A hybrid model of self-organizing map and deep learning with genetic algorithm for managing energy consumption in public buildings

Intelligent applications have become essential for effectively managing energy consumption in public buildings. Energy management, especially for critical infrastructure buildings and public buildings, often serves as an example of sustainable practices since it tries to increase energy resilience through backup power systems. These structures were dealing with more difficulties, such as tight budgets, complicated regulations, behavioral issues, and others. Managing energy consumption in buildings is a multifaceted undertaking that entails grappling with non-uniform energy usage and a lack of established design guidelines for implementing energy-efficient and sustainable solutions. Consequently, analyzing energy utilization trends in public buildings and projecting future energy requirements is of paramount importance. This comprehension is indispensable to the identification and acknowledgment of energy consumption patterns in commercial and institutional buildings. Our research is highly significant for optimizing energy usage, reducing environmental impact, and making informed decisions in building management. We emphasize the importance of our predictive capabilities, sustainability, and life cycle assessment (LCA) techniques in achieving these objectives. Our hybrid model improves energy efficiency while also making a significant contribution to the larger objective of meeting sustainability standards in the built environment through thorough analyses of real-world data. The goal of this study is to determine the most sensible classification and forecasting scheme for the energy usage rates of public structures. The goal of this study is to determine the most sensible classification and forecasting scheme for the energy usage rates of public structures. In order to determine the number of energy consumption pattern clusters, a Self-Organizing Map (SOM) model was utilized in conjunction with Principal Component Analysis (PCA). The determination of clustering levels for each structure was executed by utilizing K-means in conjunction with a Genetic algorithm (GA). This approach enabled the identification of optimal clustering levels for the given structures, thereby facilitating effective analysis and interpretation of the data. The GA was specifically employed to determine the optimal centroid points for each cluster, thus optimizing the performance of the fitting model. Cluster analysis pattern extraction has made determining which buildings consume the most energy easier. As intelligent models for predicting energy usage, Convolutional Neural Networks (CNNs) and CNNs paired with a GA have also been used. The application of a GA to modify multiple settings of a CNN was conducted at this stage. The CNN model using a GA shows the method that was used provides higher accuracy and standard error compared to the conventional approach. It achieves a 94.01% accuracy on the training dataset and a 93.74% accuracy on the validation dataset with an error rate of 0.24 and 0.26 respectively. On the training dataset, the accuracy is 89.03% with a standard error of 0.3, whereas on the validation dataset, the accuracy is 88.91% with a 0.33 standard error. This research can help policymakers in the energy sector make better decisions regarding the timing of energy supply and demand for public buildings.