Investigating flexible feeding effects on the biogas quality in full‐scale anaerobic digestion by high resolution, photoacoustic‐based NDIR sensing

Abstract

In future energy systems based on renewable energies, biogas plants can make a significant contribution to stabilizing the electricity grids. However, this requires load‐flexible and demand‐oriented electricity production by means of flexible feed management. However, these flexible feeding strategies using greatly oscillating, temporally varying high mass loads may lead to critical process failures of the anaerobic digestion process. Currently there is no online, high resolution gas quality measurement technique to detect and prevent biological process failures available. In this contribution, we present a miniaturized, low‐cost biogas quality measurement system providing data with high precision and high temporal resolution to overcome this technology gap. To highlight the capabilities of the system we have installed it using a bypass to the main biogas duct after hydrogen sulfide removal at a full‐scale research biogas plant. During a three‐month field trial, the effect of flexible feeding on the biogas quality has been monitored. The results demonstrate long‐term stability of the sensor solution and reveal the effects of changing feeding frequency and composition on gas quantity and quality, which cannot be detected with commercially available state‐of‐the‐art sensing systems.

Abbreviations

CH₄

methane

CHP

combined heat and power plant

CO₂

carbon dioxide

H₂S

hydrogen sulfide

LOD

limit of detection

NDIR

non‐dispersive infrared spectroscopy

oDM

organic dry matter

ppm

parts per million

TDLAS

tunable diode laser absorption spectroscopy

URAS

Ultrarot Absorptions Schreiber

1. INTRODUCTION

To meet long‐term climate change targets, reduction of carbon dioxide (CO₂) emissions in the power sector is needed. Intermittent renewable energy sources such as wind and solar photovoltaic can be a key component of the resulting low‐carbon power systems, but their intermittency will require more flexibility from the rest of the power system 1. For this future energy supply based on renewable energy sources, biogas plants may take on the role of a flexible and weather independent heat and electricity source to maintain system stability 2. Discontinuous, demand‐oriented feeding of biogas plants with varying substrates enables a targeted influence on the gas volume flow. At the same time, low feed frequencies in combination with high loading rates of the fermenters lead to variations in the gas quality, making the use of biogas more difficult 3. However, in future energy systems novel forms of electricity supply, such as high temperature fuel cells, may be powered by biogas and provide an alternative form of sustainable and clean energy 4, 5, 6. Also, currently used combined heat and power plants (CHP‐units) rely on a specific biogas quality and the massive feed intake has a strong influence on process stability. This makes the control of quantity and quality of biogas even more important. Research from Mauky et al. 7 showed that during flexible biogas production the gas composition varies greatly within a very short period of time of minutes. These short‐term differences in biogas composition set the lower limit for the measurement frequency of the gas quality that current state‐of‐the‐art systems oftentimes cannot provide. Furthermore, a flexible, demand‐oriented operation of the biogas digesters increases the risk of biological process disturbances 8. Process instability is particularly dangerous especially at high feeding and high organic loading rates. Process control at biogas stations is presently conducted by measuring the concentration of organic acids in the digestate, which are intermediate products during biogas formation 9. As those analyses need to be processed in external laboratories, this method cannot be used to detect process instability instantly and on‐site. Although gas composition is not a very fast indicator of process instability, it may become suitable if the resolution is sufficiently high. If gas quality measurements become available at high resolution, this allows to monitor the fragile biological process and may give information about the status of the its stability since an increased formation of organic acids is coincided with an augmented release of CO₂. Previous results show that commercially available biogas measuring devices have not yet been able to provide such high measurement resolution. However, the majority of currently operating biogas plants are not equipped with any analytic tools for analyzing gas composition. This is mainly due to the initial high unit costs, the cumbersome installation as well as the very high maintenance expenditure for currently available gas sensing equipment. While the process biology of biogas plants is complex, the produced gas mostly contains methane (CH₄), carbon dioxide (CO₂), water (H₂O), hydrogen sulfide (H₂S), ammonia (NH₃) as well as traces of oxygen (O₂) and hydrogen (H₂). For most gas utilization applications, H₂S has to be removed nearly completely because of its highly corrosive effect on machinery. At the same time, the water content of the biogas as well as its relative humidity must be considerably reduced 9. This leaves the CH₄ and CO₂ content as the most important determinants for quality, in particular with regard to the calorific value of the biogas which is important for any subsequent utilization technology. In addition, it is possible to deduce information on process stability from the relative change in the proportion of CH₄ to CO₂. Changes in the acidity of the fermenter influence the CO_2/CH₄ relation. Accordingly, commercially available gas analysis systems for biogas stations or other anaerobic microbiological processes usually feature at least sensor modules to determine CO₂ and CH₄ contents. However, overall biogas quality and bacterial health are governed by further parameters, which oftentimes make a more comprehensive monitoring necessary. The most widely used methods for fast and reliable gas analysis are based on optical methods. To this end, so‐called non‐dispersive infrared spectroscopy (NDIR) is the most used commercially available solution. Alternative market‐ready technologies include thermal conductivity, and Fourier transform spectroscopy. Table 1 lists some commercially available products along with their respective performance figures, overall size and measurement frequency.

