Energy Consumption Prediction Using CNN-LSTM Models: A Time Series Big Data Analysis of Electricity, Heating, Hot Water, and Water

With the advancement of energy management systems and smart grids, the accuracy of energy consumption prediction plays a crucial role in the efficient allocation of energy resources and cost reduction. In particular, the consumption patterns of various energy resources such as electricity, heating, hot water, and water in households can vary significantly depending on the time of day and day of the week. Effectively predicting these variations is essential for managing energy resources. Existing energy prediction models have primarily relied on simple time series analysis techniques, which have limitations in capturing complexity of consumption patterns and long-term dependencies, thus necessitating more sophisticated prediction models. The energy prediction management system aims to more accurately forecast hourly consumption of electricity, heating, hot water, and water using a CNN (convolutional neural networks) and LSTM (long short-term memory networks) hybrid model. The data preprocessing phase utilizes a moving average technique to detect outliers and improves model performance by adding feature interactions. The system applies a CNN-LSTM model to extract local patterns from time series data using 1D convolutional layers and to learn long-term temporal dependencies using LSTM layers. The CNN-LSTM model achieves high predictive accuracy by tuning various hyperparameters, demonstrating its potential to learn the complexity of energy consumption patterns effectively. This study proves that the CNN-LSTM hybrid model can improve the accuracy of energy consumption predictions. The potential is presented for contributing to the enhancement of smart grids and energy management systems. The model provides a new technique for the efficient management of energy resources and cost reduction and suggests its applicability to real-world energy management systems.