Energy use predictions with machine learning during architectural concept design

Abstract

Studies have shown that the actual energy consumption of buildings once built and in operation is often far greater than the energy consumption predictions made during design-leading to the term: 'performance gap'. An alternative to traditional simulation methods is an approach based on real-world data, where bahaviour is learned through observation. Display Energy Certificates (DECs) are a source of observed building 'behaviour' in the UK, and machine learning, a subset of artificial intelligence, can predict global behaviour in complex systems, such as buildings. In view of this, artificial neural networks, a machine learning technique, were trained to predict thermal (gas) and electrical energy use of building designs based on a range of collected design and briefing parameters. As a demonstrative case, the research focused on school design in England. Mean absolute percentage errors of 22.9% and 22.5% for thermal and electricity energy use predictions respectively were achieved. This is an improvement of 9.1% for the prediction of thermal energy use and 24.5% for the prediction of electricity energy use when compared to sources evidencing the current performance gap. Fuel Type and Sector Mean Design Total Energy Consumption (kWh/m 2 /yr) Mean Actual Total Energy Consumption (kWh/m 2 /yr) Design Prediction Error (%): 'Performance Gap' Thermal Offices 46 73 37 Education 57 84 32 Electricity Offices 71 121 41 Education 56 106 47 Data-driven Approach "A major hindrance in modelling real problems is the lack of understanding of their underlying mechanisms because of complex and nonlinear interactions among various aspects of the problem [...] in many cases, the best solution is to learn system behaviour through observations" (Samarasinghe 2007, p.1-2). In view of this, an alternative approach at predicting energy consumption to mathematical models based on building physics (traditional building energy simulation) is to collect large amounts of actual energy and

Minimising the Deviation between Predicted and Actual Building Performance via Use of Neural Networks and BIM

“…The maximum number of factors impacting the occupants' behaviour is modelled as five in the trained neural network, in line with what has been proposed in the literature (Clevenger andHaymaker, 2006, 2006;Toftum et al, 2009;Yan et al, 2015). Additional factors can be incorporated, and these could increase the prediction accuracy of the ANN, though it should be noted that increasing the number of input factors does not always correlate to an increase in accuracy, as was revealed in (Paterson et al, 2017).…”

“…In terms of applications of AI for building assessment and performance monitoring, [46] used neural networks to predict annual thermal and electrical energy use of buildings. Similarly, [47] adopted deep neural networks for space exploration in building designs that minimise the amount of energy required to operate the buildings.…”

Optimization of Performance Parameter Design and Energy Use Prediction for Nearly Zero Energy Buildings

“…For each fold, the ANN with the lowest mean squared error for the testing data was saved and the generalization errors were determined. MSE is given in Equation (1) [21]:…”

“…In order to identify the optimal solution from energy optimization functions and improve the accuracy and performance of data-driven energy modelling, Banihashemi [20] proposed a unified approach by introducing an improved hybrid energy objective function and integrating both ANN and Decision Trees, which expands the applicability and enables the model to deal with both continuous and categorical data. Paterson [21] explored an alternative way of predicting building energy consumption by using machine learning methods rather than following the traditional path of physics-based building performance simulation. In this presented model, previously collected thermal comfort and electricity consumption data subject to corresponding certain design parameters are used as inputs to train ANNs model.…”

Machine Learning-Based Short-Term Prediction of Air-Conditioning Load through Smart Meter Analytics

“…Accordingly, the consumption of air-conditioning systems has a significant impact on the electrical grid and the precise prediction of its variations can provide the grid management with notable benefits such as competitiveness in the day-ahead market, dispatch management, demand-side management and control optimization. Apparently, a straightforward solution in order to simulate the behaviour of buildings, and thus predicting the variations in their HVAC consumption, is developing physical models employing their geometrical and construction characteristics [4], infiltration properties, required ventilation rate, occupancy profiles and other details. However, grid management firms and utilities do not commonly have access to such details about the characteristics of their consumers' buildings.…”