multi-dimensional solution by creating separate estimation paths for different flux components (direct and quadrature axes). This dimensional separation enables independent optimization of each estimation path, better handling the non-linear relationships between currents and fluxes.
2Ease of manufacture
If on-board maps of magnetic flux components are constructed from a non-linear quasi-static model, then torque estimation is possible, but the estimation error is significant compared to bench measurements
Solution Approach 1:
The patent incorporates feedback mechanisms where the estimated flux values from the neural networks are used to continuously refine torque estimation. The system uses measured currents and estimated fluxes in a closed-loop manner to improve accuracy, compensating for deviations from bench measurements through adaptive estimation.
Solution Approach 2:
The patent changes the approach from using fixed non-linear quasi-static model parameters to using adaptive neural network parameters that learn optimal mappings from training data. This parameter adaptation allows the system to capture real-world non-linearities more accurately than theoretical models alone.
3Adaptability or versatility
If stator voltage measurements are used to estimate magnetic flux components, then flux estimation is possible, but the estimation error remains significant due to inverter non-linearities
Solution Approach 1:
The patent extracts the problematic inverter non-linearities from the estimation process by using current measurements directly rather than voltage measurements that pass through the inverter. This extraction removes the source of significant estimation errors associated with inverter voltage drops and non-linearities.
Solution Approach 2:
The patent uses neural networks to create accurate copies or models of the magnetic flux behavior based on current measurements, bypassing the need for voltage measurements that are corrupted by inverter non-linearities. The neural networks learn the flux-current relationship directly, creating a cleaner estimation path.
4Ease of operation
If the electric motor is oversized to remain within a linear range of flux variation, then flux estimation becomes easier, but the motor size increases which is undesirable in the automotive sector
Solution Approach 1:
The patent replaces the mechanical solution of oversizing the motor with a computational solution using neural networks. Instead of designing the motor to operate in a linear range, the system uses intelligent algorithms to accurately estimate flux and torque across the full non-linear operating range, maintaining compact motor dimensions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to estimating the torque (Tq) of an electric machine by: - establishing (E2) a direct component (Id) and a quadrature component (Iq) of the stator currents, and a rotor flux value ϕf(Τrot) corresponding to a measured temperature value (Trot) of the machine between a first predefined temperature (T1) and a second predefined temperature (T2), - obtaining (E3) a first equivalent total flux value (ρhiD(ϕ(Τ1))) at said first temperature (T1), - obtaining (E4) a second equivalent total flux value (ρhiD(ϕ(Τ2))) at said second temperature (T2), - interpolating (E5) an equivalent total flux (ρhiD(ϕ(Τrot))) at the measured temperature value (Trot) using the equivalent total flux values obtained, - estimating (E6) the torque (Tq) of the machine using the interpolated equivalent total flux, the quadrature component of the stator currents and the number of pairs of poles of the machine.