The Control
System Journey

A sequential engineering tour inside the MATLAB/Simulink simulation — from raw weather data to the last drop of water.

SCROLL TO BEGIN
📡 DATA
☀️ PV
🧠 AI
📋 MPC
🛡️ FOPID
⚙️ VFD
💧 TANK

Climate Scenarios

72 continuous hours (3 days) per scenario. The Multiport Switch mechanism allows instant scenario switching.

Select a scenario:

Clearness Index (KT) — distinguishes natural sunset from passing clouds. 7 of 15 features are zero-cost physics. ℹ️
72-HOUR IRRADIANCE
PV ARRAY CONFIGURATION
22
PANELS
590W
PER PANEL
11S×2P
TOPOLOGY
~13kW
INSTALLED
12kW
CLIPPED
MPPT
EMBEDDED
⚡ Power Clipping: Strict 12kW cap protects pump motor from uncalculated peak currents. ℹ️

The Physical Heart

Complete mathematical physics model built from scratch. MPPT embedded as dynamic equations — instantaneous maximum power tracking.

The model generates realistic, weather-volatile, engineering-constrained power ready for AI analysis.

Hybrid Intelligence

CNN-BiLSTM compressed via Knowledge Distillation + Lottery Ticket Pruning

📦 24h History Input

AI cannot start from blank state. It ingests the previous 24 hours of 4 features as historical context.

🪟 Sliding Window

Each simulation second, the window slides forward — dropping the oldest reading, absorbing the newest. The model always has the latest temporal context.

🔮 5-Hour Recursive Horizon

Predict → feed back → predict next → repeat. Output: a ray of predictions extending 5 hours, enabling fully proactive MPC control.
SLIDING WINDOW ANIMATION
◻ HISTORY◼ WINDOW◼ PREDICTION

MPC Manager + Controller

Manager decides WHEN & HOW MUCH. Controller finds the SMOOTHEST PATH.

📋 MPC Manager (Strategy)

Safe Start: 5 min above 2400W. Smart Braking: √-based flow reduction. Anti Short-Cycling: 2-min minimum off-time.

🎯 MPC Controller (Execution)

Ts=60s, 300-step horizon (5hr). Output Weight Q: LOW. Rate Weight R: 20× HIGH. Inverse tuning eliminates oscillations.
RESPONSE COMPARISON
— PID (OSCILLATING)— MPC+FOPID (SMOOTH)--- TARGET
FOPID PARAMETERS
1.0
Kp
0.5
Ki
0.01
Kd
0.9
λ LAMBDA
0.7
μ MU
±400W
HYSTERESIS
🛡️ Hysteresis: ON > 2400W · OFF < 2000W — eliminates chattering at sunrise/sunset. ℹ️

FOPID: Shock Absorber

Fractional-Order PID with 2 extra degrees of freedom (λ, μ). Parameters optimized via Finite Difference gradient search.

Output: a single, silky-smooth control signal (P_control) sent to VFD — ensuring soft starts and no inrush currents.

VFD → Pump → Tank

⚙️ VFD (Inverter + Drive)

Soft Start: Calculated burst for stiction then smooth ramp. Hydraulic Saturation: Clips excess commands. Dead Zone: Below 690 RPM → full cutoff.

💧 Pump (State Observer)

Linear centrifugal pump model with mechanical saturation at 1550 RPM. Converts flow to speed for monitoring.

🏗️ Water Tank (Gravity Battery)

Stores excess solar energy as gravitational potential. (Q_in − Q_out) × (1/A) → ∫ → Level. Feeds back to MPC Manager → loop closed!
LIVE TANK SIMULATION
20%
▲ PUMP: 0.8 L/s
▼ DEMAND: 0.3 L/s
NET: +0.5 L/s
⟳ FEEDBACK → MPC MANAGER

The Loop Is Closed

From predicting clouds to delivering every drop of water — a system that anticipates nature.

CONCLUSIONS & VISION → EXPLORE CHAPTERS INTRODUCTION