• Energy Research

AbstractIn this paper, a dataset of 92,906 dwellings was analysed adopting data mining techniques for the classification of heating and domestic hot water primary energy demand and for the evaluation of the most influencing factors. The sample was classified in three energy demand categorical variables (Low, Medium, High) considering different geometrical and physical attributes. The output of the model made it possible to set reference threshold values among the physical variables. Moreover, high energy demand dwellings were analysed in depth using a k-means algorithm in order to evaluate the design variables which need to be considered in a refurbishment process.

Abstract The energy management of buildings currently offers a powerful opportunity to enhance energy efficiency and reduce the mismatch between the actual and expected energy demand, which is often due to an anomalous operation of the equipment and control systems. In this context, the characterisation of energy consumption patterns over time is of fundamental importance. This paper proposes a novel methodology for the characterisation of energy time series in buildings and the identification of infrequent and unexpected energy patterns. The process is based on an enhanced Symbolic Aggregate approXimation (SAX) process, and it includes an optimised tuning of the time window width and of the symbol intervals according to the building energy behaviour. The methodology has been tested on the whole electrical load of buildings for two case studies, and its flexibility and robustness have been confirmed. In order to demonstrate the implications for a preliminary diagnosis, some unexpected trends of the total electrical load have also been discussed in a post-mining phase, using additional datasets related to heating and cooling electrical energy needs. The process can be used to support stakeholders in characterising building behaviour, to define appropriate energy management strategies, and to send timely alerts based on anomaly detection outcomes.

Demand Response (DR) programs represent an effective way to optimally manage building energy demand while increasing Renewable Energy Sources (RES) integration and grid reliability, helping the decarbonization of the electricity sector. To fully exploit such opportunities, buildings are required to become sources of energy flexibility, adapting their energy demand to meet specific grid requirements. However, in most cases, the energy flexibility of a single building is typically too small to be exploited in the flexibility market, highlighting the necessity to perform analysis at a multiple-building scale. This study explores the economic benefits associated with the implementation of a Reinforcement Learning (RL) control strategy for the participation in an incentive-based demand response program of a cluster of commercial buildings. To this purpose, optimized Rule-Based Control (RBC) strategies are compared with a RL controller. Moreover, a hybrid control strategy exploiting both RBC and RL is proposed. Results show that the RL algorithm outperforms the RBC in reducing the total energy cost, but it is less effective in fulfilling DR requirements. The hybrid controller achieves a reduction in energy consumption and energy costs by respectively 7% and 4% compared to a manually optimized RBC, while fulfilling DR constraints during incentive-based events.

Current efforts toward the necessary energy transition are predominantly focused on climate change mitigation in relation to decarbonization measures, mainly on the energy sector, but may not succeed in satisfying the goals of reaching the full sustainability of human activities, which should foster social equity, economic stability, and security of supply. Energy System Optimization Models, used as a key tool in guiding energy transition strategies through the formulation of energy scenarios, mostly focus on economic aspects and emissions reduction objectives only, completely neglecting the critical issues of the multifaceted “sustainability” concept. In response to that, the aim of this research is to develop an all-encompassing metric for evaluating the sustainability of decarbonization scenarios. It incorporates twelve key indicators pertaining to environmental, social, and security dimensions that are weighted and combined into a sustainability index (SI) for evaluating power sector technologies. The open-source TEMOA-Italy model is employed to create a baseline scenario and a decarbonization scenario. The computed evolution of the power sector is evaluated through a singular, multi-dimensional SI trend, enabling the monitoring of sustainability progress over time. The impact of alternative prioritization of the various sustainability factors is analyzed by exploring thousands of weights assigned to those factors within the SI. The obtained SI profiles are analyzed employing both unsupervised and supervised data analytics techniques, with the aim to extract and characterize the most representative patterns in terms of profile magnitude and trend. Eventually, explainable artificial intelligence (XAI) methods are implemented to understand the set of key indicators that mostly affect those two features of the SI profile. It turns out that the reliability of power system, geopolitical considerations, and land use play a pivotal role in influencing the SI trend and magnitude.

