2MW PMSG Generator — Digital Twin Inference Engine

Multi-Physics Simulation • Electromagnetic • Thermal • Aerodynamic • Control Systems

Live Simulation
t = 0.0 s
Agent Interaction
System Overview
Thermal Network
Electromagnetic
Control System
Agent Inferences
--
Wind Speed (m/s)
--
Power Output (MW)
--
Efficiency (%)
--
Winding Temp (°C)
--
Total Loss (kW)
--
Rotor (RPM)

2 MW Direct-Drive PMSG — System Schematic

PMSG ⚡ -- MW Heat loss -- kW Wind -- m/s -- RPM

System Architecture & Methodology

Physics-Based Inference

All outputs are derived from first-principles physics equations — no machine learning. Each parameter is a deterministic solution computed every simulation step.

Multi-Agent Architecture

5 specialized agents each solve their physics domain and communicate via message bus:

Wind → Aero → Control → EM → Thermal
↳ feedback to Control & EM

Key Specifications

Rated Power2.0 MW
GeneratorDirect-Drive PMSG
Blade Radius63 m
Hub Height80 m
Pole Pairs40
Thermal Nodes9 (LPTN)
EM Update Rate100 Hz

Real-Time Data Stream

All values, diagrams, and charts update continuously in sync with the simulation timeline — representing 300 seconds of turbine operation.

Power & Wind — Live Trace

Multi-Agent System Architecture

Five physics-domain agents communicate through a publish-subscribe message bus. Each agent independently solves its equations and publishes results that downstream agents consume. The diagram below shows real-time message flow — arrows pulse as data passes between agents.

PUBLISH-SUBSCRIBE MESSAGE BUS 🌬 Wind Profile 1 Hz | IEC 61400-1 v = -- m/s 🌀 Aerodynamic 10 Hz | BEM Theory Cp = -- 🎛 Control 100 Hz | MPPT+Pitch T_ref = -- kNm ⚡ Electromagnetic 100 Hz | dq-Frame P = -- MW 🌡 Thermal 1 Hz | 9-Node LPTN T_w = -- °C v_wind ρ T_aero Cp, λ T_ref ω_m P_loss I, V Thermal Feedback: T_winding → Torque Derating Thermal Feedback: T_winding → Rs(T) Adjustment Orchestrator Agent Sequence & Timing Control

Message Protocol

FromToMessageRate
Wind ProfileAerodynamicwind_speed, air_density, direction1 Hz
AerodynamicControlaero_torque, Cp, tip_speed_ratio10 Hz
ControlElectromagnetictorque_ref, rotor_speed, pitch100 Hz
ElectromagneticThermalcopper_loss, iron_loss, magnet_loss100 Hz
ThermalControlwinding_temp (derating trigger)1 Hz
ThermalElectromagneticwinding_temp (Rs adjustment)1 Hz

Live Agent State

AgentStatusLast Output

Message Log

🌡 9-Node Thermal Network (Live)

Winding --°C Magnet --°C End Wind --°C Tooth --°C Yoke --°C Shaft --°C Frame --°C Ambient 25°C R_wm R_wt R_my R_ty R_conv 🔥 P_Cu 🔥 P_mag 🔥 P_Fe

Thermal State (Dynamic)

Temperatures update as simulation plays. C_i × dT_i/dt = Q_i + Σ(T_j - T_i)/R_ij

ComponentTempLimitMargin

Heat Sources (Current Step)

Temperature History (Live Trace)

⚡ dq-Frame Equivalent Circuit

d-axis circuit V_d Rs(T) Ld ωLq*iq id = 0 A q-axis circuit V_q Rs(T) Lq ωψpm iq = -- A Torque: T_e = 1.5 × p × ψ_pm × iq T_e = -- kNm Power: P = -- MW

Operating Point (Dynamic)

ParameterValue

Loss Breakdown

Power & Efficiency (Live)

Loss Components (Live)

🎛 Control Loop Diagram (Live)

Wind -- m/s Aero Cp=-- CONTROLLER Region 2: MPPT T=K_opt*ω² T_ref -- kNm Pitch --° Rotor Dynamics J*dω/dt = T_aero - T_em -- RPM Thermal Derating Active if T_w > 130°C T_winding feedback

Control State (Dynamic)

ParameterValue

CONTROL REGIONS

Region 2 (v < 11.4 m/s): MPPT, T_ref = K_opt × ω²

Region 3 (v > 11.4 m/s): PI pitch control to limit power

Derating: Reduce torque if T_winding > 130°C

Speed & Pitch (Live)

Torque & Cp (Live)

🔬 All Agent Inferences (Live)

Values update dynamically as simulation plays. Each row shows the physics equation that produces the value.

🌬 Wind Profile Agent

Kaimal turbulence spectrum (IEC 61400-1), AR(1) temporal correlation, 1 Hz update

OutputValuePhysics

🌀 Aerodynamic Agent

Cp(λ,β) parametric surface, P = ½ρAv³Cp, 10 Hz update

OutputValuePhysics

🎛 Control Agent

MPPT + PI Pitch + Thermal Derating, 100 Hz update

OutputValuePhysics

⚡ Electromagnetic Agent

dq-frame PMSG + Steinmetz iron loss + thermal coupling, 100 Hz update

OutputValuePhysics

🌡 Thermal Agent

9-node LPTN, Forward Euler, 1 Hz update

OutputValuePhysics