Table 1.

An exemplary list of commercially available sensing solutions to determine the methane and carbon dioxide content

Producer	Gas	Measurement range & resolution	Technology	Measurement frequency	Size
ADOS Biogas 401a	CO2 CH4	0–50 Vol.% +/− 3% 0–100 Vol.% +/− 3%	NDIR	Discontinuous	(600 × 478 × 480) mm³
Awite AwiFLEXCool +b	CO2 CH4	0–100 Vol.% +/− 0.2% 0–100 Vol.% +/− 0.2%	NDIR	50/d	(564 × 700 × 268) mm³
Pronova SSM 6000c	CO2 CH4	0–100 Vol.% +/− 0.2% 0–100 Vol.% +/− 0.2%	NDIR	6/h	(300 × 400 × 200) mm³
SGX Sensortech VQ31 Seriesd	CO2 CH4	CO2 in CH4 0–100% (no error stated)	Thermal conductivity	Continuous/not stated	n/a
MKS AIRGARD Biogas Monitore	CO2 CH4	0–100 Vol.% (not stated) 0–100 Vol.% (not stated)	FTIR		(467 × 645 × 191) mm³

^aADOS GmbH, “Mehrkanal‐Gasanalysator, Biogas 401,” 2016.

^bAwite Bioenergie GmbH, “Technische Daten AwiFLEX COOL+,” vol. 10, no. 05.00.

^cPronova, “Biogasanalysatoren, SSM 6000 Datenblatt”, 2012.

^dSGX Sensortech, “Thermal Conductivity Gas Detector Elements, VQ31 Series”, 2007.

^e https://www.mksinst.com/docs/UR/AIRGARDsiloxane-specifications.aspx (accessed on 15/02/2019).

PRACTICAL APPLICATION

Online monitoring of the gas quality of anaerobic digestion processes in lab‐ and full‐scale, which is of particular importance for a substrate flexible and load dependent operation of biogas systems.

Regarding technologies that are currently being investigated at a research level, tunable diode laser spectroscopy 10 and Raman spectroscopy 11, 12, 13 offer provide promising possibilities in terms of resolution and broadband detection of different molecules, respectively. However, both approaches require bulky and expensive infrastructure to operate, meaning that adaptation for mass‐deployment is difficult. Further optical techniques for anaerobic digestion processes monitoring 14, 15, 16 include colorimetric sensors 17, fluorescence spectroscopy 18, fibre optics based sensing 19, and ultraviolet and visible spectroscopy 20.

