Motor fcs mpc

Skill calebzu/pmsm-control-claude-skills-for-matlab/.claude/skills/motor-fcs-mpc

PMSM Claude Skills: methodology + skill library for AI-augmented MATLAB/Simulink modeling of PMSM control (FCS-MPC, DTC, SMC) with the Reference Model Learning Workflow; plus an extra FOC + load-torque-estimator reference

Install

npx -y skills add calebzu/pmsm-control-claude-skills-for-matlab --skill motor-fcs-mpc

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

PMSM Single-Vector Finite-Control-Set MPC Builder. Build an inner-loop finite-control-set MPC current controller for a three-phase voltage-source-inverter-driven PMSM (SPMSM/IPMSM via parameterization) in Simulink, with optional outer speed PI providing iq_ref. Use when constructing, reproducing, porting, or extending an FCS-MPC simulation in Simulink (keywords FCS-MPC, finite control set MPC, single-vector MPC, 7-vector / 8-vector MPC, predictive current control). Skip for FOC, DTC, sensorless, scalar control, BLDC trapezoidal, induction-motor MPC, multi-step horizon, weak-field, MTPA optimization, or pure theory questions. Layered on motor-pmsm-base.

SKILL.md

6.3 KB, as published. Nobody here has run it

motor-fcs-mpc — PMSM Single-Vector Finite-Control-Set MPC Builder

Three-phase 2-level voltage-source inverter + PMSM (SPMSM / IPMSM via parameterization). Inner-loop = FCS-MPC current control in dq frame. Outer-loop (optional) = speed PI providing iq_ref. Generalizes across machine sub-types by parameterizing Rs, Ld, Lq, ψf, Pn, J, F — no topology change between sub-types.

Layered on motor-pmsm-base. All base discipline applies (Goto TagVisibility, Vdc/BEMF rule, Visual 4-check, broken-FOC defense).

Must-Follow Rules

Plan first. Before any add_block, write a numbered plan: parameter table, design-decision choices (design_decisions.md), build-script structure. Get user approval.
One-click reproducibility. Inject all parameters via set_param(mdl, 'InitFcn', sprintf(...)). Model must Run from .slx double-click in fresh MATLAB session. See crit_conditions.md §J-CRIT.
MPC chart params via sprintf from build-script workspace. Chart hardcodes Rs/Ld/Lq/ψf/Pn/Tsc as numeric literals at build time. Never put MPC params in an external .m file (drifts from plant). See crit_conditions.md §K-CRIT.
Chart config = INHERITED + dual ZOH (NOT DISCRETE+Tsc). ch.SampleTime='-1', ch.ChartUpdate='INHERITED', leave Inputs DataType/Size at Inherit. Rate is locked by ZOH @ Tsc on every chart input + ZOH @ Tsc on chart output. See crit_conditions.md §G-CRIT.
Outer PI saturation is mandatory. LimitOutput='on', limits [-iq_max, +iq_max] with 1.5·Pn·ψf·iq_max ≥ 1.3·TL_max. Without saturation, runaway iq_ref collapses the cost-weight ratio and MPC abandons d-axis. See outer_pi_saturation.md.
ZOH every chart input even if upstream is already discrete. INHERITED chart's effective trigger rate is set by the fastest upstream signal; one un-ZOH'd input poisons the whole chart. See crit_conditions.md §A-CRIT.
Use From Workspace + inline matrix for ω_ref, NOT Step+RateLimiter. Continuous-mode RateLimiter escapes its slew limit on first solver step. See crit_conditions.md §H-CRIT.

Build Flow

Phase	Action	Reference
0	Validate inputs + sanity grid	base/pre_build_grid.md
1	Plant layer (powergui Discrete @ Ts, DC, UB Inverter, PMSM Salient-pole, TL Step)	(this skill)
2	Measurement layer (Bus Selector → Goto/From)	(this skill)
3	Outer speed PI (optional, RPM domain, mandatory saturation)	outer_pi_saturation.md
4	Inner FCS-MPC chart	algorithm_pseudocode.md + crit_conditions.md §G/§A/§K
5	Logging (To Workspace @ Tsc)	base/measurement_logger.md
6	Solver (Variable-step Auto, MaxStep ≤ Ts; powergui handles SPS at Ts)	(this skill)
7	InitFcn injection	base/plant_modeling.md §InitFcn
8	Self-tests + acceptance	acceptance_criteria.md

If issues arise, consult crit_conditions.md (A/D/G/H/J/K) and anti_patterns.md (15 common mistakes).

Required User Inputs

Ask the user before starting. Defaults in parameter_defaults.md.

Group	Parameter
Machine	`Rs` (Ω), `Ld, Lq` (H), `ψf` (V·s), `Pn`, `J` (kg·m²), `F` (N·m·s)
Power stage	`Vdc` (V)
Sampling	`Ts` (plant solver, ~1–5 μs), `Tsc` (control period, ~20–50 μs; `Tsc/Ts ≥ 10`)
Outer loop	`Kp_w, Ki_w` (RPM domain; recommend `pi_design.m`), `iq_max` (`1.5·Pn·ψf·iq_max ≥ 1.3·TL_max`)
MPC	control objective (`ripple-priority` / `balanced` / `torque-priority`), `λ_d, λ_q` per objective, `id_ref` (SPMSM: 0; IPMSM: ask)
Scenario	`StopTime`, `ramp_time`, `ω_ref_target` (RPM), `TL_step_time`, `TL_value` (`< 1.5·Pn·ψf·iq_max`)

Triggers / Skip

✅ Use	❌ Skip
Build / port / extend FCS-MPC simulation in Simulink	FOC, DTC, sensorless, scalar V/Hz, BLDC trapezoidal
Single-vector / 7-vector / 8-vector MPC, predictive current control	Multi-step horizon (N>1), delay compensation, weak-field, MTPA optimization
Generalizing FCS-MPC to a new PMSM machine parameter set	Pure theory questions about MPC math
	MATLAB perf (`matlab-performance-optimizer`) or unit tests (`matlab-test-creator`)

Generalization Across Machine Sub-Types

Sub-type	Parameter constraint	Strategy
SPMSM	`Ld == Lq`	`id_ref = 0` (no reluctance torque)
IPMSM mild saliency	`Lq > Ld`, `Lq/Ld ≤ 1.5`	`id_ref = 0` workable; MTPA gives ~5% torque-per-amp gain
IPMSM strong saliency	`Lq/Ld ≥ 2`	`id_ref` from MTPA solver (out of scope; ask user)

Topology does not change — same blocks, same wiring, same chart algorithm, same CRIT conditions. Only parameters and id_ref strategy differ.

Out-of-scope sub-types: SynRM (ψf ≈ 0), IM, BLDC trapezoidal — different prediction equations.

Sibling Skills

motor-pmsm-base — base infrastructure (this skill layers on it)
motor-dtc-pmsm — Direct Torque Control alternative
motor-smc-pmsm — Sliding Mode Control alternative