Solar-Power-Prediction-Tensorflow-LSTM-Model 0.1.1

A package for solar power prediction using LSTM in TensorFlow

AI-Driven Solar Power Prediction

Overview

This project focuses on forecasting solar power generation using LSTM (Long Short-Term Memory) neural networks and a Dense (fully connected) model. The goal is to improve solar energy efficiency, optimize power output, and predict inverter failures.

Features

• Time Series Forecasting: Uses past energy production data to predict future output.
• LSTM Model: Captures temporal dependencies in power generation.
• Dense Model: Provides a baseline for comparison.
• Scalability: Easily deployable for real-time applications.
• Preprocessing: Automated data normalization and feature engineering.

Dataset

The dataset contains solar plant energy readings with the following key features:

• DATE_TIME (Timestamp of the reading)
• DC_POWER (Generated DC power in kW)
• AC_POWER (Generated AC power in kW)
• DAILY_YIELD (Cumulative yield per day in kWh)
• TOTAL_YIELD (Total cumulative energy generation in kWh)

Installation

Ensure you have Python 3.8+ installed, then install the required packages:

pip install numpy pandas tensorflow scikit-learn

Usage

1. Train the Model

Run the following command to train the LSTM model:

from solar_power import train_model
model, scaler = train_model("data.csv")

This will:

• Load and preprocess the dataset.
• Train the LSTM model.
• Save the trained model as solar_power_lstm.keras.

2. Predict Power Output

After training, you can use the trained model for predictions:

import numpy as np
from tensorflow.keras.models import load_model

def predict_future_power(model_path, input_data):
    model = load_model(model_path)
    input_data = np.array(input_data).reshape(1, 24, -1)  # Reshape for LSTM
    return model.predict(input_data)

predicted_power = predict_future_power("solar_power_lstm.keras", sample_input)
print("Predicted Power Output:", predicted_power)

Model Architecture

LSTM Model

def build_lstm_model(input_shape):
    model = Sequential([
        Input(shape=input_shape),
        LSTM(64, return_sequences=True),
        Dropout(0.2),
        LSTM(32, return_sequences=False),
        Dropout(0.2),
        Dense(16, activation='relu'),
        Dense(1)
    ])
    model.compile(optimizer='adam', loss='mse')
    return model

• LSTM layers capture time-series dependencies.
• Dropout layers prevent overfitting.
• Dense layers refine predictions.

Dense Model (Baseline)

def build_dense_model(input_shape):
    model = Sequential([
        Input(shape=input_shape),
        Dense(128, activation='relu'),
        Dropout(0.2),
        Dense(64, activation='relu'),
        Dropout(0.2),
        Dense(32, activation='relu'),
        Dense(1)
    ])
    model.compile(optimizer='adam', loss='mse')
    return model

Performance Optimization

To reduce training time:

1. Use a smaller batch size (e.g., batch_size=16).
2. Enable GPU acceleration with TensorFlow (tensorflow-gpu).
3. Apply XLA optimization:

   import tensorflow as tf
   tf.config.optimizer.set_jit(True)
   

4. Use Early Stopping to stop training when validation loss stops improving:

   from tensorflow.keras.callbacks import EarlyStopping
   early_stopping = EarlyStopping(monitor='val_loss', patience=3, restore_best_weights=True)
   

Future Improvements

• Real-time deployment using Flask/FastAPI.
• Integration with IoT sensors for real-time solar panel monitoring.
• Hybrid models combining CNNs and LSTMs for better accuracy.

License

This project is licensed under the MIT License.

---

🚀 Start optimizing solar energy production today!