In this work we have investigated the potential of using a photoacoustic detector in a NDIR configuration using a miniaturized setup akin to the original Ultrarot Absorptions Schreiber design 21. In the Ultrarot Absorptions Schreiber scheme, light is directed along two parallel channels serving as reference and measurement channel, respectively. After passing the channels, light is directed onto a hermetically sealed, target gas filled two chamber cell connected via a membrane. Modulation of the light source leads to the exiting soundwave whose amplitude is related to the amount of target gas in the measurement channel. About a decade ago, the first attempt to miniaturize this simple concept did use a micro‐electro‐mechanical system pressure sensor to gauge the light intensity of a thermal emitter 22, and since then different combinations of various types of broad band light sources and devices for determining the photoacoustic signal have been developed 23, 24, 25, 26, 27, 28, 29, 30, 31. While the use of lasers in combination with acoustic resonators in setups of direct photoacoustic spectroscopy allow for ultra‐sensitive trace gas detection, the use of such devices outside the laboratory is challenging 32, 33, 34, 35, 36. Therefore, we propose and test a simple, yet effective system for measuring biogas quality that may be produced at low cost to enable large‐scale deployment of a reliable, precise, and miniaturized sensor technology that may be easily installed 37 and that can meet the requirements for a control of the biogas process in terms measurement speed, accuracy, and sensitivity. To this end, a quasi‐continuous monitoring of the gas quality with measurement speed below 5 min and an absolute accuracy and precision better than 0.5% is required. In order to be able to operate in a biogas plant's harsh environment, we have built simple modules for the CO₂ and CH₄ concentration determination. The setup consists only of a light source and a non‐resonant photoacoustic detector. After a laboratory characterization and calibration showing the long‐term stability of the sensor even without a reference channel, this newly developed system was tested at the full‐scale research biogas plant “Unterer Lindenhof” located in Eningen, Germany and operated by the University of Hohenheim for a duration of 3 months.

2. MATERIALS AND METHODS

2.1. Setup of the gas sensing module

The sensing modules for CH₄ and CO₂ are based on a NDIR setup using photoacoustic detectors and the version presented here is an advanced and improved version of the device used in ref. 37. The light emitted by a mid‐infrared LED is converted into a sound wave with amplitude A_PA as a function of the LED's power E_LED, the transmission function through the probe volume T_gas, and the spectral function of absorption inside the detector A_det:

$$A_{PA}(\mathrm{c})\sim {\int }_0^{\infty }{\mathrm{E}}_{\mathrm{LED}}(\lambda )·{\mathrm{T}}_{\mathrm{gas}}(\lambda ,\mathrm{c})·{\mathrm{A}}_{det}(\lambda )\mathrm{d}\lambda .$$ (1)

The light intensity at the detector position is gauged by determining the sound wave amplitude inside the photoacoustic detectors. Here they consist of a hermetically sealed, cylindrical‐shaped miniature cell with a diameter of 7.5 mm and a length of 2 mm containing a SPU0410LR5H‐QB microphone from Knowles Electronics filled with 100% CH₄ at 1 bar and 100% CO₂ at 1 bar, respectively. Mid‐infrared LEDs (MID‐IR LEDs) serve as a light source. For the methane module, a L13771‐0330 M LED from Hamamatsu featuring a peak emission wavelength at 3.3 µm with a full width half maximum emission width of 0.6 µm is employed. The CO₂ module uses a L13201‐0430 M LED from Hamamatsu with peak emission wavelength at 4.3 µm and a full width half maximum bandwidth of 1.4 µm. The CH₄ LED output power is stated as 250 µW at 50 mA drive current while the CO₂ LED is rated at 300 µW for 80 mA drive current. Both modules are implemented in a single mount realizing an optical path length of 1 mm. The LEDs are intensity modulated using a rectangular drive pattern with 50% duty cycle at 300 Hz, which is generated and ultimately controlled using a PsoC 5LP microcontroller from Cypress. The opto‐mechanical setup, the relevant spectral spectra of the light sources and detector filling, as well as the soundwave generation process are depicted in Figure 1.

Figure 1.

(A) A CAD‐drawing of the opto‐mechanical setup showing the actual dimensions of all components and the two measurement channels to determine the content of CO₂ and CH₄, respectively. A plastic mount made of PLA allows for gas diffusion into the optical path of 1 mm in between the LED surface and the photoacoustic detector. The working principle is indicated below: via intensity modulation of the light sources a soundwave is generated inside the detectors, whose amplitude is used to determine the light intensity at the relevant wavelengths for gas sensing. (B–C) The relevant optical spectra of the LED light sources and the photoacoustic detectors, respectively. Only those spectral regions of the light emitted by the LED that coincide with absorption lines actually contribute to sound wave generation