The paper introduces a novel methodology for optimizing the operation of a centralized Air Handling Unit (AHU) in a multi-zone building served by VAV boxes with interpretable rules extracted from a Deep Reinforcement Learning (DRL) controller trained to enhance energy efficiency and indoor temperature control. To ensure practical application, a Rule Extraction (RE) framework is developed, translating the DRL complex decision-making process into actionable rules using decision trees. A multi-action approach is proposed by developing three different regression trees for adjusting the supply water temperature, the position of the chiller valve, and the position of the economizer damper of the AHU. The extracted rules are benchmarked against the original DRL controller and two conventional control sequences based on ASHRAE 2006 and ASHRAE Guideline 36 within a high-fidelity co-simulation architecture combining Spawn of EnergyPlus and Python. The co-simulation environment uses EnergyPlus for building envelope and loads while HVAC components and controls are implemented in the equation-based modeling language Modelica. Results show that the RE-based controller closely approximates the performance of the DRL policy with an electric energy consumption only 3% higher, highlighting its ability to effectively mirror a more complex control logic, representing a transparent and easily implementable alternative. The controllers based on ASHRAE 2006 and ASHRAE Guideline 36 lead to higher energy consumption (for both chiller and fan) and violations of indoor temperature compared to both RE-based control and DRL. This study underscores the potential of integrating AI-driven control methods with interpretable rule-based systems, facilitating the adoption of advanced energy management strategies in real-world building automation systems.

In this paper, a tool for the detection and diagnosis of anomalous electrical daily energy patterns relative to a transformer substation of a university campus was developed and tested. Through an innovative pattern recognition analysis consisting in a multi-step clustering process, six clusters of anomalous daily load profiles were identified and isolated in two-year historical data of total electrical energy consumption. The infrequent electrical load profiles were found to be strongly affected, in terms of both shape and magnitude, by the energy consumption behaviour related to the heating/cooling mechanical room. Then, a fault-free predictive model, which uses artificial neural network (ANN) in combination with a Regression Tree, was developed to detect anomalous trends of the electrical energy consumption. The model was able to detect the 93.7% of the anomalous profiles and only the 5% of fault-free days were wrongly predicted as anomalous. Eventually, a diagnosis phase was conceived and validated with a testing data set. A number of daily abnormal load profiles were detected and compared with the centroids of the anomalous clusters identified in the pattern-recognition stage. The work led to the development of a flexible intelligent tool useful for operating a continuous commissioning of the campus facilities.

Recently, the spread of IoT technologies has led to an unprecedented acquisition of energy-related data providing accessible knowledge on the actual performance of buildings during their operation. A proper analysis of such data supports energy and facility managers in spotting valuable energy saving opportunities. In this context, anomaly detection and diagnosis (ADD) tools allow a prompt and automatic recognition of abnormal and non-optimal energy performance patterns enabling a better decision-making to reduce energy wastes and system inefficiencies. To this aim, this paper introduces a novel meter-level ADD process capable to identify energy consumption anomalies at meter-level and perform diagnosis by exploiting information at sub-load level. The process leverages supervised and unsupervised analytics techniques coupled with the distance-based contextual matrix profile (CMP) algorithm to discover infrequent subsequences in energy consumption timeseries considering specific boundary conditions. The proposed process has self-tuning capabilities and can rank anomalies at both meter and sub-load level by means of robust severity score. The methodology is tested on one-year energy consumption timeseries of a medium/low voltage transformation cabin of the university campus of Politecnico di Torino leading to the detection of 55 anomalous subsequences that are diagnosed by analysing a group of 8 different sub-loads.

The current EU energy efficiency directive 2012/27/EU defines the existing building stocks as one of the most promising potential sector for achieving energy saving. Robust methodologies aimed to quantify the potential reduction of energy consumption for large building stocks need to be developed. To this purpose, a benchmarking analysis is necessary in order to support public planners in determining how well a building is performing, in setting credible targets for improving performance or in detecting abnormal energy consumption. In the present work, a novel methodology is proposed to perform a benchmarking analysis particularly suitable for heterogeneous samples of buildings. The methodology is based on the estimation of a statistical model for energy consumption – the Linear Mixed Effects Model –, so as to account for both the fixed effects shared by all individuals within a dataset and the random effects related to particular groups/classes of individuals in the population. The groups of individuals within the population have been classified by resorting to a supervised learning technique. Under this backdrop, a Monte Carlo simulation is worked out to compute the frequency distribution of annual energy consumption and identify a reference value for each group/class of buildings. The benchmarking analysis was tested for a case study of 100 out-patient Healthcare Centres in Northern Italy, finally resulting in 12 different frequency distributions for space and Domestic Hot Water heating energy consumption, one for each class of homogeneous class of buildings. From the median value of each frequency distribution, reference values were extracted to be used in a benchmarking analysis. Beyond being flexible, open and upgradeable over time, a benchmarking analysis relying on both a sound statistical basis and on stochastic simulation is indeed able to overcome the limitations of the more common deterministic or one-dimensional benchmarking approach.