Building Clusters Research Articles

Data-Driven Aggregate Thermal Dynamic Model for Buildings: A Regression Approach

The thermal inertia of buildings brings considerable flexibility to the building heating and cooling loads, which is believed to be a promising demand response resource in energy systems. However, it is challenging to utilize the thermal inertia of buildings in the operation of energy systems because of the complicated thermal dynamics and high computational cost. This paper proposes a data-driven aggregate thermal dynamic model (ATDM) for the multi-zone building and building cluster, respectively, which offers an equivalent and low-complexity building model for the operation and control of energy systems. The ATDM consists of the aggregation equation and the state equation. The former projects the detailed real states of buildings into the characteristic state (i.e., aggregate state) using an affine function, and the latter describes the thermal dynamics of buildings using the aggregate state. The ATDM is formulated for two practical load control strategies for the building cluster, including direct load control and indirect load control. Then, the constrained nonlinear regression model is proposed to estimate the model parameters and occupant behavior, for which an efficient algorithm based on the block coordinate descent method is developed by exploiting the decomposable structure of the regression model. Simulation results based on real-world data show that the root mean square error and mean absolute percentage error for the multi-zone building (or building cluster) are below 0.72 °C and 1.44% (or 0.32°C and 1.39%), respectively, verifying the effectiveness of the proposed methods.

Circularity of building stocks: modelling building joints and their disassembly in a 3D city model

Improving circularity in the building sector entails ensuring greater material efficiency to avoid virgin material extraction. This effort could reduce net negative environmental impacts related to building materials. The circular use of building elements requires to assess which elements from the building stock can be salvaged for re-use, repair, etc. Hereto, products and materials are to be inventoried, and info about their assembly is crucial to assess their disassembly potential. Assembly refers to the relative geometric positioning of building elements and their connections. Assemblies, such as building joints, determine the actions and costs necessary to salvage products in a damaged or undamaged state from a building. Depending on this state, the products can be re-used or recycled. In other words, to understand and alter salvaged material flows, adequate ways to model building joints and disassembly are needed. Additionally, an approach to store data on these building joints on the scale of clusters of buildings and cities should be developed. This would enable to gain insight in the salvage potential of building products for any size of building stock, and to determine the stock’s circularity potentials. To this end, the study consists of four steps. (i) Properties of building components that affect disassembly are discussed, including their storage in material passports. (ii) Methods to model the disassembly sequence of building components are reviewed and fit disassembly modelling methods are selected. (iii) These methods set the data necessary on building joints, which can be subdivided in geometry, connection means and dependencies. (iv) A proposal is made to store building joint data in a 3D city model, such as a 3D GIS model. In this workflow, a novel database is proposed that relates data on building joints and building elements to unique codes. In the 3D city model, codes of building elements can be attached to digital surfaces, and codes of building joints to edges of surfaces. Aside from storing building joint data in 3D GIS, we discuss applying this in BIM. To conclude, this study presents a novel workflow to model building joints and their disassembly, which enables the modelling of building product salvaging for any size of building stock.

Financial Fraud Detection on Social Networks Based on a Data Mining Approach

The article summarizes the arguments and counter-arguments within the scientific debate on the issue of researching financial frauds in the Internet. The main goal of the research is to develop methodological principles for identifying financial cyber fraud in social networks based on the analysis of comments to identify relevant text patterns that may indicate manipulation attempts and further fraud. The urgency of solving this scientific problem is due to the fact that the mass involvement of Internet users in social interactions in the virtual environment has contributed to the development of various criminal schemes, as well as personal data that is initially entered during registration and information that is published in social networks can be used by a fraudster to carry out illegal financial transactions. The study of the issue of identifying financial fraud in social networks in the article is carried out in the following logical sequence: collecting comments with a corresponding request under publications in the social network using the Instaloader tool; combining comments into groups based on content similarity; conducting preliminary processing of text data (decomposing the text into simpler components (tokens) and reducing similar word forms to their main dictionary form); determination of the level of similarity of text data using the cosine of similarity; building clusters of text data that can indicate the presence of signs of financial fraud under relevant comments in social networks. Instagram was chosen to identify fraudulent operations in social networks. The analysis of comments on the social network Instagram to identify text patterns showed that offers and appeals from specific groups of people and promoted in comments with the help of spam are dangerous. Based on the results of the study, it was concluded that national regulators need to strengthen public control of the Internet, as well as improve the security system at the technical level by using the latest machine learning methods to identify attempts to commit illegal actions with the subsequent imposition of sanctions on such users in social networks.