To analyze the amplitude of the soundwave, which is a measure of the light intensity entering the photoacoustic detector, the microphone signal is converted using a 12 bit AD 5593R analog‐to‐digital converter from Analog Devices operating with 2500 samples/s and read in by the PsoC 5LP microcontroller. The Goertzel algorithm 38 is implemented in the PSoC firmware and uses 1500 samples to generate an output, which corresponds to the amplitude of the photoacoustic wave A_PA, i.e. the measure for the light intensity. The Goertzel algorithm analyzes one selectable frequency component only and since we excite an acoustic wave with a single, well‐defined frequency, the Goertzel algorithm is numerically more efficient than fast Fourier‐transform algorithms. The response R(c_gas) of the sensor system towards varying concentrations c_gas of either CH₄ or CO₂ is defined as the ratio of the photoacoustic wave amplitude A_PA,CO2/CH4 without and with the respective gas present in the probing volume minus 1, i.e.:

$$\mathrm{R}({\mathrm{c}}_{\mathrm{gas}})={\mathrm{A}}_{\mathrm{PA}}({\mathrm{c}}_{\mathrm{gas}}=0)/\mathrm{A}({\mathrm{c}}_{\mathrm{gas}})-1$$

This means that the system generates an output every 1.2 s. To assign an error for a given integration time, the SD σ from subsequent output values is calculated. To calibrate the system, the response towards varying concentration levels of methane and CO₂ has been recorded using a dedicated apparatus to expose gas measurement systems to well‐defined atmospheres 39.

Because NDIR devices measure the number density of a sample, a MPL3115A2 pressure detector from NXP has been included in the system to convert the sensor reading into a fraction stated in percent. Using the apparatus, the influence of humidity on the sensor's signal has also been investigated. Finally, the influence of temperature on the sensor's signal has been evaluated by placing the system in a thermally controlled chamber and recording the signal amplitude in the range of 15–45°C. To deploy the gas sensor system in the field, a bypass system has been designed to enable a direct plug into the main gas conduct right before the combined heat and power plant (CHP), where the produced biogas was converted into electric and thermal energy. Therefore, the system was placed in a gas tight chamber with a total volume of 1.1 L, and a small amount of biogas was diverted from the main conduct and directed into the box. The average flow of gas through the bypass is 2 L/min, which means that it takes about 99 s to attain a new steady state concentration upon changes. The integration time for the results presented here has been set to 6 min, such that in effect a convolution of the in situ concentration measured in the prototype's probing volume and the dynamic concentration in the measurement chamber determine the actual concentration reading. No further pumps or membranes are employed in this setup and our device directly measures the gas concentration inside the box.

The described prototype system has analyzed the same gas samples as the commercially available biogas analyzer “InCa 4000” from Union Instruments GmbH, which it is benchmarked against. A dedicated power supply and interface to the local control system has been implemented to enable long‐term field tests. Inside the CHP‐unit, room temperature typically varies between 20–40°C and background noise exceeds 88 dB during operation.

