Prediction of residential building energy consumption: A neural network approach

Biswas, M. A. Rafe; Robinson, Melvin; Fumo, Nelson

doi:10.1016/j.energy.2016.10.066

Cited by 252 publications

(102 citation statements)

References 21 publications

Supporting

Mentioning

102

Contrasting

Order By: Relevance

“…Biswas described, that the artificial neural network has emerged as a key method to address the issue of nonlinearity of building energy data and the robust calculation of large and dynamic data [49]. Pantazaras used incorporating CO 2 concentration as a factor in predictive models may unlock further optimization opportunities in controller applications, especially in buildings with highly varied occupancy, such as institutional buildings with the results, which suggest that there is indeed potential for at least short-term prediction using a very simple identification procedure [50].…”

Section: Second Part-the Optimized Artificial Neural Network Model Wimentioning

confidence: 99%

Monitoring of the daily living activities in smart home care

Vaňuš

Belesova

Martínek

et al. 2017

Hum. Cent. Comput. Inf. Sci.

View full text Add to dashboard Cite

One of the key requirements for technological systems that are used to secure independent housing for seniors in their home environment is monitoring of daily living activities (ADL), their classification, and recognition of routine daily patterns and habits of seniors in Smart Home Care (SHC). To monitor daily living activities, the use of a temperature, CO 2 , humidity sensors, and microphones are described in experiments in this study. The first part of the paper describes the use of CO 2 concentration measurement for detecting and monitoring room´s occupancy in SHC. In second part focuses this paper on the proposal of an implementation of Artificial Neural Network based on the Levenberg-Marquardt algorithm (LMA) for the detection of human presence in a room of SHC with the use of predictive calculation of CO 2 concentrations from obtained measurements of temperature (indoor, outdoor) T i , T o and relative air humidity rH. Based on the long-term monitoring (1 month) of operational and technical functions (unregulated, uncontrolled) in an experimental Smart Home (SH), LMA was trained through the data picked up by the sensors of CO 2 , T and rH with the aim to indirectly predict CO 2 leading to the elimination of CO 2 sensor from the measurement process. Within the realized experiment, input parameters of the neuronal network and the number of neurons for LMA were optimized on the basis of calculated values of Root Mean Squared Error, the correlative coefficient (R) and the length of the measured training time ANN. With the use of the trained network ANN, we realized a strictly controlled short-term (11 h) experiment without the use of CO2 sensor. Experimental results verified high method accuracy (>95%) within the short-term and long-term experiments for learned ANN (1.6.2015ANN (1.6. -30.6.2015. For learned ANN (1.2.2014ANN (1.2. -27.2.2014) was verified worse method accuracy (>60%). The original contribution is a verification of a low-cost method for the detection of human presence in the real operating environment of SHC. In the third part of the paper is described the practical implementation of voice control of operating technical functions by the KNX technology in SHC by means of the in-house developed application HESTIA, intended for both the desktop system version and the mobile version of the Windows 10 operating system for mobile phones. The resultant application can be configured for any building equipped with the KNX bus system. Voice control implementation is an in-house solution, no third-party software is used here. Utilization of the voice communication application in SHC was proven on the experimental basis with the combination of measurement CO 2 for ADL monitoring in SHC.

show abstract

Section: Second Part-the Optimized Artificial Neural Network Model Wimentioning

confidence: 99%

Monitoring of the daily living activities in smart home care

Vaňuš

Belesova

Martínek

et al. 2017

Hum. Cent. Comput. Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…Figure 7 presents projections (average for 10 iterations) resulted from the DmGNn models. Learning models were extensively applied in the case of NG demand predictions [62,136]. Some competitive prediction models were selected to compare outputs of the proposed model and analysis of the accuracy.…”

Section: Outputs and Resultsmentioning

confidence: 99%

Developing a Data Mining Based Model to Extract Predictor Factors in Energy Systems: Application of Global Natural Gas Demand

et al. 2019

View full text Add to dashboard Cite

Recently, the natural gas (NG) global market attracted much attention as it is cleaner than oil and, simultaneously in most regions, is cheaper than renewable energy sources. However, price fluctuations, environmental concerns, technological development, emerging unconventional resources, energy security challenges, and shipment are some of the forces made the NG market more dynamic and complex. From a policy-making perspective, it is vital to uncover demand-side future trends. This paper proposed an intelligent forecasting model to forecast NG global demand, however investigating a multi-dimensional purified input vector. The model starts with a data mining (DM) step to purify input features, identify the best time lags, and pre-processing selected input vector. Then a hybrid artificial neural network (ANN) which is equipped with genetic optimizer is applied to set up ANN’s characteristics. Among 13 available input features, six features (e.g., Alternative and Nuclear Energy, CO2 Emissions, GDP per Capita, Urban Population, Natural Gas Production, Oil Consumption) were selected as the most relevant feature via the DM step. Then, the hybrid learning prediction model is designed to extrapolate the consumption of future trends. The proposed model overcomes competitive models refer to different five error based evaluation statistics consist of R2, MAE, MAPE, MBE, and RMSE. In addition, as the model proposed the best input feature set, results compared to the model which used the raw input set, with no DM purification process. The comparison showed that DmGNn overcame dramatically a simple GNn. Also, a looser prediction model, such as a generalized neural network with purified input features obtained a larger R2 indicator (=0.9864) than the GNn (=0.9679).

show abstract

“…The process of training the MFFNNBP depends on the backpropagation process, which calculates the gradient decent error between the target output and the predicted output considering the new weights each time.The Levenberg-Marquardt method involves an iterative improvement of values in order to reduce the sum of the squares of the errors. The Levenberg-Marquardt use gradient descent method and the Gauss-Newton method to solve problems [20]. In the gradient descent method, the sum of the squared errors is reduced by updating the values in the direction of the greatest reduction of the least squares objective, and the Gauss-Newton method reduces the sum of the square errors to finding the minimum of the quadratic [13].…”

Section: Proposed Modelmethodsologymentioning

confidence: 99%