Methods for assessing the performance of a bioreactor

Abstract

Described herein is a computer-implemented method for determining an operational performance of a bioreactor for the production of biogas by the anaerobic digestion of organic material inside a chamber of the bioreactor. The method comprises: collecting data related to an operation of the bioreactor; using the collected data as an input to a machine learning model; and determining, using the machine learning model the operational performance of the bioreactor. The machine learning model is trained to determine an operational performance of a bioreactor using data related to an operation of said bioreactor.

Description

[0001] METHODS FOR ASSESSING THE PERFORMANCE OF A BIOREACTOR This application claims priority from GB2501165.1 filed 27 January 2025, the contents and elements of which are herein incorporated by reference for all purposes.

[0002] Field of the Invention

[0003] The present invention relates to methods for assessing the operating performance of a bioreactor and particularly, although not exclusively, to methods for assessing the operating performance of bioreactors for use in the treatment and processing of wastewater and / or organic material for the production of electricity and / or biogas.

[0004] Background

[0005] Bioreactors, such as bioelectrochemical systems (BES), may be used to process wastewater and / or organic waste to produce electricity and / or fuel such as biogas. Some such systems may include microorganisms located inside a chamber of the bioreactor for anaerobically digesting input waste to produce biogas comprising, for example, methane.

[0006] The stability and efficiency of anaerobic digestion processes may, however, be very unstable and are susceptible to influence by variation of any of a large number of parameters associated with the environmental conditions in which the anaerobic digestion processes are taking place. For this reason, it is often necessary to monitor the environmental conditions of bioreactors to ensure that the stability and efficiency of the bioreactor can be optimised. This has generally involved collecting samples from the chamber of a bioreactor, taking the samples to a laboratory, and subjecting the samples to suitable laboratory tests to determine the environmental conditions of the bioreactor.

[0007] The present invention has been devised in light of the above considerations.

[0008] Summary of the Invention

[0009] In a general sense, described herein are methods for facilitating the real-time monitoring of the environmental conditions of a bioreactor for the purpose of optimising the stability and efficiency of biotechnological processes being carried out therein. The methods described herein leverage the selflearning and adaptive capabilities of one or more machine learning models to analyse one or more carefully selected parameters to determine, e.g., in a quantitative manner, the operational performance of the bioreactor. Some of the methods described herein may further involve implementing real-time control of the environmental conditions of the bioreactor in response to the analysis implemented by the one or more machine learning models.

[0010] In a first aspect, there is provided a computer-implemented method for determining an operational performance of a bioreactor for the production of biogas by the anaerobic digestion of organic material inside a chamber of the bioreactor. The method comprises: collecting data related to an operation of the bioreactor; using the collected data as an input to a machine learning model; and determining, using the machine learning model, the operational performance of the bioreactor. The machine learning model has8911927

[0011] 2

[0012] been trained to determine an operational performance of a bioreactor using data related to an operation of said bioreactor.

[0013] The methods described herein may be used to assess the operational performance of bioreactors that contain microorganisms inside a chamber of the bioreactor. These microorganisms may anaerobically digest organic material to produce fuel - e.g., biogas.

[0014] The bioreactors to which the methods described herein may be applied may, for example include any suitable form of bioreactor - e.g., bioelectrochemical systems, anaerobic digestors, and the like. The methods described herein may be suited to application to a microbial electrolysis cell assisted anaerobic digestion (MEC-AD).

[0015] In some examples, as discussed in more detail below, the bioreactor may be a microbial electrolysis cell system comprising a plurality of electrodes. Microorganisms may coat the plurality of electrodes such that, upon application of a voltage between adjacent electrodes of the plurality of electrodes, the microorganisms coating the plurality of electrodes are induced to anaerobically digest organic material in their vicinity. That is, the microorganisms may be electrically activated and their digestive activity may change (e.g., increase) in response to the application of an electrical signal.

[0016] As an example, the bioreactor may be an electromethanogenic reactor having methanogenic microorganisms contained therein for anaerobically digesting organic material (e.g., comprising acetate and hydrogen, amongst other materials) to produce methane. That is, the biogas may comprise methane. In such contexts, the produced methane may be referred to as “biomethane”. Additionally or alternatively, the biogas produced by the bioreactor may comprise hydrogen. In such cases, the produced hydrogen may be referred to as “biohydrogen”.

[0017] The collected data may be data indicative of the environmental conditions under which the anaerobic digestion processes are taking place. In this way, the collected data may be used by the machine learning model to determine whether the environment in which the anaerobic digestion processes are taking place is suitable and / or optimised to promote efficiency and / or stability of the anaerobic digestion processes. The collected data may compiled to form an input dataset for the trained machine learning model (e.g., for inference). For example, the collected data may be collected from a variety of sources (e.g., one or more sensors). This collected data may be sent from each respective source to a central unit. The central unit may, for example, be a controller communicatively connected to each of the one or more sensors. In some examples, as discussed below, the bioreactor may be divided into zones. In such cases, each zone may be communicatively connected to a respective zonal unit such that respective sensors within each zone are communicatively connected to a corresponding zonal unit. Each of the zonal units may be communicatively connected to the central unit. In cases with zonal units, each zonal unit may store a respective copy of the machine learning model or may be able to access (e.g., via a server of the central unit) the machine learning model for execution.

[0018] Compiling the collected data into one or more input datasets may simplify the process of providing input to the machine learning model. For example, all the collected data may be compiled into a single input8911927

[0019] 3

[0020] dataset for input into the machine learning model so that the machine learning model can parse, if necessary, and analyse the collected data. Alternatively, the collected data may be compiled into a plurality of input datasets that are each respectively input to the machine learning model at an appropriate point in its processing. This may reduce the memory requirements of the machine learning model.

[0021] The machine learning model may have any suitable architecture for analysing the collected data to determine the operational performance of the bioreactor. For example, the present inventors have found that Decision Trees, XGBoost, Support Vector Machine (SVM), and Artificial Neural Network (ANN) architectures may all be suitable for assessing the operational performance of a bioreactor, either independently or in any suitable combination. In some cases, an SVM architecture may be implemented using a Support Vector Regression-type architecture.

[0022] In particular configurations, an ANN architecture may be particularly suitable for assessing the operational performance of a bioreactor. In some cases, spatial variations in the operational performance of the bioreactor may be determined. In such cases, the ANN architecture may be modified to account for spatial variation, for example by using a convolutional neural network (CNN) architecture, or similar. Additionally or alternatively, in some cases, temporal variations in the operational performance of the bioreactor may be determined. In such cases, the ANN architecture may be modified to account for temporal variation / trends, for example by using a recurrent neural network (RNN) architecture, and / or a long short-term memory (LSTM) architecture, or similar. In some examples, the ANN may be trained to execute pattern recognition, for example the ANN may have a Hidden Markov Model (HMM) architecture, or similar.

[0023] The machine learning model may be trained using a supervised learning approach to ensure proper calibration and validation of the machine learning model.

[0024] In some configurations, the machine learning model may comprise a plurality of sub-models configured to determine the operational performance of the bioreactor, each of said sub-models relating to a respective operational state of the bioreactor. The collected data (i.e. the data related to operation of the bioreactor) may correspond to one of said operational states of the bioreactor. The machine learning model may be configured to select which of the sub-models to use to determine the operational performance of the bioreactor based on the operational state of the bioreactor that the collected data (i.e. the data related to operation of the bioreactor) corresponds to. In this way, the machine learning model can switch between different sub-models based on the operational state of the bioreactor.

[0025] Each sub-model configured to determine the operational performance of the bioreactor may be trained on data relating to the operational state of the bioreactor that said sub-model relates to. By way of example, each sub-model (e.g. a start-up sub-model) may have been trained on more data relating to the operational state it corresponds to (e.g. start-up) than data relating to an operational state that it does not correspond to (e.g. stable low-load operation). In this way, the prediction accuracy and robustness of each sub-model may be improved for when the operational state of the bioreactor is the state said submodel corresponds to, compared to where there is a single model that is used for determination of the operational performance of the bioreactor across all operational states.8911927

[0026] 4

[0027] The machine learning model may further comprise a clustering sub-model configured to determine the operational state of the bioreactor based on the collected data and accordingly select which of the submodels configured to determine the operational performance of the bioreactor to use to determine the operational performance of the bioreactor (i.e. based on the determined operational state). In some examples, the clustering model may use an unsupervised machine learning method to determine the operational state of the bioreactor (e.g. principal component analysis). The clustering sub-model may determine the operational state of the bioreactor based on data collected from online sensors. Where PCA is used by the clustering sub-model, the collected data may be projected into a low dimensional PCA space (such as to capture the dominant modes of process variability). Clustering (e.g. K-Means clustering), or regional definitions, may be applied to the low dimension PCA space to identify distinct operational states.

[0028] The collected data used to determine the operational state of the bioreactor may comprise, or consist of, one or more of: a feed rate of organic material into the chamber; a temperature inside the chamber, current data for a subset of electrode pairs within the bioreactor (where present), the pH inside the chamber; a volume and / or rate of production of biogas, and hydraulic retention time for the chamber. The operational states for the bioreactor may comprise, or consist of, one or more of: start-up, stable low load operation, stable high load operation, and recovery after disturbance.

[0029] The machine learning model (e.g. the clustering sub-model) may determine the operational state of the bioreactor in real time based on the data input into the machine learning model (i.e. the present operational state). In this way, the machine learning model is able to switch between sub-models for determining the operational performance of the bioreactor in real time as it is determined that the operational state of the bioreactor has changed.

[0030] The input(s) into the clustering sub-model may be the same as, or different to, the input(s) into the submodels configured to determine the operational performance of the bioreactor (e.g. some, but not all, of the input parameters may be common between the two sets of input(s)).

[0031] Determining the operational performance of the bioreactor may for example, involve quantifying, by the machine learning model, the operational performance of the bioreactor using the collected data. For example, the determined operational performance of the bioreactor may be an output of the machine learning model.

[0032] The output of the machine learning model may, for example, include an indication of temporal trends and / or variations in the operational performance of the bioreactor. Additionally or alternatively, the output of the machine learning model may, for example, include an indication of a quantitative or qualitative value for one or more operational parameters of the bioreactor. For example, the output of the machine learning model may, for each of a set of operational parameters of the bioreactor, include a quantitative value of said operational parameters and / or a qualitative indication or recommendation that indicates to a user of the methods described herein whether action needs to be taken to improve the efficiency and / or8911927

[0033] 5

[0034] stability of the anaerobic digestion processes within the bioreactor. For example, the recommendation may indicate to the user a specific course of action to modify one or more of the operational parameters in a particular way to improve the efficiency and / or stability of the bioreactor. Additionally or alternatively, the recommendation may indicate to the user that an improvement in the efficiency and / or stability of the bioreactor is desirable, necessary and / or possible.