2.2. Operational parameters of the biogas plant and feed intake during field trial

The full‐scale research biogas plant is equipped with two identical digesters and a secondary digester each with a gross volume of 923 m³. Both digesters are fed and mixed by different feeding and mixing systems. Usually a mixture of renewable energy crops and liquid as well as solid manure is added under mesophilic conditions at 40.5°C. Solid substrate may be grinded prior to feeding using a Bio‐QZ cross‐flow grinder from MeWa, Germany. A more detailed description of the biogas plant is given in refs. 40 and 41. Unlike most biogas plants in practice, the research biogas plant offers the unique opportunity for flexible biogas production on a research level. The research biogas plant was operated under flexible biogas production mode during the field trial. The flexibility was achieved by lowering the feeding intervals from 12 times to once a day while keeping the feed amount at the same level. Hence, gas production is expected to peak during a certain time depending on the underlying digestion processes. Digesters 1 and 2 were fed alternating in flexible mode on a weekly basis. While one digester was fed once a day on a single substrate, e.g. maize‐silage, grass‐silage, ground grain, or solid manure for about 5 days, the other digester was fed steadily 12 times a day for 7 days a week on a substrate mixture of grass‐silage, maize‐silage, ground grain, and solid manure. Liquid manure was fed into the secondary digester twice a day at an amount of 4.000 L. The target feeding amount of fresh matter in flexible mode was set at either 6.000 kg/d maize silage or 6.400 kg grass‐silage or 13.700 kg horse manure. Based on the substrate‐specific methane yields, the quantities added were chosen so that the daily methane production was kept stable throughout the experiment. The steady state digester was fed daily at a target feeding amount of fresh matter of 700 kg maize silage, 2.400 kg gras silage, 1.500 kg ground grain, 3.600 kg cow‐ or horse manure. The steady‐state operation mode of one digester was kept to provide a continuous and high biogas flow to the CHP‐Unit as the farm is supplied with electricity and heat and relies on the energy from the on‐site biogas installation. The organic loading rate related to organic dry matter (oDM) of the flexible digester was set at 5.14 kg m⁻³ d⁻¹ for horse manure, 2.45 kg m⁻³ d⁻¹ for grass silage, 2.26 kg  m⁻³ d⁻¹ for maize silage, and 2.23 kg m⁻³ d⁻¹ for ground grain. The steady state digester was targeted at an organic loading rate of 3.8 kg oDM m⁻³ d⁻¹ at a hydraulic retention time of 80 days. The feeding of the flexible digester was planned to target a design point of gas production of 730  m³ per day, thus resulting in the different feeding amounts due to its different methane potential and oDM content. The steady state digester was planned to produce 1.015 m³ methane per day.

The solid materials are premixed before feeding and at digester 1 pretreated with a cross‐flow grinder 29 before feeding. The process was kept at a mesophilic temperature of 40°C in both digesters as well as the secondary digester.

The main part of the biogas is produced in the digesters and temporarily stored in the secondary digester before combustion in the CHP‐Unit. To reduce hydrogen sulfur in the biogas, ambient air is blown into the digesters and the secondary digesters headspace at an oxygen concentration of 0.5–1% with respect to the H₂S quantity measured. Before reaching the CHP‐unit, the biogas is dewatered and fine‐cleaned with activated carbon. A preliminary version of the gas sensing system has been installed at the Hohenheim biogas plant in a field trial in 2016 37. The basic setup was similar to the one presented here but to allow for long term stable operation a rugged opto‐mechanical design has been developed and implemented. This has enabled a reduction in signal noise and robustness against acoustic noise from the CHP motor. Additionally, the optical path length for both CO₂ and CH₄ channels have been optimized to strike an ideal balance between dynamic range and resolution up to 70 Vol.%. Lastly, a sensor to determine temperature and pressure inside the measurement chamber has been added to enable drift‐compensation originating from these parameters. Here, we present results obtained with this improved and calibrated version within the period from 04.10.2017–08.12.2017.

During this period the biogas plant has been fed in a flexible manner. Flexible periods with maize silage, ground grain, and grass‐silage have been performed prior to the trial period reported here. The data presented in this publication refers to a period towards the end of the flexible trials. Table 2 displays the flexible feed consisting only of grass silage and horse manure.

Table 2.

Feed intake during the trial period indicating days of flexible feeding

	Digester 1					Digester 2
	Manure	Maize silage	Grass silage	Cereal grain	Horse manure	Manure	Maize silage	Grass silage	Cereal grain	Horse manure
Trial day	kg/d	kg/d	kg/d	kg/d	kg/d	kg/d	kg/d	kg/d	kg/d	kg/d
1	–	1768	1993	1114	2717	–	–	6399*	–	–
2	–	1675	1877	1097	2948	–	–	6346*	–	–
3	–	1174	2195	1360	3503	–	–	–	–	–
4	–	1742	1811	1323	3358	–	–	–	–	–
5	–	105	67	2312	51	–	1704	2674	1079	2438
6	–	–	–	–	8300*	–	1056	2845	1111	3871
7	–	–	–	–	9811*	–	527	1719	265	2595
8	–	593	793	802	5143	–	1077	3199	–	3898
9	–	1154	2145		3521	–	1297	2640	–	3499
10	–	870	2655	1293	3368	–	930	2549	1115	2791
11	–	432	1507	724	1981	–	–	–	–	10 400*
12	–	1084	2501	1267	3972	–	–	–	–	11 742*
13	–	662	1437	694	2534	–	–	–	–	–
14	5000	485	1300	709	1561	–	–	–	–	11 695*
15	7500	1037	2305	1400	3148	–	603	1298	999	1214
16	3700	1208	2385	1648	2916	3800	1814	2851	1640	1977
17	–	1321	2727	1462	2170	3700	2015	2992	1339	1773
18	3700	974	2290		2567	3800	1113	1622		1402
19	3700	677	2019	1187	3067	3800	826	1363	909	2266
20	3800	619	1885	729	1770	3700	674	2504	1118	2246
21	3800	984	2601	1193	2348	3700	1345	2462	1348	2582
22	3800	939	2133	1163	2163	3800	1565	2372	1343	2486
23	3700	1844	2199	1279	2574	3800	1793	2404	970	2382
24	3700	1709	2136	1281	2358	3800	1882	2462	1220	2020

