top of page
Search

PSO-Trained MPPT for Solar PV System

🧠 ANFIS Structure Used in the System

The adopted ANFIS structure (widely used in European intelligent control systems) consists of five layers:

🔹 Layer 1 – Input Membership Functions

  • Distributes inputs using Gaussian membership functions

  • Tunable parameters:

    • Center

    • Standard deviation

PSO Trained ANFIS MPPT for Solar PV system
₹4,127.00₹2,063.50
Buy Now

🔹 Layer 2 & Layer 3 – Rule & Normalization Layers

  • Rule firing strength calculation

  • Normalization of fuzzy rules

🔹 Layer 4 – Consequent Layer (Tunable)

  • Uses linear equations such as:

    f1=p1x+q1y+r1f_1 = p_1x + q_1y + r_1f1​=p1​x+q1​y+r1​

  • Tunable parameters: p, q, r

🔹 Layer 5 – Output Layer

  • Produces final crisp output

📌 Layer 1 and Layer 4 are tunable layers, and instead of conventional hybrid learning (backpropagation + least squares), Particle Swarm Optimization (PSO) is used for tuning.

🐦 Why Particle Swarm Optimization (PSO)?

PSO is used to:

  • Tune membership function parameters (center & deviation)

  • Tune consequent parameters (p, q, r)

  • Minimize Root Mean Square Error (RMSE) between predicted and target outputs

🎯 Objective: Achieve highly accurate MPPT behavior with minimal tracking error.

📊 Data Generation for Training

To train the ANFIS model:

🔹 Inputs

  • 🌞 Solar irradiation

  • 🌡️ Temperature

🔹 Output

  • ⚡ V<sub>MPP</sub> (Voltage at Maximum Power Point)

✔️ 1000 samples are generated using a MATLAB PV model✔️ Data is stored in a .mat file (PV data)✔️ Data is separated into training and testing datasets

🧪 PSO-Based ANFIS Training Process

🔧 Training Steps

  1. Load PV dataset

  2. Generate initial FIS using clustering

  3. Extract tunable parameters

  4. Define cost function (RMSE)

  5. Apply PSO algorithm:

    • Initialize population

    • Update velocity & position

    • Train ANFIS in each iteration

  6. Minimize RMSE over iterations

📉 Training Results

  • Training RMSE ≈ 4.47 × 10⁻¹⁵

  • Testing RMSE ≈ 5.64 × 10⁻⁵

✨ Extremely low error confirms excellent training accuracy.

💾 Exporting the Trained ANFIS Model

  • Final trained FIS is available in the MATLAB workspace

  • Exported as PSO-trained .fis file

  • This file is later used in Simulink MPPT implementation

⚙️ Simulink Model: PSO-Trained MPPT for Solar PV

🔋 Solar PV Panel Details

  • Rated power: 250 W

  • Open-circuit voltage: 37.3 V

  • V<sub>MPP</sub>: 30.7 V

  • I<sub>MPP</sub>: 8.15 A

📈 I–V and P–V characteristics confirm correct PV behavior under:

  • 1000 W/m² → 250 W

  • 800 W/m² → 200 W

  • 600 W/m² → 150 W

  • 400 W/m² → 100 W

🔄 Power Converter & MPPT Control

  • Boost converter connects PV to load

  • Duty cycle generated using:

    • PSO-trained ANFIS (MPPT)

    • PI controller

    • PWM generator

🎯 Goal: Always operate PV panel at MPP

🔍 Performance Analysis

✅ Case 1: Load Variation

  • Load changed every 0.3 s

  • PV consistently extracts maximum power

  • Smooth voltage & current response

  • Boost converter maintains power balance

✅ Case 2: Irradiation Variation

  • Irradiation levels: 1000 → 800 → 600 → 400 → 200 W/m²

  • MPPT accurately tracks new MPP each time

  • Extracted power matches theoretical PV characteristics

📊 Observed waveforms:

  • PV voltage & current

  • Boost converter voltage & current

  • Load power & PV power

✔️ MPPT performance remains robust under both load and irradiation changes.

⭐ Key Highlights

  • 🐦 PSO provides global optimization

  • 🤖 ANFIS offers intelligent nonlinear mapping

  • ⚡ Accurate MPPT under dynamic conditions

  • 📉 Very low RMSE

  • 🔋 Improved PV energy extraction efficiency

Comments

bottom of page