Completeness of data is vital for the decision making and forecasting on Building Management Systems (BMS) as missing data can result in biased decision making down the line. This study creates a guideline for imputing the gaps in BMS datasets by comparing four methods: K Nearest Neighbour algorithm (KNN), Recurrent Neural Network (RNN), Hot Deck (HD) and Last Observation Carried Forward (LOCF). The guideline contains the best method per gap size and scales of measurement. The four selected methods are from various backgrounds and are tested on a real BMS and metereological dataset. The focus of this paper is not to impute every cell as accurately as possible but to impute trends back into the missing data. The performance is characterised by a set of criteria in order to allow the user to choose the imputation method best suited for its needs. The criteria are: Variance Error (VE) and Root Mean Squared Error (RMSE). VE has been given more weight as its ability to evaluate the imputed trend is better than RMSE. From preliminary results, it was concluded that the best K‐values for KNN are 5 for the smallest gap and 100 for the larger gaps. Using a genetic algorithm the best RNN architecture for the purpose of this paper was determined to be GatedRecurrent Units (GRU). The comparison was performed using a different training dataset than the imputation dataset. The results show no consistent link between the difference in Kurtosis or Skewness and imputation performance. The results of the experiment concluded that RNN is best for interval data and HD is best for both nominal and ratio data. There was no single method that was best for all gap sizes as it was dependent on the data to be imputed.
MULTIFILE
Completeness of data is vital for the decision making and forecasting on Building Management Systems (BMS) as missing data can result in biased decision making down the line. This study creates a guideline for imputing the gaps in BMS datasets by comparing four methods: K Nearest Neighbour algorithm (KNN), Recurrent Neural Network (RNN), Hot Deck (HD) and Last Observation Carried Forward (LOCF). The guideline contains the best method per gap size and scales of measurement. The four selected methods are from various backgrounds and are tested on a real BMS and meteorological dataset. The focus of this paper is not to impute every cell as accurately as possible but to impute trends back into the missing data. The performance is characterised by a set of criteria in order to allow the user to choose the imputation method best suited for its needs. The criteria are: Variance Error (VE) and Root Mean Squared Error (RMSE). VE has been given more weight as its ability to evaluate the imputed trend is better than RMSE. From preliminary results, it was concluded that the best K‐values for KNN are 5 for the smallest gap and 100 for the larger gaps. Using a genetic algorithm the best RNN architecture for the purpose of this paper was determined to be Gated Recurrent Units (GRU). The comparison was performed using a different training dataset than the imputation dataset. The results show no consistent link between the difference in Kurtosis or Skewness and imputation performance. The results of the experiment concluded that RNN is best for interval data and HD is best for both nominal and ratio data. There was no single method that was best for all gap sizes as it was dependent on the data to be imputed.
DOCUMENT
In this article a generic fault detection and diagnosis (FDD) method for demand controlled ventilation (DCV) systems is presented. By automated fault detection both indoor air quality (IAQ) and energy performance are strongly increased. This method is derived from a reference architecture based on a network with 3 generic types of faults (component, control and model faults) and 4 generic types of symptoms (balance, energy performance, operational state and additional symptoms). This 4S3F architecture, originally set up for energy performance diagnosis of thermal energy plants is applied on the control of IAQ by variable air volume (VAV) systems. The proposed method, using diagnosis Bayesian networks (DBNs), overcomes problems encountered in current FDD methods for VAV systems, problems which inhibits in practice their wide application. Unambiguous fault diagnosis stays difficult, most methods are very system specific, and finally, methods are implemented at a very late stage, while an implementation during the design of the HVAC system and its control is needed. The IAQ 4S3F method, which solves these problems, is demonstrated for a common VAV system with demand controlled ventilation in an office with the use of a whole year hourly historic Building Management System (BMS) data and showed it applicability successfully. Next to this, the influence of prior and conditional probabilities on the diagnosis is studied. Link to the formal publication via its DOI https://doi.org/10.1016/j.buildenv.2019.106632
DOCUMENT
B4B is a multi-year, multi-stakeholder project focused on developing methods to harness big data from smart meters, building management systems and the Internet of Things devices, to reduce energy consumption, increase comfort, respond flexibly to user behaviour and local energy supply and demand, and save on installation maintenance costs. This will be done through the development of faster and more efficient Machine Learning and Artificial Intelligence models and algorithms. The project is geared to existing utility buildings such as commercial and institutional buildings.