[0035] The determined operational performance may be determined in absolute terms and / or may be determined relative to a baseline operational performance so that a user of the methods described herein can understand the absolute and / or performance level of the bioreactor.

[0036] The methods described herein may be carried out in real-time during operation of the bioreactor.

[0037] For example, any one or more of the operations of the method may be carried out in real-time or near real-time. It may be understood that, as used herein, the term “real-time” may infer that the method of assessing an operational performance of a bioreactor provides the determined operational performance of the bioreactor to a user of the methods during the same operational cycle of the bioreactor as the cycle during which the date related to the operation of the bioreactor is collected.

[0038] That is, the capabilities of the machine learning model may be leveraged to determine the operational performance of the bioreactor without the need to collect samples from the bioreactor and send them to a (possibly off-site) laboratory fortesting. In this way, delays in the determination of the operational performance of the bioreactor may be eliminated. This may facilitate real-time adjustments in the control of the bioreactor to achieved optimised or near-optimised efficiency and / or stability in the operation of the bioreactor.

[0039] The collected data may include one or more of: pH data indicative of a pH inside the chamber during operation; contaminant data indicative of an amount of contaminant microorganism inside the chamber during operation; production data indicative of an amount of biogas produced by the bioreactor during operation; loading data indicative of an organic load rate with which organic material is loaded into the bioreactor for processing into biogas; composition data indicative of a composition of the biogas produced by the bioreactor during operation; and / or oxygen demand data indicative of an amount of chemical oxygen demand of one or more sections of the bioreactor.

[0040] The pH data may for example, be data indicating the quantitative value of the pH detected by one or more pH sensors disposed in the chamber of the bioreactor. For example, in cases where the bioreactor is an electromethanogenic reactor having a plurality of electrodes installed therein, the one or more pH sensors may each be positioned proximal to a corresponding electrode (or pair of electrodes, or subset of ethe plurality of electrodes) to obtain corresponding pH data associated with the pH of the solution inside the chamber of the bioreactor in the vicinity of the corresponding electrode(s).

[0041] Additionally or alternatively, the pH data may include the results of a laboratory pH test (e.g. , using a pH indicator and / or a pH probe) carried out on a sample collected from the chamber of the bioreactor.

[0042] In the context of anaerobic digestion bioreactors, it is known that volatile fatty acids can accumulate over the course of the operation of the bioreactor. As the concentration of volatile fatty acids increases, the8911927

[0043] 6

[0044] anaerobic digestion of organic molecules such as acetate may be inhibited, thereby reducing the efficiency of the bioreactor. Additionally, a large build-up of volatile fatty acids may induce acidosis in the methanogenic microorganisms, thereby causing a system failure of the bioreactor (e.g., potentially severely reducing the stability of the bioreactor’s anaerobic digestion processes). As such, a pH below a predetermined threshold may indicate an excessive build-up of volatile fatty acids that may negatively impact the efficiency and / or stability of the bioreactor’s operation.

[0045] In some examples, a subset of the pH data may correspond to one or more pH values measured within the chamber of the bioreactor - that is the chamber in which the anaerobic digestion processes take place. Additionally, a subset of the pH data may correspond to one or more pH values measured within an equalisation tank (or holding tank) of the bioreactor. Organic material (e.g., organic waste) may be fed into the chamber of the bioreactor from the equalisation tank. Comparing the pH value(s) of the solution within the equalisation tank to the pH value(s) of the solution within the chamber may provide an indication of the extent to which volatile fatty acids concentrations have changed over the course of the operation of the bioreactor and / or may indicate a concentration of volatile fatty acids (or other acids) present in the organic material fed through to the chamber for anaerobic digestion.

[0046] Contaminant data may include data that quantifies and / or indicates an amount of contaminant microorganism inside the chamber during operation. For example, in the context of a methanogenic reactor, the bioreactor microorganisms may be methanogens. Methanogens are known to compete for feedstock with sulphate-reducing bacteria as both sulphate-reducing bacteria and methanogens use substrates such as hydrogen and acetate fortheir respective anaerobic digestion processes. In contrast to methanogens (which anaerobically digest hydrogen and acetate to produce methane), sulphate-reducing bacteria process hydrogen and acetate to produce sulphur-containing gases such as hydrogen sulphide.

[0047] As such, a concentration of hydrogen sulphide produce by the bioreactor may be indicative of the presence and / or amount of sulphate-reducing bacteria present in the bioreactor.

[0048] Contaminant data may therefore include data associated with one or more concentration measurements that indicate a concentration of a known contaminant-product at one or more positions within the chamber of the bioreactor. The known contaminant-product may be a known product produced by the contaminant microorganism when processing the organic material. For example, if the contaminant microorganism is a sulphate-reducing bacteria, the contaminant-product may be hydrogen sulphide.

[0049] As the amount of contaminant microorganism within the chamber increases, the competition for organic material also increases, which reduces the output of the anaerobic digestion processes by the bioreactor microorganisms. As such, the efficiency of the bioreactor (e.g., the biogas production output - either in terms of volume or concentration of target product) may be reduced.

[0050] Alternatively, if the activity of bioreactor microorganisms is inhibited (e.g., due to the pH and / or temperature within the chamber), the contaminant microorganisms may be able to thrive which may8911927

[0051] 7

[0052] increase the concentration of the known contaminant-product. In this way, the contaminant data may be indicative of an environmental issue inhibiting the performance of the bioreactor.

[0053] The production data may, for example, be a quantitative measurement of the rate with which biogas is produced, and / or an amount (e.g., a volume) of biogas produced per amount of organic material, and / or an amount (e.g., a volume) of biogas produced during an operation cycle of the bioreactor.

[0054] A relatively higher output of biogas (e.g., relative to a baseline output value) may be indicative of a relatively higher operational efficiency of the bioreactor. Conversely, a relatively lower output of biogas (e.g., relative to a baseline output value) may be indicative of a relatively lower operational efficiency of the bioreactor.

[0055] Similarly, a greater degree of variation in the output of biogas over time may be indicative of lower operational stability of the bioreactor, while a lower degree of variation in the output of biogas over time may be indicative of higher operational stability of the bioreactor.

[0056] The loading data may, for example, quantify (e.g., in terms of mass) the amount of organic material loaded into, or fed into, the chamber of the bioreactor. For example, the loading data may be indicative of the mass rate with which organic material is fed into the chamber of the bioreactor.

[0057] The loading data may be useable, for example, to infer whether sufficient organic material is being fed into the chamber of the bioreactor to generate a target quantity of biogas and / or whether the chamber of the bioreactor is becoming clogged, or overloaded, with organic material.

[0058] As with the production data, the loading data may be useable to infer the efficiency and / or stability of the operation of the bioreactor. For example, a comparison of the loading data with the production data may be indicative of an operational efficiency of the bioreactor, while variations overtime may be indicative of an operational stability of the bioreactor.

[0059] Composition data may include a quantitative indication of a composition of the biogas. For example, the composition data may include data quantifying the percentage of the produced biogas that is formed from a target compound or species. As an example, in the case of a methanogenic reactor, the composition data may include an indication of the percentage of the produced biogas that is composed of methane. Additionally or alternatively, the composition data may include data quantifying the percentage of the produced biogas that is formed from other gases, such as hydrogen and / or carbon dioxide.

[0060] The composition data may therefore be indicative of an operational efficiency with which the organic material is converted into desired biogas. Variation in the composition data overtime may be indicative of an operational stability of the bioreactor.

[0061] The oxygen demand data may quantify, for example, the chemical oxygen demand in the chamber of the bioreactor and / or in the equalisation tank (or holding tank) of the bioreactor. Fluctuations in chemical oxygen demand are known to risk inhibiting, or even starving, the bioreactor microorganisms. As such, relatively larger fluctuations in the chemical oxygen demand (e.g., the relative chemical oxygen demand8911927

[0062] 8

[0063] between the chamber and the equalisation tank, or holding tank) may be indicative of a relatively lower operational stability of the bioreactor.

[0064] In the context of the methods described herein, it has been found that the pH data, contaminant data, production data, loading data, composition data and oxygen demand data may be the parameters having the highest feature importance fortraining an accurate and unbiased machine learning model and so, subsequently, may be parameters having the highest impact on the trained machine learning model’s accuracy.

[0065] For example, when carrying out feature importance analysis on an exemplary machine learning model it was found that the most impactful feature for the output of the machine learning model was the pH data, followed by the contaminant data, followed by the production data, followed by the loading data, followed by the composition data, followed by oxygen demand data indicative of a chemical oxygen demand in the equalisation tank of the bioreactor, followed by oxygen demand data indicative of a chemical oxygen demand in the chamber of the bioreactor. In this feature importance analysis it was found that other collected data indicative of other parameters related to the operation of the bioreactor had negligible impact on the performance of the machine learning model.

[0066] Determining the operational performance of the bioreactor may comprise one or more of: determining whether an environment in the chamber is suitable for the anaerobic digestion of organic material inside the chamber; determining whether a quantity of organic material supplied to the chamber is suitable for producing a target quantity of biogas; and / or determining a FOS / TAC ratio associated with the environment in the chamber.

[0067] Determining whether the environment in the chamber is suitable for the anaerobic digestion of organic material inside the chamber may, for example, involve determining whether one or more environmental parameters are within a predefined range. For example, whether the pH, temperature, pressure and / or amount of contaminant microorganisms in the chamber of the bioreactor is suitable for the efficient and stable anaerobic digestion of organic material. Whether an environment is suitable for the anaerobic digestion of the organic material may be dependent on the context in which the bioreactor is used. Accordingly, it may be beneficial to calibrate and / or validate the machine learning model for use in assessing the operational performance of the specific bioreactor against one or more predetermined operational benchmarks. For example, each environmental condition may have associated therewith a predetermined range, within which the environmental condition may be considered to be suitable for the anaerobic digestion of the organic material. If any environmental condition is determined to lie outside its predetermined range, it may be determined that the environment is not suitable for the anaerobic digestion of the organic material. In contrast, if all the environmental conditions are determined to lie within their respective predetermined ranges, it may be determined that the environment is suitable for the anaerobic digestion of the organic material.

[0068] Determining whether a quantity of organic material supplied to the chamber is suitable for producing a target quantity of biogas may involve determining an operational efficiency of the bioreactor and determining, based on said operational efficiency, an amount of organic material needed to produce the8911927

[0069] 9

[0070] target quantity of biogas. If the loading data is indicative that the amount of organic material being fed to the chamber of the bioreactor is less than this needed weight, it may be determined that the amount of organic material is not suitable for producing the target quantity of biogas. In contrast, if the loading data is indicative that the amount of organic material being fed to the chamber is at least this needed weight, it may be determined that the amount of organic material is suitable for producing the target quantity of biogas.

