The management of energy consumption in residential buildings has been of great importance in the last decades, and the current debate on energy imbalance is also due to the lack of attention to this debate in previous years. Building Information Modeling (BIM), as one of the new technologies in the construction industry, provides the ability to analyze and optimize energy consumption by providing a digital building model. This article is a multi-stage research to complete the present research method. First, the Building Information Modeling (BIM) platform was completed, and then this platform was used to complete the database of applied deep learning methods. In this study, Design Builder software version 7.3 was used to simulate the thermal load, lighting energy consumption, and heating of educational buildings. Different neural networks were used in this research to implement the best deep learning method. Deep learning methods such as Support Vector Machine (SVM), Adaptive Neural Fuzzy Inference System (ANFIS), and other neural networks such as LSTM and XGBoost were used to increase the accuracy and efficiency of the model. The learning curves also indicated effective learning and high generalizability of the models. Finally, the most appropriate deep learning structure was introduced with a correlation coefficient of 0.98. This research showed that the use of BIM and its integration with advanced tools and techniques, such as energy simulation and optimization algorithms, has a significant impact on reducing energy consumption and increasing building efficiency, and emphasizes the importance of accurate data and intelligent design. This research focused on optimizing the energy consumption of buildings and was carried out in several phases using different techniques and tools. In the first phase, AutoCAD drawings were transferred to DesignBuilder software to estimate the annual energy consumption of the building. In this phase, standard materials and data related to topic 19 were used. This process served as a basis for subsequent analyses and providing energy consumption optimization solutions. Subsequently, by integrating smart windows and smart shades, an attempt was made to reduce energy consumption in various sectors. The implementation of the Artificial Neural Network (ANN) method for optimizing the energy consumption of air conditioning (HVAC) systems was carried out using PYTHON software. First, the database was collected according to Table (1) for the required data. Then, coding was done using Python software and the results of the coding. Implementation of the deep learning method are presented. For training and implementation of the neural network, the following data is collected.
The results indicate a significant reduction in annual energy consumption, which can lead to improved energy efficiency and cost savings. Comparing the results with the first phase, the energy consumption for lighting has decreased from 88.08 MWh to 52.58 MWh, which is equivalent to a 40.3% annual decrease, which is a significant change. On the other hand, the heating load has increased by 9% from 74.6 to 86.44 MWh, while the cooling load has decreased by 10% from 107.43 to 95.21 MWh; this decrease is due to the reduction in heat generated by the lamps. The results in this table show that the ANFIS neural network has greater accuracy and efficiency than the SVM. It is worth noting that in terms of coding complexity and coding execution, the SVM network has more advantages. Also, the combination of two types of neural networks has had a significant impact on increasing the accuracy of deep learning modeling, bringing the correlation coefficient to 0.98.
The results showed that this integration reduced the lighting energy consumption by 40.3% and the cooling load by 10%, while the heating load increased by about 9% due to the reduction in the heat generated by the lamps. The values of the equipment electricity consumption (Room Electricity) and domestic hot water (DHW) did not change, but overall these changes improved energy efficiency without major negative effects. A genetic algorithm (GA) was used to determine the best location of the cooling equipment. This algorithm identified the optimal coordinates considering the thermal comfort points (25 °C). However, some of the proposed coordinates were outside the acceptable room space, highlighting the need for position corrections. These corrections led to more accurate solutions with favorable approximations during algorithm execution. Finally, this research comprehensively demonstrated that the use of advanced techniques such as intelligent systems integration and optimization algorithms can lead to significant reductions in energy consumption and increased energy efficiency in buildings. Also, these results emphasize the importance of accurate input data and smart design to achieve effective and sustainable results. The results obtained showed that the proposed hybrid model has high accuracy and can be well used in predicting and managing energy consumption patterns of buildings. In addition to increasing the prediction accuracy, this model showed more flexibility than traditional methods and was able to examine the impact of variables such as temperature, humidity, and lighting hours on energy consumption more accurately.