Block-based variations in the impact of characteristics of urban functional zones on the urban heat island effect: A case study of Beijing

The spatial aggregation of buildings intensifies the urban heat island (UHI) effect and hinders the sustainability of a city. This study divided more than 200,000 buildings in Beijing's Fifth Ring Road area into four categories, and used geographically weighted regression (GWR) to study the spatial heterogeneity of the impact of building features on land surface temperature (LST). The results showed that: (1) building characteristics are an essential factor affecting LST. Though the overall significance level was high, the correlation coefficient was within −0.3 and 0.4; (2) the correlation was spatially nonstationary. The R-squared value of GWR was approximately 0.4 higher than that of ordinary least squares (OLS). Although the body shape coefficient and LST were negatively correlated, the relationship between the low-rise building clusters was positive. In addition, the building height of blocks with a low building density was positively correlated with LST; (3) there were differences in the strength of between building types. Residential buildings had the strongest correlation, and commercial buildings had the weakest. This study provides a reference for planners to adjust the layout of block buildings and reduce the UHI effect.

Case studies on seismic assessment of historical buildings using advanced analysis

Seismic risk mitigation in existing buildings requires an engineering assessment of the current condition and expected seismic performance and an identification of possible deficiencies that should be addressed. For heritage and historical buildings in particular, there is significant benefit in using the most detailed analysis methods available to avoid the conservatism inherent in simpler methods and thereby minimise unnecessary interventions and more precisely pinpoint where strengthening is required. On recent heritage projects, Arup has used the analysis software LS-DYNA and a new material model, calibrated against experimental tests on unreinforced masonry components and buildings to carry out (or supplement) seismic assessments. The analysis method (non-linear response history analysis) is not new, but its application on detailed finite-element models of complex historic structures has previously been computationally prohibitive and requires significant analyst experience to deliver reliable results. This paper summarises three of these recent Arup projects: Woltersum Church (the Netherlands), Procuratie Vecchie (Venice) and a building cluster in the historical centre of Appingedam (the Netherlands). The case studies show that these analyses allow complex features of seismic performance to be considered, such as damage or modifications to the building over time, pounding (separate buildings colliding into one another due to seismic movements) and load sharing between adjacent structures.

Hierarchical distributed multi-energy demand response for coordinated operation of building clusters

This paper proposes a distributed multi-energy demand response (DR) methodology for the optimal coordinated operation of smart building clusters based on a hierarchical building-aggregator interaction framework. In the proposed hierarchical framework, the aggregator acts as a digital representation of building entities to offer the multi-energy load prediction of buildings using a capsule network (CapsNet) based multi-energy demand prediction model, while these buildings leverage the load flexibility and multi-energy complementarity to implement the optimal DR for reducing individual costs. Then, a fully distributed multi-energy DR approach based on the exchange alternating direction method of multipliers (ADMM), which requires only limited information to be exchanged between the aggregator and buildings, is developed to iteratively achieve the optimal multi-energy coordination of buildings. Moreover, the proposed model can be dynamically corrected with real-time load data and weather information, and the distributed multi-energy DR approach is correspondingly optimized with rolling horizon procedures to reduce the impact of prediction uncertainties. Finally, the performance of the proposed methodology is benchmarked and validated on different scales of smart buildings, and comparative results demonstrated its superiority in solving the optimal synergistic operation problem of smart buildings.

Green roof on the ventilation and pollutant dispersion in urban street canyons under unstable thermal stratification: Aiding and opposing effects