[0071] The FOS / TAC ratio is a known quantity that may be considered to be equivalent to the ratio of volatile fatty acid concentration (FOS) to the solution’s total alkalinity or buffering capacity (TAC). The FOS / TAC ratio may be determined to infer the suitability of the environment in the chamber for the anaerobic digestion of organic material inside the chamber. For example, if the determined FOS / TAC ratio is within a predetermined range, it may be determined that the environment inside the chamber is suitable for the anaerobic digestion of the organic material. The predetermined range may, for example, be between 0.3 and 0.4. The predetermined range may, for example, vary depending on the context in which the bioreactor is used to process organic material to produce biogas.

[0072] In some examples, the FOS / TAC ratio may be inferred using at least the pH data and the contaminant data.

[0073] Deviations, variations and / or fluctuations in the FOS / TAC ratio may be indicative of an operational instability in the bioreactor. For example, an elevated FOS / TAC ratio (i.e., above the upper limit of the predetermined range) may indicate an excessive build-up of volatile fatty acids, or inadequate solution alkalinity that may inhibit the activity (e.g., methanogenic activity) of the bioreactor microorganisms. In contrast, as an example, a depressed FOS / TAC ratio (i.e., below the lower limit of the predetermined range) may indicate an insufficient organic loading rate which may reduce the biogas yield of the bioreactor.

[0074] In conventional systems, determination of the FOS / TAC ratio may be a costly and inefficient exercise. By determining the FOS / TAC ratio using the machine learning model, efficient real-time (or near real-time) assessment of the operational performance of the bioreactor may be achieved without requiring timeconsuming and possibly inaccessible or remote laboratory testing. The machine learning model may therefore be leveraged to facilitate real-time control to optimise the efficiency and / or stability of the operation of the bioreactor.

[0075] Determining the operational performance of the bioreactor may comprise: determining one or more operational parameters of the bioreactor; and determining, using the one or more determined operational parameters, an operation score for the bioreactor.

[0076] The one or more operational parameters may, for example, include any one or more of: an operational efficiency of the biogas production (e.g., an efficiency with which organic material is processed into biogas); an operational stability of the biogas production (e.g., a quantified indication of the temporal variation in biogas production); a concentration of contaminants, e.g., contaminant-products and / or volatile fatty acids, in the chamber of the bioreactor; one or more pH values inside the chamber of the8911927

[0077] 10

[0078] bioreactor (e.g., a pH value for each of a plurality of zones of the chamber); one or more temperature values inside the chamber (e.g., a temperature value for each of a plurality of zones of the chamber); a feed rate with which organic material is fed into the chamber; an output rate with which biogas is produced by the bioreactor; one or more FOS / TAC ratio values (e.g., a FOS / TAC ratio value for each of a plurality of zones of the chamber); and / or a chemical oxygen demand of the chamber and / or of an equalisation tank of the bioreactor.

[0079] Determining the operation score for the bioreactor may involve, for example, determining an individualised score for each of the determined one or more operational parameters. Each individualised score may, for example, be combined (e.g., by aggregation or averaging) to determine the operation score. Alternatively, the operation score may be a composite score defined by the one or more individualised scores. The term ‘composite score’, as used herein may refer to a score comprising a plurality of constituent scores arranged in a data structure (e.g., a plurality of constituent scores arranged in an array). Each of the plurality of constituent scores may correspond to one of the one or more individualised scores. In this way, interrogating the composite score may also be suitable for interrogating each of the one or more individualised scores.

[0080] In some cases, determining the operation score may involve, executing a gradient mapping operation and / or determining a gradient variance in any one or more of the operational parameters of the bioreactor.

[0081] Combining the individualised scores to determine the operation score may involve weighting the individualised scores based on their relative importance to the operational performance of the bioreactor. For example, pH-based scores, contaminant-based scores and / or FOS / T AC-based scores may be weighted more heavily than other individualised scores.

[0082] Each of the operation score and, optionally, the one or more individualised scores may be a numerical (or quantitative score), or may be a qualitative score indicating whether the bioreactor is operating in a manner suitable for the anaerobic digestion of the organic material to produce biogas. For example, the qualitative score may be expressed in natural language form to indicate whether the bioreactor is in “good” operational performance condition (or similar), or in “inadequate” or “poor” operational performance condition (or similar).

[0083] In cases where the machine learning model is trained to determine spatial variation in the operation performance of the bioreactor, the operation score may reflect this spatial variation, for example, in the form of an array or a graphical representation (e.g., a heatmap).

[0084] In some examples, the operation score may include a recommendation for controlling the bioreactor to improve the operation score. For example, the recommendation may indicate whether an improvement in the operation score is necessary, desirable and / or possible in view of the determined operation score (and, optionally, the determined individualised scores). Additionally or alternatively, the recommendation may indicate a recommended course of action (e.g., introduce buffer solution, adjust temperature, adjust electrode voltage and / or current, etc.) to improve the operation score. These recommendations may be8911927

[0085] 11

[0086] derived by interrogating the individualised scores or components of the operation score to determine which factors are contributing to the current state of the bioreactor.

[0087] The methods described herein may further comprise: controlling the bioreactor to adjust one or more operating parameters of the bioreactor in response to the determined operational performance.

[0088] Controlling the bioreactor may be implemented to improve and / or maintain an operational efficiency and / or stability of the bioreactor.

[0089] Controlling the bioreactor may comprise modifying or adjusting one or more operational parameters of the bioreactor. For example, it may comprise one or more of: modifying a feed rate of organic material into the chamber; adjusting a concentration of one or more buffering agents inside the chamber; adjusting a temperature inside the chamber; adjusting an amount of microorganisms inside the chamber; activating or deactivating the bioreactor; adjusting an amount of mixing inside the chamber; adjusting an amount of material recirculated into the chamber from an effluent holding tank of the bioreactor; adjusting a concentration of one or more additives for promoting microbial growth in the microorganisms inside the chamber; and / or adjusting a concentration of one or more additives for supporting organic degradation of the organic material inside the chamber.

[0090] Modifying a feed rate of organic material into the chamber may involve increasing or decreasing the rate with which organic material is fed into the chamber of the bioreactor from an equalisation tank (or holding tank) of the bioreactor.

[0091] It may be beneficial to increase the feed rate of organic material into the chamber to increase the production output of biogas from the bioreactor if the biogas yield is considered to be low and / or if it is determined that there is capacity for additional anaerobic digestion processes within the chamber.

[0092] Conversely, if the chamber is overloaded with organic material, it may be beneficial to decrease the feed rate of organic material into the chamber to allow the microorganisms within the chamber to anaerobically digest the excess organic material and clear the backlog. This may help to prevent or removed blockages caused by the presence of excess organic material.

[0093] Adjusting a concentration of one or more buffering agents inside the chamber may facilitate the controllable adjustment of the pH of the environment within the chamber. Adjusting the concentration of one or more buffering agents may involve adding a suitable buffering agent to the chamber to adjust the pH of the environment within the chamber. For example, a basic buffering agent may be added to the chamber to increase the pH of the environment within the chamber, or an acidic buffering agent may be added to the chamber to decrease the pH of the environment.

[0094] Adjusting the temperature inside the chamber may involve controllably adjusting the output of a heater and / or temperature conditioning unit (e.g., air conditioning unit or refrigerant circuit) to ensure that the temperature inside the chamber is maintained at a level at which the microorganisms inside the chamber are able to survive and efficiently anaerobically digest organic material. The level at which the temperature should be maintained may, for example, depend on the species of microorganism provided within the chamber. A temperature that is too low may inhibit anaerobic digestion processes by the8911927

[0095] 12

[0096] microorganisms, while a temperature is too high may damage or even kill a proportion of the microorganism population within the chamber. It is therefore desirable to maintain the temperature within the chamber at an appropriate level.

[0097] Adjusting the amount of microorganisms may involve adding additional microorganisms to the chamber. For example, if it is determined that an amount of contaminant microorganism within the chamber is high enough that the competition for organic material between the contaminant microorganism population and the bioreactor microorganism population is significantly inhibiting the production of biogas, it may be beneficial to increase the amount of bioreactor microorganism relative to the amount of contaminant microorganism so that the bioreactor microorganism population can out-compete the contaminant microorganism population.

[0098] Deactivating the bioreactor may involve shutting down operation of the bioreactor, for example, in the event of system failure to allow the environmental conditions within the chamber to recover to a condition suitable for the anaerobic digestion of the organic material. For example, it may be beneficial or even necessary to deactivate the bioreactor in the event of severe overheating, a severe pH imbalance, a significant growth in the contaminant microorganism population, and / or a blockage arising from overfeeding organic material into the chamber.

[0099] When the bioreactor is deactivated, any number of corrective operations may be carried out. For example, the chamber may be flushed out and refilled with new solution to restore the pH of the environment within the chamber and / or to remove contaminant microorganisms and / or to replenish the bioreactor microorganism population. Additionally or alternatively, any blockages caused e.g., by a buildup of organic material may be cleared.

[0100] Activating (or reactivating) the bioreactor may be carried out in response to a determination that, e.g., after a period of deactivation, the environmental conditions within the chamber have reached (or returned to) a condition suitable for the anaerobic digestion of organic material.

[0101] It may be beneficial to automatically deactivate and / or (re-)activate the bioreactor in response to the determined operational performance of the bioreactor to prevent system damage arising from system failure - in the case of automatic deactivation - and reduce bioreactor downtime - in the case of automatic (re-)activation.

[0102] Adjusting an amount of mixing inside the chamber may involve adjusting an extent to which a solution comprising the organic material is mixed, either before or after being introduced into the chamber of the bioreactor. Increasing the amount of mixing inside the chamber may prevent solid organic material from settling at a bottom of the chamber and, as such, may reduce the risk of the chamber becoming overloaded with undigested organic material.

[0103] Adjusting an amount of material recirculated into the chamber from an effluent holding tank may improve the operational efficiency of the bioreactor. In some cases, after passing through the chamber, undigested organic material may be conveyed to an effluent holding tank for recirculation and / or disposal. Recirculating organic material back into the chamber from the effluent holding tank may reduce the waste8911927

[0104] 13

[0105] produced by the bioreactor and may increase the proportion of organic material that is successfully digested to produce biogas. However, at times when the chamber is already filled with organic material, recirculating organic material back into the chamber may increase the risk of the chamber becoming overloaded. As such, controlling the recirculation rate of organic material back into the chamber may be needed to both increase and decrease the recirculation rate as appropriate.

