ElectroCNN: Regressive CNN-Based Energy Consumption Forecasting Leveraging Weather Data
Dharmi Patel, Mann Patel, Krisha Darji, Rajesh Gupta, Sudeep Tanwar, Jitendra Bhatia, Hossein Shahinzadeh
Abstract
Energy efficiency has become essential in the modern power sector. This research suggests an approach for identifying weather patterns to improve energy efficiency. ElectroCNN, which uses a Convolutional Neural Network (CNN) as its foundation, makes it easier to anticipate energy usage in a variety of meteorological scenarios. ElectroCNN modifies forecasts based on referred-to weather patterns to account for the three primary energy considerations associated with a given geographic location: industrial, commercial, and domestic. Four optimizers which are Adagrad, Adam, RMSprop, and FTRL used in the study to effectively improve prediction accuracy. In the context of model performance assessment, a stringent methodology is employed, integrating validation r2-score and validation loss curve to identify potential overfitting during the training process. Additionally, the study presents exhaustive curves for diverse optimizers such as Adagrad, Adam, RMSprop, and FTRL. These curves improve our comprehension of the model's performance under different optimization procedures by offering a thorough insight into the way every optimizer impacts testing samples in three distinct locations.