Evolutionary Deep Learning-Based Energy Consumption Prediction for Buildings

Abstract

Today's energy resources are closer to consumers due to sustainable energy and advanced technology. To that end, ensuring a precise prediction of energy consumption at the buildings' level is vital and significant to manage the consumed energy efficiently using a robust predictive model. Growing concern about reducing the energy consumption of buildings makes it necessary to predict the future energy consumption precisely using an optimizable predictive model. Most of the previously proposed methods for energy consumption prediction are conventional prediction methods that are normally designed based on the developer's knowledge about the hyper-parameters. However, the time lag inputs and the network's hyper-parameters of learning methods need to be adjusted to have a more accurate prediction. This paper proposes a novel hybrid prediction approach based on the evolutionary deep learning (DL) method that is combining genetic algorithm with long short-term memory and optimizing its objective function with time window lags and the network's hidden neurons. The performance of the presented optimization predictive model is investigated using public building datasets of residential and commercial buildings for very short-term prediction, and the results indicate that the evolutionary DL models have better performance than conventional and regular prediction models.

Gated Recurrent Unit with Genetic Algorithm for Product Demand Forecasting in Supply Chain Management

“…In this paper, the raw dataset was preprocessed through the min-max normalization method. The equation of the min-max normalization method is presented as follows [24]: In Equation (8), is the original data of the raw dataset. and are the maximum value of the features and the minimum value of the features, respectively.…”

“…In this paper, the raw dataset was preprocessed through the min-max normalization method. The equation of the min-max normalization method is presented as follows [24]:…”

“…In addition, the crossover probabilities and mutation factors in the literature are generally set as 0.5-1.0 and 0.005-0.05, respectively [33]. On the basis of similar work [24], crossover probability and mutation factor were set at 0.7 and 0.015, respectively. Table 3 reports the performance of the six forecasting models.…”

“…The use of these meta-heuristic algorithms aims to obtain hyperparameters of NNs such as the number of input parameters, the number of hidden neurons, and batch size, etc. Almalaq and Zhang [24] focused on the future energy demand prediction for a building, which is an important problem in energy demand reduction. Due to this, they proposed a hybrid model which combined LSTM with GA, which optimizes window size and both the number of hidden layers and the number of neurons.…”

“…The size of the test set is fixed while the size of the training set increases for every fold. Testing the cross-validation offers a robust performance estimation of demand forecasting when each data is used for training and testing at each k-fold [24]. We utilized 10-fold cross-validation using the best hyperparameters of GA-GRU obtained in Section 4.1.…”