[0106] Adjusting a concentration of one or more additives may improve the operational performance of the bioreactor. For example, one of the one or more additives may be a composition that promotes microbial growth to increase the population of microorganisms inside the chamber. As another example, another of the one or more additives may be a composition (e.g., an enzyme) that supports the degradation of organic material in the chamber, thereby improving the efficiency of the anaerobic digestive processes taking place inside the chamber, and / or reducing the risk of the chamber from becoming overloaded by excess organic material.

[0107] The methods described herein may be repeated continuously or semi-continuously so as to monitor the operational performance of the bioreactor.

[0108] For example, the methods described herein may be repeated on a continual basis to provide a continuous stream of determined operational performances of the bioreactor.

[0109] Alternatively, the methods described herein may be repeated at intervals, e.g., regular intervals, to provide (optionally regular) repeated determined operational performances of the bioreactor. For example, the methods described herein may be repeated at a regular interval, the regular interval being, for example, every 30 minutes, every hour, every 4 hours, every 6 hours, every 12 hours, every day, every 2 days, every week, or any other suitable interval.

[0110] Continuous or semi-continuous repetition of the methods described herein may facilitate real-time monitoring of the operational performance of the bioreactor, thereby providing the user or controller of the bioreactor with detailed information that can be used to improve the efficiency and / or stability of the operation of the bioreactor.

[0111] The machine learning model may be trained by carrying out an initial training of the machine learning model using a training dataset; and calibrating the machine learning model using sensor data and calibration data. The sensor data may be collected from the bioreactor and relate to the operation of the bioreactor. The calibration data may be derived from laboratory tests carried out on one or more samples collected from the bioreactor.

[0112] For example, the machine learning model can be trained by initially training it with a training dataset and then calibrating it using newly acquired sensor and calibration data. This calibration data may have been derived from a new operational period of the bioreactor, during which both sensor readings and sample data are collected. Comparing different training sets allows for statistical tests, such as the Kolmogorov-Smirnov test, to be conducted. These tests may be useable to assess the suitability of the new training data for model refinement.8911927

[0113] 14

[0114] The machine learning models described herein may be trained using supervised learning techniques. That is the machine learning model may be a supervised learning model - e.g., a supervised learning ANN.

[0115] Training the machine learning model may be carried out before the machine learning model is provided to the bioreactor and may be trained in a manner agnostic of the specific bioreactor in which it may be used. For example, the machine learning model may be trained remote from the bioreactor (e.g., communicatively disconnected in relation to the bioreactor) so that the machine learning model is trained to a predetermined degree of accuracy and / or reliability before being implemented in the context of the bioreactor.

[0116] Calibrating the machine learning model may be carried out during its implementation in the context of the (specific) bioreactor. That is, the calibration of the machine learning model may be carried out in situ. Calibrating the machine learning model (e.g., retraining the machine learning model in situ) may ensure that the machine learning model is accurately and / or reliably calibrated to the specific environmental conditions of the specific bioreactor whose operational performance it is implemented to assess. In this way, for different bioreactors, different machine learning models may be deployed, each of which is specifically trained and calibrated to optimise the accuracy and / or reliability of the machine learning model for assessing the operational performance of the corresponding bioreactor.

[0117] The sensor data may include one or more of: calibration pH data indicative of a pH inside the chamber during operation; calibration contaminant data indicative of an amount of contaminant microorganism inside the chamber during operation; calibration production data indicative of an amount of biogas produced by the bioreactor during operation; calibration loading data indicative of an organic load rate with which organic material is loaded into the bioreactor for processing into biogas; calibration composition data indicative of a composition of the biogas produced by the bioreactor during operation; and / or calibration oxygen demand data indicative of an amount of chemical oxygen demand of one or more sections of the bioreactor.

[0118] Each of the calibration pH data, calibration contaminant data, calibration production data, calibration loading data, calibration composition data, and calibration oxygen demand data may be in the same form, and correspond to the same environmental / operational parameters / conditions as the pH data, contaminant data, production data, composition data, and oxygen demand data discussed above in relation to the collected data.

[0119] It is beneficial for the sensor data used to calibrate the machine learning model to correspond with the collected data used to determine the operational performance of the bioreactor so that the calibration is relevant to the implementation of the machine learning model.

[0120] The calibration data may be externally tested and validated data that provides information about the environmental conditions within the chamber of the bioreactor. For example, the calibration data may include data derived from the results of laboratory tests to determine e.g., the pH of the solution within the chamber, the FOS / TAC ratio within the chamber, the temperature of the chamber, the production output8911927

[0121] 15

[0122] of the bioreactor and / or any other suitable parameter (such as any one or more of the parameters discussed above in relation to the sensor data).

[0123] In this way, the calibration data may be used to benchmark or validate the output of the machine learning model when the machine learning model uses the sensor data as its input. This ensures that the machine learning model is specifically calibrated to assess the operational performance of the specific bioreactor for which the machine learning model is implement.

[0124] Calibrating the machine learning model may be carried out in 180 days or less.

[0125] In the context of bioreactor operation, the lifetime of a bioreactor may be several years - even decades. Accordingly, providing accurate and reliable calibration of the machine learning model within the first days, weeks, and months of the bioreactor’s lifetime may be beneficial for optimising the lifetime efficiency and / or stability of the bioreactor.

[0126] In some cases, calibrating the machine learning model may be carried out in 150 days or less, 120 days or less, 60 days or less, or even 30 days or less.

[0127] The training dataset may include one or more of: laboratory data derived from laboratory tests carried out on a plurality of samples collected from one or more representative bioreactors; sensor data collected from one or more representative bioreactors, and relating to the operation of said one or more representative bioreactors; and / or prior data used in a method for assessing an operational performance of a prior bioreactor, said method being as described herein. The machine learning model of said method may be a prior machine learning model configured to output the prior data to facilitate transfer learning for one or more other machine learning models.

[0128] The laboratory data and the sensor data of the training dataset may take the same form as the sensor data and calibration data discussed above in relation to the calibrating of the machine learning model. The one or more representative bioreactors may include actual or simulated bioreactors that have similar properties to the bioreactor for which the machine learning model is to be trained to assess the operational performance. In the case of a simulated bioreactor, the laboratory data and sensor data may be simulated data obtained by performing suitable simulations of the operational performance and / or environmental conditions associated with said simulated bioreactor.

[0129] For example, the laboratory data may be externally tested and validated data that provides information about the environmental conditions within the chamber of the one or more representative bioreactors. For example, the laboratory data may include data derived from the results of laboratory tests to determine e.g., the pH of the solution within each respective chamber of one of the one or more representative bioreactors, the FOS / TAC ratio within each chamber, the temperature of each chamber, the production output of each of the one or more representative bioreactors and / or any other suitable parameter (such as any one or more of the parameters discussed below in relation to the sensor data of the training dataset).8911927

[0130] 16

[0131] The sensor data of the training dataset may include one or more of: training pH data indicative of a pH inside each chamber during operation of the one or more representative bioreactors; training contaminant data indicative of an amount of contaminant microorganism inside each said chamber during operation; training production data indicative of an amount of biogas produced by each of the one or more representative bioreactors during operation; training loading data indicative of an organic load rate with which organic material is loaded into each said representative bioreactor for processing into biogas; training composition data indicative of a composition of the biogas produced by each said representative bioreactor during operation; and / or calibration oxygen demand data indicative of an amount of chemical oxygen demand of one or more sections of each said representative bioreactor.

[0132] Each of the training pH data, training contaminant data, training production data, training loading data, training composition data, and training oxygen demand data may be in the same form, and correspond to the same environmental / operational parameters / conditions as the pH data, contaminant data, production data, composition data, and oxygen demand data discussed above in relation to the collected data. It is beneficial for the sensor data used to train the machine learning model to correspond with the collected data used to determine the operational performance of the bioreactor so that the training is relevant to the implementation of the machine learning model.

[0133] The prior data may be data obtained from one or more prior machine learning models used to assess the performance of corresponding bioreactors that are themselves similar to the bioreactor for which the machine learning model is being trained to assess. For example, each of the one or more prior machine learning models may be useable to carry out the methods described herein to assess the operational performance of a respective (optionally similar) bioreactor.

[0134] Implementing transfer learning techniques may reduce the time needed to train the machine learning model as the machine learning model is able to benefit from the training and learning of other similar prior machine learning models.

[0135] The machine learning model may, for example, be stored on a device or server that is communicatively connected to other similar devices and / or servers storing the prior machine learning models. In this way, a network of connected devices and / or servers may be provided that are able to exchange data from their respective machine learning models to build a transfer learning network to improve the training and / or calibration or retraining of a plurality of machine learning models, each of which is implemented to assess the operational performance of a corresponding (optionally similar) bioreactor.

[0136] The methods described herein may further comprise compiling, using the collected data and the determined operational performance, a transfer training dataset for use in transfer learning for one or more other machine learning models.

[0137] The transfer training dataset compiled by the machine learning model may take the same form as the prior data discussed above in relation to the training of the machine learning model.8911927

[0138] 17

[0139] In this way, the machine learning model may both benefit from transfer learning derived from one or more prior machine learning models, and provide the benefit of transfer learning to one or more latter, or successor, machine learning models that are implemented on corresponding similar bioreactors.

[0140] The bioreactor may be an electromethanogenic reactor. The biogas may comprise methane. The bioreactor may comprise methanogens for anaerobically digesting the organic material to produce methane.

[0141] As described above, the electromethanogenic reactor may have methanogenic microorganisms contained therein for anaerobically digesting organic material such as acetate and hydrogen to produce methane.

[0142] The bioreactor may comprise an electrode system. The electrode system may comprise: a plurality of pairs of electrodes spaced apart in the chamber. One or more subsets of the plurality of pairs of electrodes may be independently addressable to apply a voltage therebetween, to thereby instigate the anaerobic digestion of the organic material inside the chamber.

[0143] For example, as described above, the bioreactor may be a microbial electrolysis cell system comprising a plurality of electrodes arranged in respective electrode pairs. Microorganisms (e.g., methanogenic microorganisms) may coat each of the plurality of electrodes such that, upon application of a voltage between adjacent electrodes of the plurality of electrodes, the microorganisms coating the plurality of electrodes are induced to anaerobically digest the organic material in their vicinity to produce the biogas. That is, the microorganisms may be electrically active and have a digestive response that is increased by the application of an electrical signal.

[0144] Each subset of the plurality of pairs of electrodes may comprise a respective plurality of electrode pairs or may comprise a single pair of electrodes. In either case, each subset may be independently addressable such that voltage may be applied to each subset independently and / or current signals may be measured from each subset independently. This may provide control over the spatial distribution of digestive activity within the chamber and may provide spatial resolution in the collected data such that the determined operational performance of the bioreactor may include information indicative of the spatial distribution of different performance levels within the chamber.