• Green roof influenced on the outdoor thermal buoyancy flow distributions in a wide range of H/W. • Aiding and opposing effects were observed as wind and thermal buoyancy simultaneously acted. • A “free air flow region” arose when the leeward green roof deployment was adopted. • Lateral deployment of green roof system performed better than that of rooftop deployment. • Rational design of the green roof system could enhance its cooling capacity. The “thermal effect” of the green roof system on the pollutant dilution process would be analyzed in this research. Different arrangement schemes (on the rooftop, windward side and leeward side) of the green roof system would alter the thermal buoyancy distribution in the street canyon, thus change the overall flow field and pollutant dilution condition Computational Fluid Dynamics (CFD) technique along with the Renormalization Group (RNG) k-ε turbulence model has been adopted for the simulation work. Cases without the green roof system are assumed to be uniformly heated and are prepared for benchmark comparisons. Six aspect ratios ( H/W ) and ten ranks of unstable thermal stratifications ( Rb ) are selected to consider the influence of street canyon morphology and thermal buoyancy. Results demonstrate that the lateral deployment of green roof system performs better than the rooftop deployment and the situations without green roof system. In the windward cooling cases, a single vortex is formed in the canyon center and dramatically reduces the pollutant retention time τ when H/W > 1, while in the leeward cooling cases, a “free airflow” region is formed in all the tested aspect ratios expect for H/W = 0.5. The rooftop deployment scheme has negative effects, not just decaying ventilation performance, but also delaying pollutant dispersion period. Present research could benefit the rational design of the green roof system in the urban building clusters, and further enhance its cooling capacity for the built energy conservation.

A critical review of control schemes for demand-side energy management of building clusters

The proliferation of smart meters, energy storage systems (ESS), and renewable energy systems (RES) creates a unique opportunity for the building sector to enhance the efficacy of demand-side energy management. The objective of this review is to capture the trends, strategies, and lessons learned from studies focused on building-to-grid interactions at the cluster level, including residential and mixed-use neighborhoods, microgrids, and multi-unit residential buildings. In approximately 75% of the 82 reviewed studies, ESS such as electric vehicles or batteries were used to reduce the peak load of the building cluster and capture excess renewable energy generation. All the reviewed studies ensured that computers and televisions were non-controllable loads to ensure occupant needs while HVAC, washing machines, and dishwashers were targeted as controllable loads to help reduce and shift energy peaks. A 20% and 30% reduction in total energy cost and peak energy consumption respectively were achieved under the various control schemes, with approximately 50% of the studies using price-based strategies to shift energy consumption away from peak hours. Most importantly, the reviewed studies achieved these reductions while ensuring or maximizing occupant comfort, thus presenting buildings as a great opportunity to help manage energy consumption for the grid.

Testing of Software for the Planning of a Linear Object GNSS Measurement Campaign under Simulated Conditions

The precision of a linear object measurement using satellite techniques is determined by the number and the relative position of the visible satellites by the receiver. The status of the visible constellation is described by the Dilution Of Precision (DOP). The obtained geometric coefficient values are dependent on many variables. When determining these values, field obstacles at the receiver location and satellite positions changing with time must be taken into account. Carrying out a series of surveys as part of a linear object Global Navigation Satellite System (GNSS) measurement campaign requires the optimisation problem to be solved. The manner of the inspection vehicle’s movement should be determined in such a way that the surveys are taken only within the pre-defined time frames and that the geometric coefficient values obtained at subsequent points of the route are as low as possible. The purpose of this article is to develop a software for the planning of a linear object GNSS measurement campaign to implemented in motion and taking into account the terrain model and its coverage. Additionally, it was determined how much the developed program improves DOP values on the planned route under simulated conditions. This software has no equivalent elsewhere in the world, as the current solutions for the planning of a GNSS measurement campaign, e.g., Trimble GNSS Planning, GNSS Mission Planning, or GPS Navigation Toolbox, allow the satellite constellation geometry to be analysed exclusively for specific coordinates and at a specific time. Analysis of the obtained simulation test results indicates that the campaign implementation in accordance with the pre-determined schedule significantly improves the quality of the recorded GNSS data. This is particularly noticeable when determining the position using the Global Positioning System (GPS) and GLObal NAvigation Satellite System (GLONASS) satellite constellations at the same time. During the tests conducted on the road along a three-kilometre-long route (tram loop) in Gdańsk Brzeźno, the average value of the obtained Position Dilution Of Precision (PDOP) decreased by 22.17% thanks to using the software to plan a linear object GNSS measurement campaign. The largest drop in the geometric coefficient values was noted for an area characterised by a very large number of field obstacles (trees with crowns and high buildings). Under these conditions, the PDOP value decreased by approx. 25%. In areas characterised by a small number of field obstacles (single trees in the vicinity of the track, clusters of trees and buildings located along the track), the changes in the PDOP were slightly smaller and amounted to several percent.