DOCUMENT
The built environment requires energy-flexible buildings to reduce energy peak loads and to maximize the use of (decentralized) renewable energy sources. The challenge is to arrive at smart control strategies that respond to the increasing variations in both the energy demand as well as the variable energy supply. This enables grid integration in existing energy networks with limited capacity and maximises use of decentralized sustainable generation. Buildings can play a key role in the optimization of the grid capacity by applying demand-side management control. To adjust the grid energy demand profile of a building without compromising the user requirements, the building should acquire some energy flexibility capacity. The main ambition of the Brains for Buildings Work Package 2 is to develop smart control strategies that use the operational flexibility of non-residential buildings to minimize energy costs, reduce emissions and avoid spikes in power network load, without compromising comfort levels. To realise this ambition the following key components will be developed within the B4B WP2: (A) Development of open-source HVAC and electric services models, (B) development of energy demand prediction models and (C) development of flexibility management control models. This report describes the developed first two key components, (A) and (B). This report presents different prediction models covering various building components. The models are from three different types: white box models, grey-box models, and black-box models. Each model developed is presented in a different chapter. The chapters start with the goal of the prediction model, followed by the description of the model and the results obtained when applied to a case study. The models developed are two approaches based on white box models (1) White box models based on Modelica libraries for energy prediction of a building and its components and (2) Hybrid predictive digital twin based on white box building models to predict the dynamic energy response of the building and its components. (3) Using CO₂ monitoring data to derive either ventilation flow rate or occupancy. (4) Prediction of the heating demand of a building. (5) Feedforward neural network model to predict the building energy usage and its uncertainty. (6) Prediction of PV solar production. The first model aims to predict the energy use and energy production pattern of different building configurations with open-source software, OpenModelica, and open-source libraries, IBPSA libraries. The white-box model simulation results are used to produce design and control advice for increasing the building energy flexibility. The use of the libraries for making a model has first been tested in a simple residential unit, and now is being tested in a non-residential unit, the Haagse Hogeschool building. The lessons learned show that it is possible to model a building by making use of a combination of libraries, however the development of the model is very time consuming. The test also highlighted the need for defining standard scenarios to test the energy flexibility and the need for a practical visualization if the simulation results are to be used to give advice about potential increase of the energy flexibility. The goal of the hybrid model, which is based on a white based model for the building and systems and a data driven model for user behaviour, is to predict the energy demand and energy supply of a building. The model's application focuses on the use case of the TNO building at Stieltjesweg in Delft during a summer period, with a specific emphasis on cooling demand. Preliminary analysis shows that the monitoring results of the building behaviour is in line with the simulation results. Currently, development is in progress to improve the model predictions by including the solar shading from surrounding buildings, models of automatic shading devices, and model calibration including the energy use of the chiller. The goal of the third model is to derive recent and current ventilation flow rate over time based on monitoring data on CO₂ concentration and occupancy, as well as deriving recent and current occupancy over time, based on monitoring data on CO₂ concentration and ventilation flow rate. The grey-box model used is based on the GEKKO python tool. The model was tested with the data of 6 Windesheim University of Applied Sciences office rooms. The model had low precision deriving the ventilation flow rate, especially at low CO2 concentration rates. The model had a good precision deriving occupancy from CO₂ concentration and ventilation flow rate. Further research is needed to determine if these findings apply in different situations, such as meeting spaces and classrooms. The goal of the fourth chapter is to compare the working of a simplified white box model and black-box model to predict the heating energy use of a building. The aim is to integrate these prediction models in the energy management system of SME buildings. The two models have