A novel method of determining events in combination gas boilers: Assessing the feasibility of a passive acoustic sensor
Introduction
Approximately 72% of energy consumption in the domestic environment is done so in gas-boilers for the purposes of domestic central heating (DCH) and domestic hot water (DHW) production [4]. Being such a large contributor to domestic energy consumption, gas-boilers have received significant attention from policy makers and researchers alike. In accordance to the UK Government's CO2 target of an 80% reduction in emissions by 2050 [28], amongst other actions, policy makers have modified building regulations [11] so that boilers installed after 2006 are required to have a SEDBUK [2] efficiency of 88% or above. In addition to policy changes, many researchers look to evaluate environmental or behavioural interventions designed to reduce DCH and/or DHW consumption. To evaluate such interventions it is necessary to accurately monitor changes in disaggregated operational energy consumption, often in field trials.
Measuring energy consumed for DCH in field trials has however proven a challenge, as Hong et al. [12] note: due to the “… high cost associated with sophisticated fuel-use monitoring equipment required for different types of heating system and the complexity of its installation”. Estimates of DCH energy use have been made in a number of studies by subtracting the summer fuel load from the winter fuel load [12], [36]. These methods however require more than one years' worth of data and assume no variation occurs for non-space heating energy use between seasons. Other studies, such as Martin and Watson [22] and Love [23], used temperature sensors on flow pipes or radiators and relate temperature increases to boiler firing. These methods however miss short burner cycles and burner modulation. If methods were available that could accurately detect burner duration and modulation, space heating energy consumption estimates would be more accurate.
In addition to the above, it is well established that unnecessary cycling of boilers is undesirable [24]. Frequent on/off cycles, for example, cause boilers to operate at efficiencies “well below (their) full-load bench value” [38]. This issue has been observed in practice, for example Ren et al. [32] note “Space heating systems cycled more frequently than anticipated due to a tight range of room thermostat settings and potentially oversized heating capacities.” Likewise a survey of 35 non-residential buildings revealed that “all of the installed heating capacity was oversized by a minimum of 30% …” [21]. Furthermore unnecessary cycling causes additional wear and tear on components. These problems can be due to oversizing [10], incorrect return flow temperature sensor settings, inadequate boiler modulation ranges etc. Data on cycling rates and boiler modulation could thus help identify potential issues causing efficiency losses.
Thus to accurately evaluate interventions designed to reduce demand, new field-trial applicable methods are required that can accurately measure operational burner duration for space heating purposes. Additionally such methods could be used to provide data on operational boiler cycling behaviour and help identify potential efficiency losses due to setup, design and/or sizing issues.
The aim of this study is to assess the feasibility of a single-point acoustic sensor and associated detection device that can be retrofitted to Combination Gas Fired Boilers (C-GFB) to provide data on energy consumption, demand type, boiler operation and boiler failure modes. Primary events for detection include demand type and the burner duration. Secondary events include the activity of the pre-mix fan and circulation-pump. Note, the circulation-pump circulates heating water around the closed heating system circuit (boiler, radiators etc.) and the pre-mix fan draws both natural gas and air into the burner chamber and helps to expel the exhaust gases. C-GFBs were selected for this study because, as of 2011, 60% of DCH and DHW systems in the UK were of this type. Additionally the number of C-GFBs has risen steadily since 1990, with no indication of it decreasing in popularity [4].
For this investigation, event detection will use the acoustic signal produced by the physical process of the event itself. This is to avoid making the assumption that the expected process flow of the system is being followed. For example, changes in the pre-mix fan motor speed are directly related to the period of firing, however using the pre-mix fan motor speed as a basis for detecting firing would assume the boiler is behaving as expected. Thus, the investigation bases burner identification on the acoustic signal produced by the burner and not the acoustic signal produced by any other component of the system. By doing so, events occurring outside of the expected process flow of the boiler can potentially be detected.