[0145] During operation of the bioreactor, the voltage applied across each pair of electrodes may be a relatively low voltage - e.g., approximately 1 V. Such voltages may be sufficient to induce or increase digestive activity in the microorganisms, without risking damaging or even killing the microorganisms (e.g., by electrical burn). The voltage may, in some cases, be an operational parameter that may be controllably adjusted, e.g., in response to the determined operational status of the bioreactor.

[0146] The collected data may comprise: current data indicative of one or more current signals collected from a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor; and / or voltage data indicative of a respective voltage applied between each pair of a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor.8911927

[0147] 18

[0148] The current data may, for example, include data indicative of a value of current and / or current density drawn through each subset of the plurality of pairs of electrodes during operation of the bioreactor. The current data may, for example, include data that aggregates the total flow of current and / or current density for the entire subset. Additionally or alternatively, the current data may, for example, include data indicative of a value of current and / or current density drawn through each individual electrode pair of the given subset.

[0149] Current data may be useful for determining the efficiency and / or stability of the digestive processes inside the chamber because, the digestion of the organic material by the microorganisms may generate electrons which are transferable from the microorganisms to a nearby electrode, thereby inducing a return current flow in the corresponding electrode. Current measurements may therefore provide information indicative of the digestive activity of the microorganisms corresponding each electrode.

[0150] By collecting the current data using the same electrodes as the electrodes used to deliver voltage to induce or increase digestive activity, the structural complexity of the bioreactor may be reduced, not least because only one electrode system need be provided.

[0151] The voltage data may, for example, include data indicative of the aggregated voltage applied to each subset of the plurality of pairs of electrodes during operation of the bioreactor. Additionally or alternatively, the voltage data may, for example, include data indicative of the voltage applied to each individual electrode pair within a corresponding subset.

[0152] Voltage data may be useful for determining the efficiency and / or stability of the digestive processes inside the chamber because, the voltage data may be useable to infer the extent to which digestive processes are (or could be) enhanced and / or driven by the application of voltage.

[0153] By providing the current data and / or voltage data for each subset of the plurality of pairs of electrodes, spatial distributions in the operational performance of the bioreactor may be determined. That is the determined operational performance of the bioreactor may account for spatial variations within the chamber (e.g., across different subsets of the plurality of electrode pairs).

[0154] Similarly, the collected current and / or voltage data (as discussed above for the collected data) may include data indicative of temporal variations in the current and / or voltage data. These variations may be useable to determine the operational stability of the bioreactor.

[0155] In cases where the collected data includes current data and / or voltage data, the sensor data of the training dataset and / or the sensor data used to calibrate the machine learning model may correspondingly including training current data and / or training voltage data, and / or calibration current data and / or calibration voltage data accordingly.

[0156] Determining the operational performance of the bioreactor may comprise dividing the bioreactor into a plurality of zones. Each zone may correspond to a respective subset of the plurality of pairs of electrodes. Determining the operational performance of the bioreactor may further comprise: determining, for each zone, a respective zonal operational performance indicative of the operational performance of the bioreactor within the corresponding zone.8911927

[0157] 19

[0158] The division of the bioreactor into zones may not be a physical division but rather may be an organisational division for the purpose of determining zonal operational performance. In this way, spatial variations in the operational performance of the bioreactor may be determined.

[0159] Determining the zonal operational performance for a given zone may comprise one or more of: determining whether an environment in the zone is suitable for the anaerobic digestion of organic material inside the chamber; determining whether a quantity of organic material supplied to the zone is suitable for producing a target quantity of biogas; and / or determining a FOS / TAC ratio associated with the environment in the zone.

[0160] Determining whether the environment in the zone is suitable for the anaerobic digestion of organic material inside the chamber may, for example, involve determining whether the pH, temperature, pressure and / or amount of contaminant microorganisms in the zone is suitable for the efficient and stable anaerobic digestion of organic material. As discussed above, whether an environment is suitable for the anaerobic digestion of the organic material may be dependent on the context in which the bioreactor is used. Accordingly, it may be beneficial to calibrate and / or validate the machine learning model for use in assessing the operational performance of the specific bioreactor against one or more predetermined operational benchmarks. For example, each environmental condition may have associated therewith a predetermined range, within which the environmental condition may be considered to be suitable for the anaerobic digestion of the organic material. If any environmental condition is determined to lie outside its predetermined range, it may be determined that the environment is not suitable for the anaerobic digestion of the organic material. In contrast, if all the environmental conditions are determined to lie within their respective predetermined ranges, it may be determined that the environment is suitable for the anaerobic digestion of the organic material.

[0161] Determining whether a quantity of organic material supplied to the zone is suitable for producing a target quantity of biogas may involve determining an operational efficiency of the zone and / or overall bioreactor and determining, based on said operational efficiency, an amount of organic material needed to produce the target quantity of biogas. If the loading data is indicative that the amount of organic material being fed to the zone is less than this needed weight, it may be determined that the amount of organic material is not suitable for producing the target quantity of biogas. In contrast, if the loading data is indicative that the amount of organic material being fed to the zone is at least this needed weight, it may be determined that the amount of organic material is suitable for producing the target quantity of biogas.

[0162] In some examples, the FOS / TAC ratio may be inferred using at least pH data for the zone and contaminant data for the zone. The FOS / TAC ratio may be determined to infer the suitability of the environment in the chamber for the anaerobic digestion of organic material inside the chamber. For example, if the determined FOS / TAC ratio is within a predetermined range, it may be determined that the environment inside the chamber is suitable for the anaerobic digestion of the organic material. The predetermined range may, for example, be between 0.3 and 0.4. The predetermined range may, for example, vary depending on the context in which the bioreactor is used to process organic material to produce biogas.8911927

[0163] 20

[0164] Deviations, variations and / or fluctuations in the FOS / TAC ratio may be indicative of an operational instability in the zone. For example, an elevated FOS / TAC ratio (i.e., above the upper limit of the predetermined range) may indicate an excessive build-up of volatile fatty acids, or inadequate solution alkalinity that may inhibit the activity (e.g., methanogenic activity) of the microorganisms in the zone. In contrast, as an example, a depressed FOS / TAC ratio (i.e., below the lower limit of the predetermined range) may indicate an insufficient organic loading rate which may reduce the biogas yield of the bioreactor.

[0165] The methods described herein may further comprise generating a representation of the plurality of zonal operational performances for analysis by an operator of the bioreactor. The method may include displaying the generated representation, for example on a user device or terminal.

[0166] The representation may, for example, be a graphical representation (e.g., a heatmap) to facilitate easy interpretation of the spatial distribution of zonal operational performances within the bioreactor. For example, a heatmap may facilitate an identification of whether a particular zonal operational performance is an isolated outlier or whether there is a trend in the zonal operational performances across the reactor. The representation of the plurality of zonal operational performances may, for example, be interactive (e.g., via a user interface) to allow the operator to controllably adjust one or more of the operating parameters of the bioreactor within a given zone in response to the determined zonal operational performance.

[0167] Determining each zonal operational performance may comprise: determining one or more operational parameters of the bioreactor within the corresponding zone; and determining, using the one or more determined operational parameters, a zonal score for the corresponding zone of the bioreactor.

[0168] Each of the zonal scores may be determined in a manner analogous to that described above for the operation score.

[0169] For example, the one or more operational parameters within a given zone may, for example, include any one or more of: an operational efficiency of the biogas production within the zone (e.g., an efficiency with which organic material is processed into biogas); an operational stability of the biogas production within the zone (e.g., a quantified indication of the temporal variation in biogas production); a concentration of contaminants, e.g., contaminant-products and / or volatile fatty acids, in the zone; one or more pH values inside the zone; one or more temperature values inside the zone; a feed rate with which organic material is fed into the zone; an output rate with which biogas is produced by the zone; one or more FOS / TAC ratio values for the zone; and / or a chemical oxygen demand of the zone.

[0170] Determining the zonal score for the bioreactor may involve, for example, determining an individualised score for each of the determined one or more operational parameters. Each individualised score may, for example, be combined (e.g., by aggregation or averaging) to determine the zonal score. Alternatively, the zonal score may be a composite score defined by the one or more individualised scores.

[0171] Combining the individualised scores to determine the zonal score may involve weighting the individualised scores based on their relative importance to the operational performance of the zone. For8911927

[0172] 21

[0173] example, pH-based scores, contaminant-based scores and / or FOS / T AC-based scores may be weighted more heavily than other individualised scores.

[0174] Each of the zonal score and, optionally, the one or more individualised scores may be a numerical (or quantitative score), or may be a qualitative score indicating whether the zone is operating in a manner suitable for the anaerobic digestion of the organic material to produce biogas. For example, the qualitative score may be expressed in natural language form to indicate whether the bioreactor is in “good” operational performance condition (or similar), or in “inadequate” or “poor” operational performance condition (or similar).

[0175] In cases where the machine learning model is trained to determine spatial variation in the operation performance of the bioreactor, the operation score may reflect this spatial variation, for example, in the form of an array or a graphical representation (e.g. , a heatmap).

[0176] In some examples, the zonal score may include a recommendation for controlling the bioreactor to improve the zonal score. For example, the recommendation may indicate whether an improvement in the zonal score is necessary, desirable and / or possible in view of the determined zonal score (and, optionally, the determined individualised scores). Additionally or alternatively, the recommendation may indicate a recommended course of action (e.g., introduce buffer solution, adjust temperature, adjust electrode voltage and / or current, etc.) to improve the zonal score.

[0177] In some examples, controlling the bioreactor may involve adjusting one or more operating parameters of the bioreactor on a zone-by-zone basis. For example, the division of the bioreactor into a plurality of zones may, in addition to facilitating a spatial resolution in the determination of the operating performance of the bioreactor, facilitate a spatial resolution in the control of the bioreactor. For example, controlling the bioreactor may involve, adjusting the one or more operating parameters of each zone (independently), in response to the determined zonal operational performance of each zone.

[0178] Adjusting one or more operating parameters of each zone may involve, for each zone, one or more of: modifying a feed rate of organic material into said zone, adjusting a concentration of one or more buffering agents inside said zone, adjusting a temperature inside zone, adjusting an amount of microorganisms inside said zone, adjusting a voltage applied across the electrode pair(s) in said zone, and / or activating or deactivating one or more of the electrode pairs in said zone.

[0179] The overall operation score may be determined by combining and / or compiling the plurality of zonal scores. For example, the plurality of zonal scores may be combined by aggregation or averaging, or any other suitable operation, to determine - at least in part - the operational score.