been tested with data from a residential unit since at the time of the analysis the data of a SME building was not available. The prediction models developed have a low accuracy and in their current form cannot be integrated in an energy management system. In general, black-box model prediction obtained a higher accuracy than the white box model. The goal of the fifth model is to predict the energy use in a building using a black-box model and measure the uncertainty in the prediction. The black-box model is based on a feed-forward neural network. The model has been tested with the data of two buildings: educational and commercial buildings. The strength of the model is in the ensemble prediction and the realization that uncertainty is intrinsically present in the data as an absolute deviation. Using a rolling window technique, the model can predict energy use and uncertainty, incorporating possible building-use changes. The testing in two different cases demonstrates the applicability of the model for different types of buildings. The goal of the sixth and last model developed is to predict the energy production of PV panels in a building with the use of a black-box model. The choice for developing the model of the PV panels is based on the analysis of the main contributors of the peak energy demand and peak energy delivery in the case of the DWA office building. On a fault free test set, the model meets the requirements for a calibrated model according to the FEMP and ASHRAE criteria for the error metrics. According to the IPMVP criteria the model should be improved further. The results of the performance metrics agree in range with values as found in literature. For accurate peak prediction a year of training data is recommended in the given approach without lagged variables. This report presents the results and lessons learned from implementing white-box, grey-box and black-box models to predict energy use and energy production of buildings or of variables directly related to them. Each of the models has its advantages and disadvantages. Further research in this line is needed to develop the potential of this approach.
DOCUMENT
The Heating Ventilation and Air Conditioning (HVAC) sector is responsible for a large part of the total worldwide energy consumption, a significant part of which is caused by incorrect operation of controls and maintenance. HVAC systems are becoming increasingly complex, especially due to multi-commodity energy sources, and as a result, the chance of failures in systems and controls will increase. Therefore, systems that diagnose energy performance are of paramount importance. However, despite much research on Fault Detection and Diagnosis (FDD) methods for HVAC systems, they are rarely applied. One major reason is that proposed methods are different from the approaches taken by HVAC designers who employ process and instrumentation diagrams (P&IDs). This led to the following main research question: Which FDD architecture is suitable for HVAC systems in general to support the set up and implementation of FDD methods, including energy performance diagnosis? First, an energy performance FDD architecture based on information embedded in P&IDs was elaborated. The new FDD method, called the 4S3F method, combines systems theory with data analysis. In the 4S3F method, the detection and diagnosis phases are separated. The symptoms and faults are classified into 4 types of symptoms (deviations from balance equations, operating states (OS) and energy performance (EP), and additional information) and 3 types of faults (component, control and model faults). Second, the 4S3F method has been tested in four case studies. In the first case study, the symptom detection part was tested using historical Building Management System (BMS) data for a whole year: the combined heat and power plant of the THUAS (The Hague University of Applied Sciences) building in Delft, including an aquifer thermal energy storage (ATES) system, a heat pump, a gas boiler and hot and cold water hydronic systems. This case study showed that balance, EP and OS symptoms can be extracted from the P&ID and the presence of symptoms detected. In the second case study, a proof of principle of the fault diagnosis part of the 4S3F method was successfully performed on the same HVAC system extracting possible component and control faults from the P&ID. A Bayesian Network diagnostic, which mimics the way of diagnosis by HVAC engineers, was applied to identify the probability of all possible faults by interpreting the symptoms. The diagnostic Bayesian network (DBN) was set up in accordance with the P&ID, i.e., with the same structure. Energy savings from fault corrections were estimated to be up to 25% of the primary energy consumption, while the HVAC system was initially considered to have an excellent performance. In the third case study, a demand-driven ventilation system (DCV) was analysed. The analysis showed that the 4S3F method works also to identify faults on an air ventilation system.