Two‐Dimensional Striking Distance of Lightning Flashes to a Cluster of Tall Buildings in Guangzhou

AbstractBased on the high‐speed video records of 54 lightning flashes striking on a cluster of tall buildings in Guangzhou and the return stroke peak currents provided by lightning location systems, this paper analyzed the two‐dimensional (2D) first‐return‐stroke striking distances (SD) of lightning flashes to buildings with a height ranging from 100 to 600 m. Three methods, including one proposed in this study, have been used to estimate the 2D SD. The results with different methods basically agree with each other and show that the height and the top geometry of the structure on which lightning terminates are key factors that affect the SD. The correlation between the SD and the peak return stroke current appeared very scattered, particularly in the cases when each building was considered separately. Besides, we found that: (a) the estimated 2D average initiation speed of upward connecting leader (UCL) on different buildings are similar, ranging from 4.9 to 23 × 104 m/s, with an average of 13.4 × 104 m/s; (b) about 87% (27/31) flashes that struck on buildings with simple top shape only have one upward leader observed, while for buildings with complicated top shape, only 26% (6/23) flash cases have one upward leader observed; (c) for the 36 flashes to four tall buildings with heights higher than 300 m, the 2D average speed ratio of the downward leader and the UCL is 0.74 during the last 0.1 ms before the first return stroke.

Heat Balance Calculation and Energy Efficiency Analysis for Building Clusters Based on Psychrometric Chart

How to perform accurate calculation of heat balance and quantitative analysis of energy efficiency for building clusters is an urgent problem to be solved to reduce building energy consumption and improve energy utilization efficiency. This article proposes a method for the heat balance calculation and energy efficiency analysis of building clusters based on enthalpy and humidity diagrams and applies it to the energy management of building clusters containing primary return air systems and heating pipe networks. Firstly, the basic structure and energy management principle of building clusters with a primary return air system and a heating pipe network were given, and the heat balance calculation and energy efficiency analysis method based on i-d diagram was proposed to realize the accurate calculation of heat load and the quantification of energy utilization. Secondly, the energy management model of the building cluster with a primary return air system and a heating pipe network was established to efficiently manage the indoor temperature and the heating schedule of ASHP, HN and HI. Finally, the proposed method was validated by calculation examples, and the results showed that the proposed method is beneficial for improving the energy economy and energy efficiency of building clusters.

Application of the TEAC software for analysis of Energy Flexible Building Clusters – a case study

Nowadays, natural environment protection and sustainable development became common and necessary issues for all the economic sectors. It is extremely important to focus on all the efforts resulting in the most efficient and sustainable power sources and electric power grid. Typically, the residential districts are connected by electric grids, which with an application of the appropriate technologies might be considered as so-called smart-grids. In the smart-grid neighbourhoods, houses are the consumers, energy supply is performed by the local or/and national power plants, while energy distribution is performed using some monitoring and management systems. Such a residential area can be considered as a Building Cluster, the novel paradigm in the energy and environmental analysis of the built environments. In this article, the exemplary single-family houses neighbourhood is examined, following the Building Cluster paradigm. The analysed area is located in Lodz (Poland), consisting of 202 buildings. The study is performed by means of the home-developed software named TEAC (Tool for Energy Efficiency Analyses of an Energy Cluster). The analysis is focused on the energy, economic and environmental issues of the considered Building Cluster.

Building Typification in Map Generalization Using Affinity Propagation Clustering

Building typification is of theoretical interest and practical significance in map generalization. It aims to transform an initial set of buildings to a subset, while maintaining the essential distribution characteristics and important individual buildings. This study focuses on buildings located in residential suburban or rural areas and generalizes them to medium or small scale, for which the typification process can be viewed as point-similar object selection that generates exemplars in local building clusters. From this view, we propose a novel building typification approach using affinity propagation exemplar-based clustering. Based on a sparse graph constructed on the input building set, the proposed approach considers all buildings as potential cluster exemplars and keeps passing messages between those objects; thus, high-quality representative objects (i.e., exemplars) of the initial building set can be obtained and further outputted as the typified result. Experiments with real-life building data show that the proposed method is superior to the two existing representative methods in maintaining the overall distribution characteristics. Meanwhile, the importance of each individual building and the constraints of the road network can be embedded flexibly in this method, which gives some advantages in terms of preserving important buildings and the local structural distribution along the road, etc.