[0180] In another aspect, there is provided a computer comprising a memory for storing data, and a processor configured to carry out the methods described herein. The computer and / or the processor may form a part of the bioreactor being monitored.

[0181] The memory may also, for example, store the trained machine learning model.8911927

[0182] 22

[0183] In a further aspect, there is provided a computer-readable medium comprising instructions that, when executed by a computer, cause the computer to carry out the methods described herein.

[0184] The computer-readable medium may, for example, be stored in a memory of the computer, or may be stored remotely (e.g. , on a server accessible to the computer).

[0185] In the described embodiments of the invention, the system may be implemented as any form of a computing and / or electronic device. Such a device may comprise one or more processors which may be microprocessors, controllers or any other suitable type of processors for processing computer executable instructions to control the operation of the device in order to gather and record routing information. In some examples, for example where a system on a chip architecture is used, the processors may include one or more fixed function blocks (also referred to as accelerators) which implement a part of the method in hardware (rather than software or firmware). Platform software comprising an operating system or any other suitable platform software may be provided at the computing-based device to enable application software to be executed on the device.

[0186] Moreover, the acts described herein may be embodied using computer-executable instructions that can be implemented by one or more processors and / or stored on a computer-readable medium or media. The computer-executable instructions can include routines, sub-routines; programs; threads of execution, and / or the like. Still further, results of acts of the methods can be stored in a computer-readable medium, displayed on a display device, and / or the like.

[0187] The order of the operations of the methods described herein is exemplary, but the steps may be carried out in any suitable order, or simultaneously where appropriate. Additionally, steps may be added or substituted in, or individual steps may be deleted from any of the methods without departing from the scope of the subject matter described herein. Aspects of any of the examples described above may be combined with aspects of any of the other examples described to form further examples without losing the effect sought.

[0188] Various functions described herein can be implemented in hardware, software, or any combination thereof. If implemented in software, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media may include, for example, computer-readable storage media. Computer-readable storage media may include volatile or non-volatile, removable or non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. A computer-readable storage media can be any available storage media that may be accessed by a computer. By way of example, and not limitation, such computer-readable storage media may comprise RAM, ROM, EEPROM, flash memory or other memory devices, CD-ROM or other optical disc storage, magnetic disc storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer.8911927

[0189] 23

[0190] Although illustrated as a local device it will be appreciated that the computing device may be located remotely and accessed via a network or other communication link (for example using a communication interface).

[0191] The term 'computer' is used herein to refer to any device with processing capability such that it can execute instructions. Those skilled in the art will realise that such processing capabilities are incorporated into many different devices and therefore the term 'computer' includes PCs, servers, mobile telephones, personal digital assistants and many other devices.

[0192] Those skilled in the art will realise that storage devices utilised to store program instructions can be distributed across a network. For example, a remote computer may store an example of the process described as software. A local or terminal computer may access the remote computer and download a part or all of the software to run the program. Alternatively, the local computer may download pieces of the software as needed, or execute some software instructions at the local terminal and some at the remote computer (or computer network). Those skilled in the art will also realise that by utilising conventional techniques known to those skilled in the art that all, or a portion of the software instructions may be carried out by a dedicated circuit, such as a DSP, programmable logic array, or the like.

[0193] It will be understood that the benefits and advantages described above may relate to one embodiment or may relate to several embodiments. The embodiments are not limited to those that solve any or all of the stated problems orthose that have any or all of the stated benefits and advantages. Variants should be considered to be included into the scope of the invention.

[0194] The invention includes the combination of the aspects and preferred features described except where such a combination is clearly impermissible or expressly avoided.

[0195] Summary of the Figures

[0196] Embodiments and experiments illustrating the principles of the invention will now be discussed with reference to the accompanying figures in which:

[0197] Figure 1 shows a schematic illustration of a bioreactor.

[0198] Figure 2 shows a schematic illustration of a chamber for a bioreactor having an electrode system installed therein.

[0199] Figure 3 shows a schematic illustration of a chamber for a bioreactor having an electrode system installed therein, wherein subsets of electrodes of the electrode system are independently addressable.

[0200] Figure 4 is a graph showing the current response over time of different subsets of an electrode system.

[0201] Figure 5A schematically illustrates an exemplary machine learning model suitable for assessing the operational performance of a bioreactor.

[0202] Figure 5B schematically illustrates an exemplary machine learning model suitable for assessing the operational performance of a bioreactor, the machine learning model employing a plurality of sub-models.

[0203] Figure 6 shows a method of assessing the operational performance of a bioreactor.8911927

[0204] 24

[0205] Figure 7 shows the results of a trained machine learning model trained to predict a FOS / TAC value for a bioreactor comparing the predicted FOS / TAC values with true, measured, FOS / TAC values. A. True values vs. predictions from an ANN over out-of-fold ensemble evaluation. B. True values vs. predictions ordered over time.

[0206] Detailed Description of the Invention

[0207] Aspects and embodiments of the present invention will now be discussed with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art. Figure 1 shows a schematic illustration of a bioreactor 100. The bioreactor 100 comprises an equalisation tank 110 (or holding tank) for storing homogenised influent organic material. For example, in the context of a wastewater treatment system, the equalisation tank 110 stores homogenised wastewater. Organic material stored into the equalisation tank 110 is fed, via feeding inlet 112, into a chamber 120 of the bioreactor 110. The chamber 120 has contained therein a microorganism population for anaerobically digesting organic material received from the equalisation tank 110 to produce biogas. For example, the chamber may contain a methanogen population for anaerobically digesting organic material such as hydrogen and acetate to produce methane. The produced biogas is expelled from the chamber 120 via a gas outlet 122.

[0208] As part of the operation of the bioreactor 100, not all of the organic material may be fully digested in a single feed event. In such cases, undigested organic material is fed, via chamber outlet 124 into an effluent holding tank 130. Solids in the organic material may settle in the effluent tank 130 and be fed back into the chamber 120 for anaerobic digestion via a recirculation inlet 132, while excess organic material (e.g., excess liquid) may be expelled from the effluent holding tank 130 via an effluent outlet 134. This may prevent the bioreactor 100 from becoming overloaded with organic material.

[0209] The bioreactor 100 further comprises one or more sensors 140 for measuring one or more parameters associated with the operation and / or environmental conditions of the bioreactor 100. As discussed herein, the one or more sensors 140 may be provided as components of, or connected to, the chamber 120 (as shown in Figure 1. Additionally or alternatively, one or more of the one or more sensors 140 may be provided as components of, or connected to, any of the equalisation tank 110, the feeding inlet 112, the gas outlet 122, the feeding outlet 124, the effluent holding tank 130, the recirculation inlet 132, and / or the effluent outlet 134.

[0210] Any of the one or more sensors 140 may, for example, be configured to measure any suitable parameters - e.g., pH, temperature, chemical oxygen demand, loading rate, contaminant concentration (e.g., H2S concentration), produced biogas volume, biogas composition, and / or pressure, or any other suitable parameter useable to assess the operational performance of the bioreactor.

[0211] Figure 2 shows a schematic illustration of a chamber 120 of a bioreactor 100 having an electrode system 200 installed therein.8911927

[0212] 25

[0213] For illustrative simplicity, the chamber 120 is depicted with a single inlet 112, 132, which may include one or both of the feeding inlet 112 and the recirculation inlet 132.

[0214] The electrode system 200 comprises a plurality of pairs of electrodes arranged into subsets 202, 204, 206, 208. Each of the subsets 202, 204, 206, 208 of electrode pairs is independently addressable via a respective zonal control unit 212, 214, 216 ,218.

[0215] Each zonal control unit 212, 214, 216, 218 is configured to deliver a signal to the corresponding subset of electrode pairs 202, 204, 206, 208 to cause a voltage (e.g. , a 1 V voltage) to be applied across each pair of electrodes within the subset.

[0216] The chamber 120 depicted in Figure 2 may be an electromethanogenic reactor (EMR) chamber. In such cases, each electrode of the electrode system 200 is coated with an electroactive methanogen population. Under the application of a voltage across each pair of electrodes, the anaerobic digestive activity of the electroactive methanogens is instigated and / or enhanced to promote the anaerobic digestion of organic material in the chamber 120 to produce a biogas comprising methane. This arrangement therefore provides an MEC-AD system.

[0217] In addition to delivering a signal to apply voltage across each pair of electrodes within a given subset 202, 204, 206, 208, each zonal control unit 212, 214, 216, 218 is further configured to receive sensor data from a zone defined by the corresponding subset of electrode pairs 202, 204, 206, 208. The received sensor data defines collected data that is useable, by a trained machine leaning model, to determine the zonal operational performance for the corresponding zone.

[0218] The sensor data may include data indicative of any of the parameters set out in T able 1 below. This data has been found to be suitable for building and implementing a machine learning model that can accurately and reliably assess the operational performance of a bioreactor 100. The sensor data received by each of the zonal control units 212, 214, 216, 218 and (in some cases) the central control unit 220 are received from one or more sensors 140 discussed above in relation to Figure 1.

[0219]

[0220] 8911927

[0221] 26

[0222]

[0223] Table 1: List of exemplary parameters useable by a machine learning model to determine the operational performance of a bioreactor

[0224] Each of the zonal units 212, 214, 216, 218 may have stored thereon instructions that cause the zonal unit to carry out part or all of the methods described herein.

[0225] Additionally, collected data and / or zonal operational performance may be communicated from the chamber 120 and / or each of the zonal units 212, 214, 216, 218 to a central control unit 220. The control unit 220 is configured to use a trained machine learning model to determine, using the collected data, and / or the determined zonal operational performances, an operational performance of the bioreactor. Determining each of the zonal operational performances may involve determining a zonal score that characterises the operational performance of the corresponding zone. Similarly, determining the operational performance of the bioreactor may involve determining an operation score that characterises the operational performance of the bioreactor. Determining the operational performance of the bioreactor may involve aggregating, averaging, or otherwise combining the determined zonal operational performances.

[0226] Figure 3 shows a schematic illustration of a chamber for a bioreactor having an electrode system installed therein. Figure 3 shows each subset of the electrode pairs 302, 304, 306, 308 including multiple electrode pairs arranged in an array. The zonal control units 312, 314, 316, 318 are configured to each respectively and independently apply a voltage across the electrodes of the corresponding subset 302, 304, 306, 308 and to measure the current and / or current density drawn through the corresponding subset 302, 304, 306, 308.

[0227] Each of the zonal units 312, 314, 316, 318 is further configured to communicate with a central control unit 320 for determining the operational performance of the bioreactor.

[0228] Further, each of the zonal units 312, 314, 316, 318 is configured to communicate with respectively adjacent zonal units. This may facilitate the determination of spatial variations and / or trends in the determined zonal operational performances and so may facilitate a refined spatially aware control of the operational parameters of the bioreactor within each zone.