Flexible feeding days are marked bold and with an asterisk (*). During these days the feed has been supplied into the respective digester within 1 h.

Trial day 1 corresponds to 04/10/2017. On trial days 1–4, digester 1 was fed in steady state while digester 2 was fed with grass silage in flexible mode with intense feeding on days 1 and 2 and none on days 3 and 4. The pause at day 3 and 4 was set to simulate a weekend break where less energy is needed. In the following trial days 5–10, digester 2 was fed at steady state and digester 1 was planned to be fed on horse manure. Due to technical difficulties during feeding in this period, disturbances prevented the targeted quantity from being reached, and the experiment with horse manure was cancelled. The standard feeding mixture was later fed into digester 1. From day 11 to 15, digester 2 was fed flexible with horse manure. The data from feed intake shows that due to feed availability on biogas plants or technical failures, the feed composition may vary within a very short time, thus leading to a rapid change in gas quantity and quality. Figure 2 summarizes the feeding during the trial period and presents an overview photo of the experimental biogas plant in Eningen, Germany.

Figure 2.

(A) A graph of the feeding rate for digester 1. The complete amount of horse manure has been fed within a short period of time on day 6 and 7, respectively. During the remainder of time continuous feeding fuelled digester 1. (B) The corresponding graph for digester 2, showing the flexible feeding on days 1, 2, 11, 12, and 14. (C) Aerial view of the experimental biogas plant at Eningen, Germany. Both digester feed into the secondary fermenter, which is why the measured biogas quality is always a mixture as generated by both digesters and the secondary fermenter

3. RESULTS AND DISCUSSION

To check the suitability of the system for biogas quality control, a basic test of resolution, LOD, and dynamic range has been performed in the laboratory for both sensor modules. In order to check the repeatability of the prototype device, the measurement has been repeated five times for both gases. The results are depicted in Figure 3 and show the response to varying levels of both gases in the range up to 70% at 1 bar ambient pressure. The error bars in Figure 3 correspond to one standard deviation σ calculated from 300 measurement points corresponding to an integration time of 360 s for each run of the measurement. Even though advanced techniques for the estimation of the LOD are available 42, the current International Union of Pure and Applied Chemistry (IUPAC) definition has been applied here 43. Based on these measurements, the sensitivity has been determined using polynomial fits of first and third order to the CO₂ and CH₄ module, respectively. To determine the LOD and according to ref. 44, we use 3.3 σ and assuming the validity of a linear approximation of around 0% probe gas concentration, a linear sensitivity with slope α_CO2/CH4, LOD has been determined using LOD = 3.3σ/α_CO2/CH4, which yields 5024 parts per million (ppm) for CH₄ and 2053 ppm for CO₂. Because of the linear behavior of the CO₂ channel, the dynamic range and resolution of this channel extend the complete concentration range. For the CH₄ channel, the absolute resolution diminishes slightly, but average resolution over the complete range is 7587 ppm, which is sufficient for application in a biogas plant (Table 3).

Figure 3.

(A) Transient response of both modules to varying levels of methane and carbon dioxide, respectively. No cross‐talk between the modules is detected. (B, C) Response of both channels including the corresponding fit functions averaging five repetitions of the same gas measurement cycle. The error bars displayed correspond to 1σ

Table 3.