DOCUMENT
In practice, faults in building installations are seldom noticed because automated systems to diagnose such faults are not common use, despite many proposed methods: they are cumbersome to apply and not matching the way of thinking of HVAC engineers. Additionally, fault diagnosis and energy performance diagnosis are seldom combined, while energy wastage is mostly a consequence of component, sensors or control faults. In this paper new advances on the 4S3F diagnose framework for automated diagnostic of energy waste in HVAC systems are presented. The architecture of HVAC systems can be derived from a process and instrumentation diagram (P&ID) usually set up by HVAC designers. The paper demonstrates how all possible faults and symptoms can be extracted on a very structured way from the P&ID, and classified in 4 types of symptoms (deviations from balance equations, operational states, energy performances or additional information) and 3 types of faults (component, control and model faults). Symptoms and faults are related to each other through Diagnostic Bayesian Networks (DBNs) which work as an expert system. During operation of the HVAC system the data from the BMS is converted to symptoms, which are fed to the DBN. The DBN analyses the symptoms and determines the probability of faults. Generic indicators are proposed for the 4 types of symptoms. Standard DBN models for common components, controls and models are developed and it is demonstrated how to combine them in order to represent the complete HVAC system. Both the symptom and the fault identification parts are tested on historical BMS data of an ATES system including heat pump, boiler, solar panels, and hydronic systems. The energy savings resulting from fault corrections are estimated and amount 25%. Finally, the 4S3F method is extended to hard and soft sensor faults. Sensors are the core of any FDD system and any control system. Automated diagnostic of sensor faults is therefore essential. By considering hard sensors as components and soft sensors as models, they can be integrated into the 4S3F method.
DOCUMENT
Current symptom detection methods for energy diagnosis in heating, ventilation and air conditioning (HVAC) systems are not standardised and not consistent with HVAC process and instrumentation diagrams (P&IDs) as used by engineers to design and operate these systems, leading to a very limited application of energy performance diagnosis systems in practice. This paper proposes detection methods to overcome these issues, based on the 4S3F (four types of symptom and three types of faults) framework. A set of generic symptoms divided into three categories (balance, energy performance and operational state symptoms) is discussed and related performance indicators are developed, using efficiencies, seasonal performance factors, capacities, and control and design-based operational indicators. The symptom detection method was applied successfully to the HVAC system of the building of The Hague University of Applied Sciences. Detection results on an annual, monthly and daily basis are discussed and compared. Link to the formail publication via its DOI https://doi.org/10.1016/j.autcon.2020.103344
DOCUMENT
Current methods for energy diagnosis in heating, ventilation and air conditioning (HVAC) systems are not consistent with process and instrumentation diagrams (P&IDs) as used by engineers to design and operate these systems, leading to very limited application of energy performance diagnosis in practice. In a previous paper, a generic reference architecture – hereafter referred to as the 4S3F (four symptoms and three faults) framework – was developed. Because it is closely related to the way HVAC experts diagnose problems in HVAC installations, 4S3F largely overcomes the problem of limited application. The present article addresses the fault diagnosis process using automated fault identification (AFI) based on symptoms detected with a diagnostic Bayesian network (DBN). It demonstrates that possible faults can be extracted from P&IDs at different levels and that P&IDs form the basis for setting up effective DBNs. The process was applied to real sensor data for a whole year. In a case study for a thermal energy plant, control faults were successfully isolated using balance, energy performance and operational state symptoms. Correction of the isolated faults led to annual primary energy savings of 25%. An analysis showed that the values of set probabilities in the DBN model are not outcome-sensitive. Link to the formal publication via its DOI https://doi.org/10.1016/j.enbuild.2020.110289
DOCUMENT
To understand how transition across different thermal zones in a building impacts the thermal perception of occupants, the current work examines occupant feedback in two work environments — nursing staff in hospital wards and the workers in an office. Both studies used a mix of subjective surveys and objective measurements. A total of 96 responses were collected from the hospital wards while 142 were collected from the office. The thermal environment in the hospital wards was perceived as slightly warm on the ASHRAE thermal sensation scale (mean TSV = 1.2), while the office workers rated their environment on the cool side (mean TSV = 0.15). The results also show that when the transitions were across temperature differences within 2 °C, the thermal perception was not impacted by the magnitude of the temperature difference — as reflected in occupant thermal sensation and thermal comfort/thermal acceptability vote. This would imply that the effect of temperature steps on thermal perception, if any, within these boundaries, was extremely short lived. These findings go towards establishing the feasibility of heterogeneous indoor thermal environments and thermal zoning of workspaces for human comfort.
DOCUMENT