Demand Side Management Based Power-to-Heat and Power-to-Gas Optimization Strategies for PV and Wind Self-Consumption in a Residential Building Cluster

The volatility of renewable energy sources (RES) poses a growing problem for operation of electricity grids. In contrary, the necessary decarbonisation of sectors such as heat supply and transport requires a rapid expansion of RES. Load management in the context of power-to-heat systems can help to simultaneously couple the electricity and heat sectors and stabilise the electricity grid, thus enabling a higher share of RES. In addition power-to-hydrogen offers the possibility of long-term energy storage options. Within this work, we present a novel optimization approach for heat pump operation with the aim to counteract the volatility and enable a higher usage of RES. For this purpose, a detailed simulation model of buildings and their energy supply systems is created, calibrated and validated based on a plus energy settlement. Subsequently, the potential of optimized operation is determined with regard to PV and small wind turbine self-consumption. In addition, the potential of seasonal hydrogen storage is examined. The results show, that on a daily basis a 33% reduction of electricity demand from grid is possible. However, the average optimization potential is reduced significantly by prediction inaccuracy. The addition of a hydrogen system for seasonal energy storage basically eliminates the carbon dioxide emissions of the cluster. However, this comes at high carbon dioxide prevention costs of 1.76€kg−1.

Building energy consumption prediction for residential buildings using deep learning and other machine learning techniques

The high proportion of energy consumed in buildings has engendered the manifestation of many environmental problems which deploy adverse impacts on the existence of mankind. The prediction of building energy use is essentially proclaimed to be a method for energy conservation and improved decision-making towards decreasing energy usage. Also, the construction of energy efficient buildings will aid the reduction of total energy consumed in newly constructed buildings. Machine Learning (ML) method is recognised as the best suited approach for producing desired outcomes in prediction task. Hence, in several studies, ML has been applied in the field of energy consumption of operational building. However, there are not many studies investigating the suitability of ML methods for forecasting the potential building energy consumption at the early design phase to reduce the construction of more energy inefficient buildings. To address this gap, this paper presents the utilization of several machine learning techniques namely Artificial Neural Network (ANN), Gradient Boosting (GB), Deep Neural Network (DNN), Random Forest (RF), Stacking, K Nearest Neighbour (KNN), Support Vector Machine (SVM), Decision tree (DT) and Linear Regression (LR) for predicting annual building energy consumption using a large dataset of residential buildings. This study also examines the effect of the building clusters on the model performance. The novelty of this paper is to develop a model that enables designers input key features of a building design and forecast the annual average energy consumption at the early stages of development. This result reveals DNN as the most efficient predictive model for energy use at the early design phase and this presents a motivation for building designers to utilize it before construction to make informed decision, manage and optimize design.

Microgrid for a Cluster of Grid Independent Buildings Powered by Solar and Wind Energy

The reduction of CO2 emissions and the avoidance of Global Climate Change necessitate the conversion of the electricity generation industry to rely on non-carbon sources. Additionally, the mitigation of the duck-curve effects in microgrids requires the development of grid-independent buildings. Computations were performed for a cluster of one thousand grid-independent buildings in the North Texas area, where air-conditioning demand is high in the summer months. The electricity demand is balanced with energy supply generated from wind turbines, photovoltaic cells, or stored energy in hydrogen tanks. The results indicate that with one wind turbine operating, each building must be fitted with 10.2 kW rating photovoltaics capacity and a tank with 5.2 m3 of hydrogen storage capacity to satisfy the hourly demand of the buildings’ community. The addition of more wind turbines significantly reduces the needed PV rating but increases the required storage. Investing in energy conservation measures in the buildings significantly reduces both the needed storage capacity and the PV cell ratings.

Distributed urban drag parametrization for sub‐kilometre scale numerical weather prediction

AbstractA recently developed, height‐distributed urban drag parametrization is tested with the London Model, a sub‐kilometre resolution version of the Met Office Unified Model over Greater London. The distributed‐drag parametrization requires vertical morphology profiles in the form of height‐distributed frontal‐area functions, which capture the full extent and variability of building heights. London's morphology profiles are calculated and parametrized by an exponential distribution with the ratio of maximum to mean building height as the parameter. A case study evaluates the differences between the new distributed‐drag scheme and the current London Model setup using the MORUSES urban land‐surface model. The new drag parametrization shows increased horizontal spatial variability in total surface stress, identifying densely built‐up areas, high‐rise building clusters, parks, and the river. Effects on the wind speed in the lower levels include a lesser gradient and more heterogeneous wind profiles, extended wakes downwind of the city centre, and vertically growing perturbations that suggest the formation of internal boundary layers. The surface sensible heat fluxes are underpredicted, which is attributed to difficulties coupling the distributed momentum exchange with the surface‐based heat exchange.