Calibration results for both modules

Module	Sensitivity	Limit of detection	Average resolution
CO2	0.084 mPa/%	2053 ppm	2053 ppm
CH4	0.058 mPa/%	5024 ppm	7587 ppm

LOD and average resolution are calculated using 3.3σ from 300 measurement points.

To explore the stability of the sensor signal and the possibility to lower the LOD by varying the integration time, an Allan plot has been calculated for both gas sensing modules using 1σ for all signals. The result is shown in Figure 4 and it highlights the high degree of stability as well as the potential for improved resolution and LOD as a result of increased integration times. Depending on the actual requirements for active control of a biogas reactor, these values can thus easily be adjusted for the miniature biogas sensing device.

Figure 4.

Allan deviation for the methane and carbon dioxide channels, respectively. Each measurement point is obtained after analyzing 1500 samples, i.e. 1.2 s integration of the microphone signal. Both channels feature a minimum deviation after about 2000 measurement points, corresponding to 40 min of integration time thus highlighting the stability of the device

An evaluation of the dependence of the sensor's signal on temperature and humidity has been performed in the laboratory in a pure nitrogen environment. The results are depicted in Figure 6A and show no influence of humidity on the respective sensor signals. However, the change in ambient temperature in the measurement chamber does influence the signal significantly. To enable a test of larger temperature ranges, the system, including the electronics, has been placed in a climatic chamber and the change in signal as a function of temperature has been recorded in the range of 15–45°C. Figure 5B shows the results revealing a linear dependence of the signal on temperature, which is in accordance with previously obtained results 23 and in contrast to systems employing only semiconductor components, which show exponential dependence on changes in temperature.

Figure 6.

Section of the real‐world test showing a comparison between the commercial InCa device with the photoacoustic NDIR prototypes. The sensor readings agree within the margin of error during the test. The most notable difference from trial day 15–17 (27/10/2019–29/10/2019) coincides with exceptionally high hydrogen sulfide concentrations, and adverse effects on the sensing system cannot be ruled out. However, the signal recuperates and the reasons for this should be investigated further. (A) Reading of the methane content in the bypass. (B) Reading of the carbon dioxide content in the bypass

Figure 5.

(A) Influence of humidity and temperature on both channels of the biogas quality sensing system. Even changes as large as 50% r.H. at 25°C do not affect the sensor signal. On the other hand, thermal drifts in the output signal may be observed for changes in temperature of about 0.5°C. (B) A thermal calibration of the signal output reveals a linear dependency on temperature

Based on these results, a compensation function has been implemented for both modules to correct for changes in ambient temperature during operation. The following formulae correct the output of the Goertzel algorithm:

$${S_{CH}}_4\left(T\right)=-\left(2,30\ast {10}^{-5}Pa\right)\ast \left(T_{gas}-{25}^{\circ }C\right)$$

$${S_{CO}}_2\left(T\right)=-\left(1,91\ast {10}^{-5}Pa\right)\ast \left(T_{gas}-{25}^{\circ }C\right)$$

An uncompensated temperature change of 10 K reveals a calculated concentration error of 4% for methane and 1.5% for CO₂. This means that the compensation is necessary for a reliable and stable measurement system.

The gas composition has been monitored for a period of over 2 months and here we concentrate on the flexible feeding period of 24 days starting from 13/10/2017 as trial day 1. During this time, the time‐resolved variation of the two main components methane and carbon dioxide are shown. In Figure 6A, direct comparison between the commercially available, state‐of‐the‐art InCa 4000 sensing system from Union Instruments GmbH, Germany is shown against the readings from the miniaturized prototype.

The readings are in agreement within the respective errors of the systems except for trials days 15–17, where the photoacoustic based system shows a drop of about 7% as compared to the InCa system. One reason might be high levels of hydrogen sulfide during this period but the exact cause is unknown and because the signal later recuperates, the origin cannot be traced via post‐analysis of the system.

Most importantly, the photoacoustic system provides gas quality data every few minutes, while the InCa system is only capable to do so every 30 min. This provides the necessary measurement frequency to monitor flexible feeding processes in real‐time. For the remainder of the discussion, we use the reading from the photoacoustic system. The data shows a slightly varying CH₄ quality with some periods of massive deviations. In Figure 7, the gas volume produced in each fermenter as well as the gas quality readings are depicted.

