AI Challenge, Water Forecasting via Deep Learning
Water Forecasting
This project focused on developing a forecasting algorithm for a water treatment plant control company, as part of an Industrial AI challenge competition. The goal was to accurately predict incoming water volume at the plant's intake. This prediction plays a crucial role in optimizing the activation cycles of bacteria through air pumps during the water cleaning process, ensuring efficient and effective treatment.
The project involved testing and evaluating multiple algorithms such as XGBoost, Temporal Convolutional Network (TCN), and Long Short-Term Memory (LSTM). These algorithms were trained on historical data to capture patterns in water flow rates, weather conditions, seasonal variations, and other relevant factors. Feature engineering techniques were applied to enhance the algorithms' ability to extract meaningful insights from the data.
After rigorous training and fine-tuning, the algorithms were extensively evaluated to compare their forecasting accuracy using various performance metrics. This thorough analysis aimed to identify the most effective algorithm for predicting water volume intake, thereby supporting optimal operational decisions in water treatment processes.