Alternative methods of event detection exist. One method is to monitor the boiler's Central Processing Unit (CPU). Junkers (the German brand of Bosch), for example, have developed a smartphone application which is linked via a wireless network with the CPU of some Junkers boilers [18]; however the application doesn't give detailed information on boiler events such as ignition time (only failure codes are reported back to the user). In general, accessing a boiler's CPU requires manufacturers' consent and substantial proprietary software, and if performed by research fieldwork teams, could invalidate boiler warranties. The advantage of using a non-invasive retrofitted method, such as an acoustic sensor, is that warranties are not invalidated and theoretically all boilers are accessible irrespective of age, software or hardware. Another option would be to use multiple sensors: The sensors however would need to be situated in various locations depending on the boiler. This increases both cost, complexity and the probability of sensor failure. The advantage of a single acoustic sensor is that most events of interest could, in principle, still be detected, and that the exact positioning of such an acoustic sensor would be of less importance and could sit externally to the boiler.
Section snippets
Related work
This study is focused on identifying techniques to determine events of interest from the acoustic signals produced by domestic C-GFBs. No studies can be found in the literature of this nature, however related studies exist that apply signal analysis techniques to determine resource usage in the domestic environment and they include: the determination of electrical component usage from the electrical mains signals [9], [29]; the localisation of water valve usage from water pipe pressure
Methods
Analysis was made of the acoustic signals emanating from C-GFBs. Relevant techniques were applied to attempt to automatically detect demand and ignition events. The acoustic signatures corresponding to the operation of the pre-mix fan and circulation-pump were investigated.
Demand type
The C-GFB under investigation (C-GFB A) meets the household demands for DHW and DCH. In response to a particular demand type, the initial acoustic signal (first 5 s) produced by the C-GFB consistently followed a repeatable pattern. From the data collected (40 instances in total) DHW demand produced a single acoustic configuration whereas DCH demand produced two acoustic configurations. DCH demand was thus split into two classes; DCH1 and DCH2. Fig. 3 shows the acoustic patterns for the three
Conclusions
Our investigation has shown that acoustic sensing methods, using a single-point sensor, can be used to detect specific events of interest in domestic C-GFBs. The methods developed allow for the accurate automatic detection of demand type (100 ± 0.0% accuracy) and ignition time (97.1% accuracy). Algorithms to determine demand type were trained on features based on the normalised signal energy for the first 4 s of C-GFB activity. Infrasound pressure pulsations produced during ignition were
Acknowledgements
Stephen Hailes and Sarah Chisholm of the Computer Science department at UCL for providing assistance in the application and gathering of equipment for detecting the burner firing event, and for initial discussions regarding application of machine learning algorithms to the analysis of acoustic signals. George Bennett of Bosch and UCL for reviewing the work completed and advising on combination boiler operation. Lastly I acknowledge the London-Loughborough Centre for Doctoral Research in Energy
References (42)
- et al.
The impact of energy efficient refurbishment on the space heating fuel consumption in English dwellings
Energy Build.
(2006) - et al.
Deducing the classification rules for thermal comfort controls using optimal method
Build. Environ.
(2016) - et al.
Modelling Personal Thermal Sensations Using C-Support Vector Classification (C-SVC) algorithm
Build. Environ.
(2016) - et al.
On the control of heating systems in the UK
Build. Environ.
(2005) - et al.
Data mining of space heating system performance in affordable housing
Build. Environ.
(2015) - et al.
Energy consumption for space heating of West-German households: empirical evidence, scenario projections and policy implications
Energy Policy
(2000) - et al.
Improving the effectiveness of a domestic central-heating boiler by the use of heat storage
Appl. energy
(1987) Recording Software
(2016)Building Energy Performance Assessment: DECC
(2015)- et al.
GasSense: Appliance-Level, Single-Point Sensing of Gas Activity in the Home
United Kingdom housing Energy Fact File: 2013
On the optimality of the simple Bayesian classifier under zero-one loss
Mach. Learn.
Sensing from the basement: a feasibility study of unobtrusive and low-cost home activity recognition
HydroSense: infrastructure-mediated single-point sensing of whole-home water activity
ElectriSense: single-point sensing using EMI for electrical event detection and classification in the home
Cycling efficiency: A basis for replacing outsized boilers
Energy Eng.
Approved Document L1A: Conservation of Fuel and Power in New Dwellings
Speech recognition using MFCC
Estimating continuous distributions in Bayesian classifiers
Flame failure detection and modern boilers
J. Phys. E Sci. Instrum.
Cited by (1)
Study on the in situ performance of a 60 kW condensing gas boilers thermal installation and economic assessment
2020, IOP Conference Series: Materials Science and Engineering