Figure 7.

Gas volume and gas quality data during the field trial with flexible feeding. The shaded areas indicate day with feeding daily fuel at once, whereas the rest of the days feed has been executed quasi‐continuously. The grey marked area stands for grass silage feeding and the green areas for horse manure. The effects of the feeding strategy are visible both in gas volume and gas quality

The rise of methane on days 7, 8, 9, 12, and 13 may be explained by the flexible feed input of horse manure. Due to mentioned technical problems in digester 1, the horse manure could not be fed. The missing feed might have led to a higher production of methane due to lower hydrolyses activity, where CO₂ is produced, as the microorganisms had nothing to digest. The production of CO₂ was restarted with the supply of feedstock which led to a lower methane quality short after. The change in gas composition on days 16–17 may be explained by the high quantity of the dosed horse manure quantity from days 11 to 14. As a result of horse manure's slow decomposition, methanogenesis is delayed. This can be observed when the gas quantity increases and the gas quality simultaneously decreases, i.e. an increase in CO₂. Later the methane quality rises again, which indicates that the degradation of the horse manure was almost completed.

The massive input of untreated horse manure in digester 2 changes the conditions for microorganisms in the digester drastically. Due to its composition, which includes a high proportion of lignin, cellulose, or hemi‐cellulose, horse manure degrades more slowly than maize silage and grass silage, resulting in a corresponding decrease in gas quality over the following days. Lignin, cellulose, or hemi‐cellulose have to be metabolized before methane can be produced. The methane decline is much higher than the CO₂ increase, due to the fact that methanogenesis is inhibited and the CO₂ consumption ceases accordingly.

The results presented here do concur to a considerable degree with the readings from a state of the art gas measurement device. The main deviation occurs in the methane channel reading and the timing coincides with extraordinary high concentrations of H₂S. Cross‐sensitivities towards this latter gas can be ruled out due to the working principle of the photoacoustic sensor setup employed 23, 26. Likewise, mechanical instabilities are unlikely the cause of this since the adjacent CO₂ channel does not suffer at all. Since no optics is necessary in this miniaturized setup and because the signal recovers without any intervention, the cause is likely from a nonideal contact of one of the electronic components. The sensor's measurement speed is fast as compared to typically time‐scales in a bioreactor during flexible feeding and thus sufficient to monitor changes in the biogas composition resulting thereof. Even though the flexible feeding period monitored here is short as compared to typical lifecycles of a biogas plant, the high degree of stability and the ease of installation point towards the suitability of the approach for deployment at biogas plants that currently do not use gas measurement technology at all. Future work will consequently focus on increasing the robustness of the system and test possible long‐term drifts.

4. CONCLUDING REMARKS

We have presented results of a field trial using a prototype gas measurement system for biogas quality that has been characterized previously in the laboratory. The optical path used to determine the content of carbon dioxide and methane is 1 mm, which is sufficient to obtain an average absolute resolution of 2084 and 7475 ppm for CO₂ and CH₄, respectively. No reference channel has been implemented because changes in the baseline signal are only caused by temperature variations, which can be corrected for. The output of the prototype system provides a gas concentration reading with a resolution similar to commercially available systems but at a much reduced size and considerably higher measurement frequency, thus we could detect changes in process‐biology much faster now. Because of the miniaturized design, the system may be easily installed as an in situ device in a bypass to the main gas conduct. This approach may pave the way towards a large‐scale deployment of reliable, robust, and sensitive gas measurement equipment for biogas plants because it is easy to install and cost‐efficient method for the determination of biogas quality. This in turn may lead to a more efficient operation of those plants.

CONFLICT OF INTEREST

The authors have declared no conflict of interest.

Bierer B, Kress P, Nägele H‐J, Lemmer A, Palzer S. Investigating flexible feeding effects on the biogas quality in full‐scale anaerobic digestion by high resolution, photoacoustic‐based NDIR sensing. Eng Life Sci. 2019;19:700–710. 10.1002/elsc.201900046