DigiBUILD project

Scientific Publications

Scientific publications

DigiBUILD research articles
A DATA ACQUISITION FRAMEWORK FOR BUILDING ENERGY MANAGEMENT

Date: 2023

Publication type: Research article.

Author(s): Konstantinos Touloumis, Panagiotis Kapsalis, Elissaios Sarmas, Stathis Stamatopoulos, Vangelis Marinakis and Haris Doukas

Keywords: Building Energy, Analytics, Big Data Pipeline, Energy Data Modelling, Building Data Processing

Abstract: Collecting data for providing energy analytics services has closely been dependent on Building Management Systems (BMSs). A BMS connected to a set of sensors and other electronic devices aims at observing a collection of rooms in terms of temperature, humidity, and other energy related metrics with the goal of maximining user comfort while preserving energy consumption. The tasks of collecting, preprocessing, storing and querying data on a BMS faces the same challenges with those of querying big data in terms of scalability, data heterogeneity, and integration. This paper presents a framework that addresses the problem of collecting sensor data in one enriched data warehouse which follows a common ontology model. Data from various data sources is homogenized by appropriate data preprocessing and feature extraction techniques. The framework allows timeseries-oriented querying with the target outcome of providing stakeholders high detailed analytics services for decision making on energy consumption optimization, building renovation and financing.

UNSUPERVISED DOMAIN ADAPTATION METHODS FOR PHOTOVOLTAIC POWER FORECASTING

Date: October 2023

Publication type: Research article.

Author(s): Loukas Ilias, Elissaios Sarmas, Vangelis Marinakis, Dimitris Askounis, Haris Doukas

Keywords: PV power forecasting, Photovoltaics, Deep learning, Unsupervised domain adaptation, Adversarial training, Domain adversarial neural network, Margin disparity discrepancy

Abstract: The accurate forecasting of photovoltaic (PV) power generation is of great significance in renewable energy systems, as it enables optimal energy management and grid stability. Despite the importance of this issue, substantial limitations still exist in the majority of existing research initiatives, which employ shallow machine learning algorithms. Recently, some studies have proposed employing convolutional and long short-term memory neural networks (LSTMs) in conjunction with transfer learning techniques; however, these approaches require that the production of PV systems is known during training. To overcome these limitations, we present the first study in the task of PV power forecasting utilizing unsupervised domain adaptation methods. Specifically, we employ two unsupervised methods, namely Domain Adversarial Neural Network and Margin Disparity Discrepancy. Both approaches use a source and a target domain during training, where the target labels of the target domain are unknown during training. We use production and weather data from seven PV systems with nominal capacities ranging from 23.52 kW to 271.53 kW, located in different areas. The findings demonstrate that our proposed architectures improve root mean squared error (RMSE), normalized RMSE, and scores over the smart persistence model across all the PV systems used for testing. Furthermore, our approaches improve the performance of the smart persistence model, with a forecast skill index reaching up to 45.35%. Our extensive experiments demonstrate that our introduced approaches offer valuable advantages over state-of-the-art ones, as the target variable of the target domain is unknown during training. We also demonstrate the robustness of our approaches by conducting a series of ablation experiments.

A PROGRAMMING MODEL FOR PORTABLE DATA-DRIVEN BUILDING APPLICATIONS

Date: November 2023

Publication type: Research article.

Author(s): Dimitris Mavrokapnidis, Gabe Fierro, Maria Husmann, Ivan Korolija, Dimitrios Rovas

Keywords: SeeQ, Programming model, Building applications, Software

Abstract: This paper introduces SeeQ, a programming model and an abstraction framework that facilitates the development of portable data-driven building applications. Data-driven approaches can provide insights into building operations and guide decision-making to achieve operational objectives. Yet the configuration of such applications per building requires extensive effort and tacit knowledge.

In SeeQ, we propose a portable programming model and build a software system that enables self-configuration and execution across diverse buildings. The configuration of each building is captured in a unified data model — in this paper, we work with the Brick ontology without loss of generality. SeeQ focuses on the distinction between the application logic and the configuration of an application against building-specific data inputs and systems. We test the proposed approach by configuring and deploying a diverse range of applications across five heterogeneous real-world buildings. The analysis shows the potential of SeeQ to significantly reduce the efforts associated with the delivery of building analytics.

SEMI-AUTOMATED EXTRACTION OF HVAC SYSTEM TOPOLOGY FROM IMPERFECT BUILDING INFORMATION MODELS

Date: September 2023

Publication type: Research article.

Author(s): Dimitris Mavrokapnidis, Georgios N. Lilis, Kyriakos Katsigarakis, Ivan Korolija, Dimitrios Rovas

Keywords: HVAC, BIM, Builder energy simulation models

Abstract: Knowledge of the HVAC system topology is required in several design and operation activities, including setting up building energy simulation models, fault detection and control applications, and facility management tasks. In theory, we can use the geometric context captured in Building Information Models (BIM) to infer these topological relationships; in practice, these models are far from perfect, with modelling errors and omissions. The implication is that manual remodelling is required, which results in unnecessary and redundant work. This paper discusses first the data quality issues faced in real-world data transformation workflows. We propose a methodology for generating good-quality HVAC topological models from imperfect BIM models. Our method’s core is a rule-based model-checking framework to identify missing elements. Once an issue is detected, a geometric-relationship enrichment tool is invoked to infer missing topological information. The proposed approach is tested in a real-world complex BIM model, and we discuss lessons learnt.

Follow us