Comparative Analysis of Deep Learning Models for Predictive Energy Management: A Case Study of 1st Avenue Estate, Ikoyi, Lagos

Electricity has always been an important factor for technological progress, and finding cleaner, more efficient generation methods has become more important than ever. In Nigeria, diesel generators are widely used due to their durability, better torque, and lower fuel consumption compared to petrol generators. However, with fuel scarcity and rising costs, optimizing fuel usage is crucial. This study focuses on predicting diesel consumption for 1st Avenue Estate, Ikoyi, Lagos, using deep learning. Data on diesel consumption from 2020 to 2024 was collected to create a predictive model for future use. The estate, powered by three generators, includes 36 flats and a water treatment plant. Three models were used for the diesel consumption forecast and each of them were evaluated with Mean Squared Error (MSE), Mean Absolute Error (MAE) and the Coefficient of evaluation. The model performance for the LSTM were 258939.57,400.357, and -0.094, the Bi-directional LSTM had a result of 265807.09,410.66 and -0.123, while the CNN had 236370.27,388.485 and -0.0058 in the MSE, MAE and COE respectively. Nevertheless, CNN outperformed other models adopted in this study according the performance evaluation The study offers insights into model effectiveness in diesel consumption forecasting by comparing three different models using performance metrics. The predictive model aids in optimizing fuel purchases, which addresses fuel scarcity and rising costs by enabling a more efficient diesel supply planning. This can be used to improve maintenance schedule and financial planning to aid Facility Managers.