[0229] For example, as shown in Figure 4, data from different (e.g. , adjacent) zones may be used to identify trends in the operational performance of the bioreactor. Figure 4 is a graph showing the current response over time of different subsets of an electrode system.8911927

[0230] 27

[0231] Figure 4 shows a spike in the current response of a first subset of the electrode pairs at an initial time. This spike in the current response coincides with a spike in the organic loading rate from the equalisation tank 110 into the chamber 120. That is, in response to organic feed being fed into the chamber 120, the anaerobic digestive rate in a zone receiving the organic material increases, thereby causing an increase in the current response measured by the corresponding zonal unit.

[0232] At a later time, an increase in the current response of a second subset of the electrode pairs can also be seen in Figure 4. This second subset of the electrode pairs is downstream of the first subset - that is the second subset is arranged further from the feeding inlet 112 than the first subset. Accordingly, the current spike in the current response of the second subset at a time later than the initial time is indicative that the organic material is flowing through the chamber and through each of the zones. Additionally, the current spike in the current response of the second subset is indicative that at least some of the organic material fed into the chamber is not fully digested by microorganisms in upstream zones.

[0233] For example, if there were no current spike in the current response of any one of the subsets corresponding to a zone downstream of the first subset, this may indicate that the organic material is fully digested before flowing all the way through the chamber and, therefore, that the organic loading rate may be increased to increase the biogas yield.

[0234] Figure 5A shows an exemplary machine learning model useable in the methods described herein. The machine learning model of Figure 5A is an artificial neural network, ANN, comprising an input layer for receiving the collected data, a first hidden layer comprising 64 nodes, a second hidden layer comprising 128 nodes, a third hidden layer comprising 128 nodes, and an output node comprising at least one node. The ANN of Figure 5A is trained to determine, using the collected data, a prediction of the FOS / TAC ration of the bioreactor.

[0235] As discussed above, while the machine learning model of Figure 5A is depicted as an ANN, other architectures are also possible - for example, Decision Trees, XGBoost, SVMs, Support Vector Regression, ANNs, RNNs, CNNs, LSTM may all be suitable architectures.

[0236] It was determined, through testing, that an ANN architecture such as that shown in Figure 5A, when used in the context of the methods described herein, may benefit from a mean absolute error of 0.045-0.17 (with an average of 0.092), a normalised root mean square error of 0.38-1 (with an average of 0.63), and an explained variance 0.15-0.86 (with an average of 0.62).

[0237] The machine learning model was trained using supervised learning techniques and a training dataset. As discussed above, the training dataset may include one or more of: laboratory data derived from laboratory tests carried out on a plurality of samples collected from one or more representative bioreactors; sensor data collected from one or more representative bioreactors, and relating to the operation of said one or more representative bioreactors; and / or prior data used in a method for assessing an operational performance of a prior bioreactor, said method being as described herein. The machine learning model of said method may be a prior machine learning model configured to output the prior data to facilitate transfer learning for one or more other machine learning models.8911927

[0238] 28

[0239] The machine learning model is also calibrated overtime for implementation with the specific bioreactor to which it is deployed using sensor data and calibration data, as discussed above.

[0240] Figure 5B shows another example of a machine learning model useable in the methods described herein.

[0241] The machine learning model of Figure 5B comprises a clustering sub-model configured to firstly determine an operational state of the bioreactor based on data for selected input parameters (e.g. data obtained from online sensors) input into the clustering sub-model. The architecture of the clustering submodel is not illustrated in Figure 5B. Typically, the clustering sub-model comprises PCA prior to clustering (e.g. via K-Means clustering).

[0242] The machine learning model of Figure 5B further comprises a plurality of sub-models configured to determine the operational performance of the bioreactor, and each sub-model configured to determine the operational performance of the bioreactor relates to a respective operational state of the bioreactor (e.g. the start-up sub-model corresponding to the operational state of the bioreactor being that of startup). Based on the operational state of the bioreactor determined to be present by the clustering submodel, the appropriate sub-model for determining the operational performance of the bioreactor (i.e. the sub model relating to the operational state determined by the clustering model) is selected to determine the operational performance of the bioreactor (e.g. where the clustering module determines that the operational state of the bioreactor is ‘stable low load’, it is subsequently the stable low load sub-model that is used to predict the operational performance of the bioreactor.

[0243] In this way, the prediction accuracy and robustness of each sub-model for predicting the operational performance may be improved for when the operational state of the bioreactor is the state said sub-model corresponds to, compared to where there is a single model that is used for determination of the operational performance of the bioreactor across all operational states. By way of example, the training of each sub-model for predicting the operational performance may be tailored to the operational state it corresponds to, e.g. by the training data containing a larger proportion of data relating to the operational state it corresponds to versus the amount of data relating to other operational states.

[0244] As illustrated in Figure 5B, data for selected input parameters are also passed to the sub-models configured to determine the operational performance of the bioreactor, along with the predicted operational state from the clustering sub-model. The selected input parameters inputted into the clustering sub-model may be the same as, or different to, those inputted into the sub-models configured to determine the operational performance of the bioreactor (e.g. some, but not all, of the input parameters may be common between the two sets of input parameters). The sub-models configured to determine the operational performance of the bioreactor may be trained in the same manner as described above for the model of Figure 5A.

[0245] Figure 6 shows a method of assessing the operational performance of a bioreactor.8911927

[0246] 29

[0247] In an operation 602, the method comprises collecting data related to the bioreactor operation. The collected data may be associated with the overall bioreactor, or may be specific to a particular zone from amongst the plurality of zones of a bioreactor, e.g., an EMR.

[0248] As discussed above, the collected data may include one or more of include one or more of: pH data indicative of a pH inside the chamber during operation; contaminant data indicative of an amount of contaminant microorganism inside the chamber during operation; production data indicative of an amount of biogas produced by the bioreactor during operation; loading data indicative of an organic load rate with which organic material is loaded into the bioreactor for processing into biogas; composition data indicative of a composition of the biogas produced by the bioreactor during operation; oxygen demand data indicative of an amount of chemical oxygen demand of one or more sections of the bioreactor; current data indicative of one or more current signals collected from a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor; and / or voltage data indicative of a respective voltage applied between each pair of a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor.

[0249] The method further comprises, in an operation 604, using the collected data as input to a machine learning model. The collected data may, for example, be compiled and / or pre-processed into a format suitable for input into the machine learning model.

[0250] The method may further comprise, in an operation 606, determining, by the machine learning model, operational parameters of each zone of the bioreactor using the collected data associated with each respective zone.

[0251] The one or more operational parameters within a given zone may, for example, include any one or more of: an operational efficiency of the biogas production within the zone (e.g., an efficiency with which organic material is processed into biogas); an operational stability of the biogas production within the zone (e.g., a quantified indication of the temporal variation in biogas production); a concentration of contaminants, e.g., contaminant-products and / or volatile fatty acids, in the zone; one or more pH values inside the zone; one or more temperature values inside the zone; a feed rate with which organic material is fed into the zone; an output rate with which biogas is produced by the zone; one or more FOS / TAC ratio values for the zone; and / or a chemical oxygen demand of the zone.

[0252] The method further comprises, in an operation 608, determining, by the machine learning model, operational parameters of the bioreactor using the collected data and, optionally, the determined operational parameters of each zone.

[0253] The one or more operational parameters may, for example, include any one or more of: an operational efficiency of the biogas production (e.g., an efficiency with which organic material is processed into biogas); an operational stability of the biogas production (e.g., a quantified indication of the temporal variation in biogas production); a concentration of contaminants, e.g., contaminant-products and / or volatile fatty acids, in the chamber of the bioreactor; one or more pH values inside the chamber of the bioreactor (e.g., a pH value for each of a plurality of zones of the chamber); one or more temperature8911927

[0254] 30

[0255] values inside the chamber (e.g., a temperature value for each of a plurality of zones of the chamber); a feed rate with which organic material is fed into the chamber; an output rate with which biogas is produced by the bioreactor; one or more FOS / TAC ratio values (e.g., a FOS / TAC ratio value for each of a plurality of zones of the chamber); and / or a chemical oxygen demand of the chamber and / or of an equalisation tank of the bioreactor.

[0256] The method may further comprise, in an operation 610, determining, for each zone of the bioreactor, a zonal score representative of the operational performance of that zone.

[0257] As discussed above, the zonal score may be indicative of the zonal operational performance for a given zone. Determining the zonal score for the bioreactor may involve, for example, determining an individualised score for each of the determined one or more operational parameters. Each individualised score may, for example, be combined (e.g., by aggregation or averaging) to determine the zonal score. Alternatively, the zonal score may be a composite score defined by the one or more individualised scores. Combining the individualised scores to determine the zonal score may involve weighting the individualised scores based on their relative importance to the operational performance of the zone. For example, pH-based scores, contaminant-based scores and / or FOS / T AC-based scores may be weighted more heavily than other individualised scores.

[0258] The zonal score may be useable to determine a zonal operational performance for a given zone. As discussed above, determining the zonal operational performance for a given zone may comprise one or more of: determining whether an environment in the zone is suitable for the anaerobic digestion of organic material inside the chamber; determining whether a quantity of organic material supplied to the zone is suitable for producing a target quantity of biogas; and / or determining a FOS / TAC ratio associated with the environment in the zone.

[0259] The method may further comprise, in an operation 612, determining an operation score for the bioreactor representative of the operational performance of that zone.

[0260] As discussed above, determining the operation score for the bioreactor may involve, for example, determining an individualised score for each of the determined one or more operational parameters. Each individualised score may, for example, be combined (e.g., by aggregation or averaging) to determine the operation score. Alternatively, the operation score may be a composite score defined by the one or more individualised scores.

[0261] Combining the individualised scores to determine the operation score may involve weighting the individualised scores based on their relative importance to the operational performance of the bioreactor. For example, pH-based scores, contaminant-based scores and / or FOS / T AC-based scores may be weighted more heavily than other individualised scores.

[0262] Each of the operation score and, optionally, the one or more individualised scores may be a numerical (or quantitative score), or may be a qualitative score indicating whether the bioreactor is operating in a manner suitable for the anaerobic digestion of the organic material to produce biogas. For example, the qualitative score may be expressed in natural language form to indicate whether the bioreactor is in8911927

[0263] 31

[0264] “good” operational performance condition (or similar), or in “inadequate” or “poor” operational performance condition (or similar).

[0265] Additionally or alternatively, the operation score may be determined by combining and / or compiling the plurality of zonal scores. For example, the plurality of zonal scores may be combined by aggregation or averaging, or any other suitable operation, to determine - at least in part - the operational score.

