S. Karatasou, Group of Building Environmental Research, Physics Department, University of Athens, Building PHYS-5, University Campus, 15784, Athens, Greece

M. Santamouris, Group of Building Environmental Research, Physics Department, University of Athens, Building PHYS-5, University Campus, 15784, Athens, Greece V. Geros, Group of Building Environmental Research, Physics Department, University of Athens, Building PHYS-5, University Campus, 15784, Athens, Greece

Abstract

Artificial neural networks (ANNs) have been used for the prediction of cooling loads of residential buildings in Athens, Greece. The investigation was performed for the summer period, where for Southern European countries, short time cooling load forecasting in residential buildings with lead times from 1 hour to 7 days can play a key role in the economic and energy efficient operation of cooling appliances.

The objective of this work is to produce a simulation algorithm, using ANNs, capable to forecast the following 24-hour cooling load profiles. Reliable cooling consumption measurements are required but are not usually available for residential buildings. State-of — the-art building simulation software, TRNSYS, was used to calculate energy demand for cooling for five selected apartments in Athens, Greece, using detailed building data (geometry, wall construction, occupancy etc) and Athens climate conditions. In order to validate simulation results and to have a more accurate modeling of difficult to estimate factors (like infiltration, absorptivity, inhabitants influence, etc), measurements of the indoor and ambient air temperature were performed during a period of two months, for April to May where neither heating nor cooling loads are demanding, and then used to calibrate the developed model of each house. Following this procedure, a database consisting of reliable cooling data was produced.

These data are used to train artificial neural networks in order to generate the relationship between selected inputs and the desired output, the next day building energy consumption for cooling. A multiplayer perceptron architecture using the standard back-propagation learning algorithm has been applied yielded to satisfactory results and the conclusion that when ANNs are trained on reliable data they can simulate the behavior of the building, thus they can be effectively used to predict future performance.

Introduction

In Southern European Countries, short term cooling load forecasting in residential buildings with lead times from 1 hour to 7 days can play a key role in the economic and energy efficient operation of cooling appliances.

In Greece, the operation of cooling has an important impact on the power demand profile. During the summer period, the highest peak power demand is approximately 2GW bigger than the bottom line peak, constant for the months April to May and October, when no cooling is in operation. The power profile between those months follows a similar pattern to the monthly average temperature profile. Furthermore, the fact that energy consumption for domestic use represents almost 35% of the total annual electricity consumption indicates that domestic load predictive models could be very useful from the Energy Utilities perspective.

To predict building energy consumption a large number of building software tools are available, making feasible to model a building for thermal evaluation and study it’s exact thermal behavior. Building thermal models which have been widely used in a variety of buildings and for a range of applications, in practice diversify on many factors: the modeling methodology, the physical laws, parameters and data that they encase, the integration of HVAC, passive solar, photovoltaic systems. Thus, depending on the application, these models vary on complexity and can be simple and easy to use, or more sophisticated and time-consuming to set-up and run.

In general, for the majority of applications, most of the appropriate software tools are time consuming and computationally heavy, especially when transient numerical methods are used. A large number of assumptions often need to be made when the quantitatively measurement of factors like infiltration or the estimation of parameters like occupancy is not possible. Also, parameters like the cost, the level of expertise and the exhaustive information needed to be collected could be prohibitive for a massive implementation.

Furthermore, almost all energy consumption predictive schemes are based on the prior prediction of weather data. As many weather variables are considered such as dry bulb temperature, relative humidity, solar radiation and cloudiness conditions, the most common practice is to use weather forecasts issued by meteorological centres, yet the direct link with such a centre make the procedure even more complicated. Artificial Neural Networks (ANN) can provide an alternative approach, as they are widely accepted as a very promising technology offering a new way to solve complex problems. ANNs ability in mapping complex non-linear relationships, have succeeded in several problems such as planning, control, analysis and design. They have been extensively used in the design, operation and fault detection of HVAC applications, in short time electric load forecasting, and in many fields of energy analysis and prediction.

In this study, a methodology based on a combination of an ANN and a thermal model is being used, in order to demonstrate the feasibility of using neural networks to forecast the following 24-hour cooling load profile for residential buildings. Neural network architectures are characterized by a collection of processing nodes connected by a series of weighted links. The relationship between the individual node’s inputs and outputs is typically a nonlinear function (for example a sigmoid function). A neural network can carry out complex calculations from global inputs to global outputs. In carrying out this process, and with the absence of a general theory for non-linear time — series prediction, good performance comes only by carefully analyzing available data (i. e. by using so-called ‘training data’).

Reliable cooling load data for at least one season are required but are not usually available for residential buildings, the focus of interest. Thus, dynamic building simulations were carried out with TRNSYS for five selected apartments in Athens, Greece, using known building data (geometry data, wall construction data, etc.) The simulation was carried out for a period of year, using as input for the climatic data the typical meteorological year. The training and forecast for the ANN model is then performed on the simulation results.