Capstone-Project-3-Prediction-of-turbine-energy-yield-TEY-

Prediction of turbine energy yield (TEY) using Neural Networks

Problem Statement:

Prediction of turbine energy yield (TEY) using ambient variables and Turbine parameters as features

Project Overview:

Gas turbines are essential in many industrial applications, particularly for electricity generation. Monitoring and predicting their performance is crucial for optimizing efficiency and reducing emissions. In this project, a neural network model is developed to predict gas turbine energy yield and emissions from operational sensor data.

Dataset:

The dataset used for this project consists of sensor measurements from a gas turbine, recorded under various operating conditions. The variables include ambient factors like temperature, pressure, and humidity and turbine-specific readings such as exhaust pressure and turbine temperatures. The dataset contains 36733 instances of 11 sensor measures aggregated over one hour (utilizing average or sum) from a gas turbine. The Dataset includes gas turbine parameters (such as Turbine Inlet Temperature and Compressor Discharge pressure).

Variables:

1. Ambient temperature (AT): The temperature around the turbine (in °C).
2. Ambient pressure (AP): The pressure of the surrounding air (in mbar).
3. Ambient humidity (AH): The relative humidity of the surrounding air (in %).
4. Air filter difference pressure (AFDP): The pressure difference across the turbine's air filter (in mbar).
5. Gas turbine exhaust pressure (GTEP): The pressure at the turbine's exhaust (in mbar).
6. Turbine inlet temperature (TIT): The temperature at the turbine's inlet (in °C).
7. Turbine after temperature (TAT): The temperature after the turbine (in °C).
8. Compressor discharge pressure (CDP): The pressure at the compressor discharge (in mbar).
9. Turbine energy yield (TEY): The energy generated by the turbine (in MWH).
10. Carbon monoxide (CO): The concentration of CO emissions (in mg/m³).
11. Nitrogen oxides (NOx): The concentration of NOx emissions (in mg/m³).

Model Architecture

The neural network used for this project includes:

1. Input Layer: 8 variables (AT, AP, AH, AFDP, GTEP, TAT, CO, NOx).

2. Hidden Layers: Dense layers with ReLU activations.

3. Output Layer: Separate nodes predicting TEY.

Key Technologies:

1. Python: Core programming language for data processing and model building.

2. TensorFlow/Keras: For neural network model development.

3. NumPy/Pandas: For data manipulation and preprocessing.

4. Matplotlib/Seaborn: For data visualization.

5. Scikit-learn: For data preprocessing and evaluation metrics.

6. Pickle: For dumping model API

7. Streamlit: For developing the web application

View of the project:

Video.Software.1.mp4

Project working: