摘要
FPGA实时仿真中通过查表法建立的电机非线性模型存在通用性不高、精确度受限于片上存储资源等问题,为此建立了基于矩阵求逆的电机FPGA实时仿真模型。为了减小电感矩阵求逆对仿真实时性的影响,算法中选择了前向欧拉离散方法以及基于LDL T分解的矩阵求逆方法;模型框架中设计了电感矩阵更新及其求逆与电机方程更新并行处理的顶层框架,设计了以高速浮点乘累加器和浮点除法器为基本单元的底层计算架构;程序实现中采用了数据驱动的方法控制基本计算单元的任务调度。以Xilinx Virtex7 xc7vx485t为目标芯片,评估了以三相感应电机为例建立的电机模型FPGA程序,进行了电感参数定常与饱和两种工况下的离线仿真与实时仿真。所建电机FPGA实时仿真模型能够达到320 MHz的运行频率、793.75 ns的仿真步长,定子电流和转子电角速度的最大相对稳态误差分别为3.13%、0.70%。
In order to address the issues of low generality and accuracy limited by on-chip storage resources of electric machine nonlinear models established by the table look-up method in the current field programmable gate array(FPGA)-based real-time simulation,an electric machine model with matrix inversion is constructed.To reduce latencies induced by inductance matrix inversion,it chose discrete electric machine models with the forward Euler method,and the matrix inversion with the LDL T decomposition method.In terms of model framework,a top-level framework for parallel processing of inductance matrix updating and inversion as well as electric machine equation updating was designed,and a bottom-level computing architecture with high-speed floating-point multiplier accumulator and floating-point divider as the basic units was put forward.A data-driven method was used to control the task scheduling of basic computing units.With Xilinx Virtex7 xc7vx485t as the target device,the performance of the simulation model established by taking a three-phase induction motor as an example was evaluated,and its off-line and real-time simulation tests under both constant and saturated inductance parameters were completed.The proposed electric machine model can reach an operating frequency of 320 MHz,a simulation time step of 793.75 ns,and its maximum relative steady-state errors of stator current and rotor electric angular velocity are 3.13%and 0.70%,respectively.
作者
周斌
汪光森
李卫超
王志伟
王康
ZHOU Bin;WANG Guang-sen;LI Wei-chao;WANG Zhi-wei;WANG Kang(National Key Laboratory For Vessel Integrated Power System Technology,Naval University of Engineering,Wuhan 430033,China)
出处
《电机与控制学报》
EI
CSCD
北大核心
2023年第2期17-26,共10页
Electric Machines and Control
基金
国防科技重点实验室项目(6142217200305)。
关键词
FPGA实时仿真
电机非线性模型
矩阵求逆
LDL
T分解
数据驱动
浮点数
field programmable gate array-based real-time simulation
electric machine nonlinear model
matrix inversion
LDL T decomposition
data-driven method
floating point number