Temperature and flow rates dynamics in a district heating (DH) network fed by a combined heat-and-power (CHP) plant

Research focus (problem description, short background):The energy efficiency of a district heating (DH) system can be improved by fine-tuning supply and return temperatures and flow rates without need for invasive renovation of the end user heating systems. Research methods:We employ a correlation coefficient to examine the relationship between the local supply temperature and the use of heat; research relies on data hourly heat generation records from each heat generation facility in the DH system of Omsk, Russia, 2016–2017. Key research results/findings:The correlation declines the further we go from winter. Firstly, the relationship between the transition states and the outside temperature results from the relationship between external conditionings and heat parameters in DH. Secondly, both the energy consumption and the supply temperature are affected by the infrastructure, other weather conditions, and energy price; they are also highly dependent on consumer behavior, which is arbitrary, heterogeneous, and interdependent. Thirdly, the regional heating load is influenced by the characteristics of a building cluster. Although nearly all DH plants have a control system in place, we find that human factor jeopardizes the performance of such systems. Main conclusions and recommendations:To generalize these results to other systems, we emphasize that the effective diffusion coefficients and effective heat-loss coefficients are influenced by the thermal dynamics in the network and change on typical time scale (one year), hence their uncertainty. This study has produced a technique for visualizing the relationship that exists in a DH system between supply temperature and heat consumption; this technique is expected to contribute to the dynamic modeling, control, and planning of future integrated energy systems. The load profiles are dynamic; a thermal storage could be used to shift the timing of heating loads to minimize the flow through DH networks.

Cluster analysis of agricultural products storage infrastructure in Kazakhstan: directions and development prospects

Purpose - the article is devoted to identifying the needs of the economy of Kazakhstan in infrastructure facilities for storing agricultural products and the prospects for its further development. Methods - statistical, factorial and k-mean cluster analysis. Results - the influence of an efficiently functioning network of storage facilities for agricultural products on the stability of providing the population with food, reducing the volatility of prices and supplies in the off-season is justified. The authors revealed that modern premises for warehousing and preserving the quality of agricultural products in foreign countries combine the functions of storage and marketing by creating logistics wholesale distribution centers (WDC), which allow the use of modern information technologies and computer systems aimed to form consumer applications for formulating commodity groups and delivery of the required product volumes to the consumer with minimal costs. Based on the analysis of indicators of harvesting and use of the main types of agricultural products in the republic (stocks at the beginning of the year, gross harvest, imports, industrial use and domestic consumption, losses, exports, personal consumption by the population, reserves at the end of the year, availability of storage facilities, shortage), the need in the storage infrastructure has been determined. The clustering of warehouse buildings was carried out according to the main features: the volume of one-time preservation, level of losses, availability of climatic warehouses and transport mechanism, on the basis of which five clusters have been identified. Conclusions - cluster analysis of storages showed their uneven and irrigation distribution across the regions of the country. It is necessary to modernize and build new such facilities of various capacities on the basis of the WDC - large hubs concentrated in various regions of Kazakhstan, depending on their needs.

Data-driven district energy management with surrogate models and deep reinforcement learning

Demand side management at district scale plays a crucial role in the energy transition process, being an ideal candidate to balance the needs of both users and grid, by managing the volatility of renewable sources and increasing energy flexibility. The presented study aims to explore the benefits of a coordinated approach for the energy management of a cluster of buildings to optimise the electrical demand profiles and provide services to the grid without penalising indoor comfort conditions. The proposed methodology makes use of a fully data-driven control scheme which exploits Long Short-Term Memory (LSTM) Neural Networks, and Deep Reinforcement Learning (DRL). A simulation environment is introduced to train a DRL controller to manage the operation of heat pumps and chilled and domestic hot water storage for a cluster of four buildings. LSTM models are trained with synthetic data set created in EnergyPlus and are integrated into simulation environment to evaluate the indoor temperature dynamics in each building. The developed DRL controller is tested against a manually optimised Rule Based Controller (RBC). Results show that the DRL algorithm is able to reduce the overall cluster electricity costs, while decreasing the peak energy demand by 23% and the Peak to Average Ratio (PAR) by 20%, without penalizing indoor temperature control.