[0266] The operation score may be useable to determine an operational performance of the bioreactor. As discussed above, determining the operational performance of the bioreactor may comprise one or more of: determining whether an environment in the chamber is suitable for the anaerobic digestion of organic material inside the chamber; determining whether a quantity of organic material supplied to the chamber is suitable for producing a target quantity of biogas; and / or determining a FOS / TAC ratio associated with the environment in the chamber.

[0267] The operation score and any of the zonal scores may, in some cases, include a recommendation for controlling the bioreactor to improve said score(s), as discussed above.

[0268] The method may further comprise, in an operation 614, controlling the bioreactor to adjust one or more operating parameters of the bioreactor in response to the determined operational performance of the bioreactor and / or one or more zones of the bioreactor.

[0269] As discussed above, controlling the bioreactor may comprise one or more of: modifying a feed rate of organic material into the chamber; adjusting a concentration of one or more buffering agents inside the chamber; adjusting a temperature inside the chamber; adjusting an amount of microorganisms inside the chamber; and / or activating or deactivating the bioreactor so as to achieve a target operational performance for the bioreactor and / or one or more zones of the bioreactor.

[0270] The method may further comprise, in an operation 616, compiling a transfer learning dataset from the collected data and the outputted determinations of the machine learning model. The transfer learning dataset may be useable to train latter, or successor, machine learning models to assess the operational performance of other, similar, bioreactors. The transfer learning dataset may be communicated to an external device, e.g., a device configured to train machine learning models, a device configured to use a successor machine learning model to assess the operational performance of another, similar, bioreactor, and / or a repository fortraining datasets.

[0271] The methods described herein may be used to accurately and reliably assess the operational performance of a bioreactor, for example by accurately predicting the FOS / TAC ratio of a bioreactor. Figure 7 shows the results of a trained machine learning model (such as the ANN of Figure 5A) trained to predict a FOS / TAC value for a bioreactor comparing the predicted FOS / TAC values with true, measured, FOS / TAC values. Figure 7A plots the true values vs. predictions from an ANN over out-of-fold ensemble evaluation, while Figure 7B plots the same values ordered over time to demonstrate the improvement in the performance of the ANN as it is calibrated overtime. As can be seen from Figure 7, the positive correlation between the true and predicted values demonstrates the suitability of machine8911927

[0272] 32

[0273] learning models (e.g., models having an ANN architecture as described herein) for assessing the operational performance of the bioreactor.

[0274] The features disclosed in the foregoing description, or in the following claims, or in the accompanying drawings, expressed in their specific forms or in terms of a means for performing the disclosed function, or a method or process for obtaining the disclosed results, as appropriate, may, separately, or in any combination of such features, be utilised for realising the invention in diverse forms thereof.

[0275] While the invention has been described in conjunction with the exemplary embodiments described above, many equivalent modifications and variations will be apparent to those skilled in the art when given this disclosure. Accordingly, the exemplary embodiments of the invention set forth above are considered to be illustrative and not limiting. Various changes to the described embodiments may be made without departing from the spirit and scope of the invention.

[0276] For the avoidance of any doubt, any theoretical explanations provided herein are provided for the purposes of improving the understanding of a reader. The inventors do not wish to be bound by any of these theoretical explanations.

[0277] Any section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

[0278] The terms “a” (or “an”), as well as the terms “one or more” and “at least one” can be used interchangeably herein.

[0279] The term “and / or” as used herein is to be taken as specific disclosure of each of specified listed features or components with or without one or more of the others. Thus, the term “and / or” as used in a phrase such as “A, B and / or C” encompasses each of: A and B and C; A and B; A and C; B and C; A or B or C; A or C; A or C; B or C; only A; only B; and only C.

[0280] The use of the term “comprise” and “include” to refer to the inclusion of integers, steps and / or operations nonetheless also encompasses aspects, examples and embodiments that may be analogously described with the term “consist” in respect of those integers, steps and / or operations.

[0281] Throughout this specification, including the claims which follow, unless the context requires otherwise, the word “comprise” and “include”, and variations such as “comprises”, “comprising”, and “including” will be understood to imply the inclusion of a stated integer or step or group of integers or steps but not the exclusion of any other integer or step or group of integers or steps.

[0282] It must be noted that, as used in the specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from “about” one particular value, and / or to “about” another particular value. When such a range is expressed, another embodiment includes from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by the use of the antecedent “about,” it8911927

[0283] 33

[0284] will be understood that the particular value forms another embodiment. The term “about” in relation to a numerical value is optional and means for example + / - 10%.

Claims

891192734Claims:

1. A computer-implemented method for determining an operational performance of a bioreactor for the production of biogas by the anaerobic digestion of organic material inside a chamber of the bioreactor, the method comprising:collecting data related to an operation of the bioreactor;using the collected data as an input to a machine learning model, the machine learning model having been trained to determine an operational performance of a bioreactor using data related to an operation of said bioreactor; anddetermining, using the machine learning model, the operational performance of the bioreactor.

2. The computer-implemented method according to claim 1, wherein the method is carried out in realtime during operation of the bioreactor.

3. The computer-implemented method according to claim 1 or 2, wherein the collected data includes one or more of:pH data indicative of a pH inside the chamber during operation;contaminant data indicative of an amount of contaminant microorganism inside the chamber during operation;production data indicative of an amount of biogas produced by the bioreactor during operation;loading data indicative of an organic load rate with which organic material is loaded into the bioreactor for processing into biogas;composition data indicative of a composition of the biogas produced by the bioreactor during operation; and / oroxygen demand data indicative of an amount of chemical oxygen demand of one or more sections of the bioreactor.

4. The computer-implemented method according to any preceding claim, wherein determining the operational performance of the bioreactor comprises one or more of:determining whether an environment in the chamber is suitable for the anaerobic digestion of organic material inside the chamber;determining whether a quantity of organic material supplied to the chamber is suitable for producing a target quantity of biogas; and / ordetermining a FOS / TAC ratio associated with the environment in the chamber.

5. The computer-implemented method according to any preceding claim, wherein determining the operational performance of the bioreactor comprises:determining one or more operational parameters of the bioreactor; anddetermining, using the one or more determined operational parameters, an operation score for the bioreactor.8911927356. The computer-implemented method according to any preceding claim, further comprising:controlling the bioreactor to adjust one or more operating parameters of the bioreactor in response to the determined operational performance.

7. The computer-implemented method according to claim 6, wherein controlling the bioreactor comprises one or more of:modifying a feed rate of organic material into the chamber;adjusting a concentration of one or more buffering agents inside the chamber; adjusting a temperature inside the chamber;adjusting an amount of microorganisms inside the chamber; and / oractivating or deactivating the bioreactor;adjusting an amount of mixing inside the chamber;adjusting an amount of material recirculated into the chamber from an effluent holding tank of the bioreactor;adjusting a concentration of one or more additives for promoting microbial growth in the microorganisms inside the chamber; and / oradjusting a concentration of one or more additives for supporting organic degradation of the organic material inside the chamber.

8. The computer-implemented method according to any preceding claim, wherein the method is repeated continuously or semi-continuously so as to monitor the operational performance of the bioreactor.

9. The computer-implemented method according to any preceding claim, wherein the machine learning model is trained by:carrying out an initial training of the machine learning model using a training dataset; and calibrating the machine learning model using sensor data and calibration data, wherein the sensor data is collected from the bioreactor and relates to the operation of the bioreactor, and wherein the calibration data is derived from laboratory tests carried out on one or more samples collected from the bioreactor.

10. The computer-implemented method according to any preceding claim, wherein the calibrating the machine learning model is carried out in 180 days or less.

11. The computer-implemented method according to claim 9 or 10, wherein the training data set includes one or more of:laboratory data derived from laboratory tests carried out on a plurality of samples collected from one or more representative bioreactors;sensor data collected from one or more representative bioreactors, and relating to the operation of said one or more representative bioreactors; and / or891192736prior data used in a method for assessing an operational performance of a prior bioreactor, said method being the method of any preceding claim, wherein the machine learning model of said method is a prior machine learning model configured to output the prior data to facilitate transfer learning for one or more other machine learning models.

12. The computer-implemented method according to any preceding claim, further comprising:compiling, using the collected data and the determined operational performance, a transfer training dataset for use in transfer learning for one or more other machine learning models.

13. The computer-implemented method according to any preceding claim, wherein the bioreactor is an electromethanogenic reactor, the biogas comprises methane, and the bioreactor contains methanogens for anaerobically digesting the organic material to produce methane.

14. The computer-implemented method according to any preceding claim, wherein the bioreactor comprises an electrode system, the electrode system comprising:a plurality of pairs of electrodes spaced apart in the chamber, wherein one or more subsets of the plurality of pairs electrodes are independently addressable to apply a voltage therebetween, to thereby instigate the anaerobic digestion of the organic material inside the chamber.

15. The computer-implemented method according to claim 13, wherein the collected data comprises:current data indicative of one or more current signals collected from a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor; and / orvoltage data indicative of a respective voltage applied between a each pair of a corresponding subset of the plurality of pairs of electrodes during operation of the bioreactor.

16. The computer-implemented method according to claim 14 or 15, wherein determining the operational performance of the bioreactor comprises:dividing the bioreactor into a plurality of zones, each zone corresponding to a respective subset of the plurality of pairs of electrodes; anddetermining, for each zone, a respective zonal operational performance indicative of the operational performance of the bioreactor within the corresponding zone.

17. The computer-implemented method according to claim 16, further comprising:generating a representation of the plurality of zonal operational performances for analysis by an operator of the bioreactor.

18. The computer-implemented method according to claim 16 or 17, wherein determining each zonal operational performance comprises:determining one or more operational parameters of the bioreactor within the corresponding zone; and891192737determining, using the one or more determined operational parameters, a zonal score for the corresponding zone of the bioreactor.

19. The computer-implemented method according to any preceding claim, wherein:the machine learning model comprises a plurality of sub-models configured to determine the operational performance of the bioreactor, each of said sub-models relating to a respective operational state of the bioreactor; andthe machine learning model is configured to select which of the sub-models to use to determine the operational performance of the bioreactor based on the operational state of the bioreactor that the collected data corresponds to.

20. The computer-implemented method according to claim 19, wherein the machine learning model comprises a clustering sub-model configured to:determine the operational state of the bioreactor based on the collected data; and select which of the sub-models configured to determine the operational performance of the bioreactor to use to determine the operational performance of the bioreactor based on the determined operational state21. A computer comprising a memory for storing data, and a processor configured to carry out the method of any preceding claim.

22. A computer-readable medium comprising instructions that, when executed by a computer, cause the computer to carry out the method of any of claims 1 to 20.

Images