Computer program, information processing device, and information processing method

Abstract

Provided are a computer program and the like capable of appropriately estimating information related to a microbial agent used in a wastewater treatment system.　This computer program causes a computer to execute processing for: acquiring sludge image information related to a sludge image from a wastewater treatment system which uses an activated sludge process; inputting the acquired sludge image information into a model that outputs microbial agent information related to a microbial agent used in a wastewater treatment system when sludge image information of the wastewater treatment system is input into the same; and outputting the microbial agent information.

Description

Computer program, information processing device, and information processing method

[0001] The present invention relates to a computer program, an information processing device, and an information processing method.

[0002] Wastewater treatment systems using the activated sludge method are widely used to treat wastewater discharged from food factories, chemical factories, etc., and public sewage. In wastewater treatment systems using the activated sludge method, wastewater to be treated is introduced into a treatment tank, and an aeration process is performed to supply oxygen to various types of aerobic microorganisms (activated sludge) present in the treatment tank. Organic pollutants contained in the wastewater in the treatment tank are decomposed by the action of the aerobic microorganisms, thereby purifying the wastewater. Techniques for managing the operation of such wastewater treatment systems using the activated sludge method have been disclosed.

[0003] For example, Patent Document 1 discloses a wastewater treatment system that can maintain the quality of treated water in a good state by making the aeration amount by the aeration means the target aeration amount when the target aeration amount calculated based on a treatment index value measured by a sensor exceeds the minimum aeration amount of the aeration means, and maintaining the aeration amount by the aeration means at the target aeration amount when the target aeration amount is equal to or less than the minimum aeration amount, and supplying raw water to an aerobic treatment tank by a water conveying means.

[0004] Japanese Patent Application Laid-Open No. 2018-103113

[0005] However, the wastewater treatment system described in Patent Document 1 has a problem in that it is not possible to estimate information about the microbial agent used in the wastewater treatment system. In wastewater treatment systems using the activated sludge method, a microbial agent containing microorganisms is added to improve the quality of the wastewater. Since the quality and volume of wastewater vary significantly depending on the season and day, it is necessary to add a microbial agent according to the state of the wastewater. Conventionally, the addition conditions of the microbial agent have been adjusted based on the operator's empirical rules, which does not necessarily achieve appropriate control.

[0006] An object of the present disclosure is to provide a computer program or the like that can appropriately estimate information regarding a microbial agent used in a wastewater treatment system.

[0007] A computer program according to one aspect of the present disclosure acquires sludge image information relating to sludge images of a wastewater treatment system using an activated sludge method, and causes a computer to execute a process of inputting the acquired sludge image information into a model that outputs microbial agent information relating to a microbial agent to be used in the wastewater treatment system when the sludge image information of the wastewater treatment system is input, and outputting the microbial agent information.

[0008] An information processing method according to one aspect of the present disclosure deploys a model on a computer that controls a wastewater treatment system, which model outputs microbial agent information related to a microbial agent used in the wastewater treatment system when sludge image information of the wastewater treatment system using an activated sludge method is input, acquires sludge image information related to sludge images of the wastewater treatment system, and inputs the acquired sludge image information into the deployed model to output microbial agent information.

[0009] An information processing method according to one aspect of the present disclosure deploys, on each of a plurality of computers controlling each wastewater treatment system, a model that outputs microbial agent information regarding the microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system using an activated sludge method is input, collects from each of the plurality of computers parameters for each model in each wastewater treatment system that have been adjusted based on a data group including sludge image information newly obtained in each wastewater treatment system and microbial agent information regarding the microbial agent to be used in each wastewater treatment system, performs federated learning of the model based on the collected parameters of each model, and deploys the model after federated learning on each of the plurality of computers.

[0010] According to the present disclosure, information about a microbial agent used in a wastewater treatment system can be appropriately estimated.

[0011] FIG. 1 is a schematic diagram of a management system according to a first embodiment. FIG. 2 is a block diagram showing the configuration of an information processing device. FIG. 3 is a block diagram showing the configuration of a control device. FIG. 4 is an explanatory diagram showing an overview of a learning model. FIG. 5 is a flowchart showing an example of a processing procedure executed by an information processing device. FIG. 6 is an explanatory diagram showing an overview of a learning model of a second embodiment. FIG. 7 is a flowchart showing an example of a processing procedure executed by an information processing device of the second embodiment. FIG. 8 is an explanatory diagram showing an overview of a learning model of a third embodiment. FIG. 9 is a flowchart showing an example of a processing procedure related to reinforcement learning executed by an information processing device of the third embodiment. FIG. 10 is an explanatory diagram showing an overview of a learning model of a fourth embodiment. FIG. 11 is a diagram showing the configuration of a management system according to a fifth embodiment. FIG. 12 is a flowchart showing an example of a processing procedure executed by a management system according to the fifth embodiment. FIG. 13 is a flowchart showing an example of a processing procedure executed by an information processing device according to a sixth embodiment.

[0012] The present disclosure will be specifically described with reference to the drawings showing embodiments thereof.

[0013] (First embodiment) Fig. 1 is a schematic diagram of a management system 100 according to a first embodiment. The management system 100 includes an information processing device 1 as a main device. The information processing device 1 is communicably connected to each of a plurality of wastewater treatment systems 2 via a network N such as the Internet. The wastewater treatment systems 2 are systems that perform wastewater treatment using an activated sludge method. The management system 100 of this embodiment optimizes the operation management of the wastewater treatment systems 2. The number of wastewater treatment systems 2 may be one or three or more. The information processing device 1 may be integrated into any of the wastewater treatment systems 2.

[0014] The information processing device 1 is a device capable of various types of information processing and sending and receiving information, such as a server computer, a personal computer, or a quantum computer. The information processing device 1 is managed, for example, by a service provider of a management service provided by the management system 100. The information processing device 1 estimates a microbial agent suitable for use in each wastewater treatment system 2 based on sludge image information representing the state of activated sludge in the wastewater in each wastewater treatment system 2. The information processing device 1 provides information on the estimated microbial agent to the wastewater treatment system 2.

[0015] The wastewater treatment system 2 includes a wastewater treatment device 20 that treats wastewater using activated sludge, and a control device 3 that controls the operation of the wastewater treatment device 20. The activated sludge used in the wastewater treatment device 20 is a complex microbial system that includes microorganisms such as bacteria, protozoa, and metazoa. The wastewater treated by the wastewater treatment system 2 contains organic pollutants. The wastewater is, for example, industrial wastewater, public sewerage, etc.

[0016] The wastewater treatment device 20 includes a treatment tank 21 and a settling tank 22. Wastewater (water to be treated) flows into the treatment tank 21 through a water supply line 23a. A constant amount of the wastewater may be continuously supplied to the treatment tank 21.

[0017] The treatment tank 21 is equipped with an aeration means (not shown), and aerobically treats the water to be treated using activated sludge present in the treatment tank 21. The treatment tank 21 uses gases (air) such as air or oxygen supplied by a blower (not shown) to diffuse air into the treatment tank 21 and aerate the stored activated sludge. The aerobic microorganisms in the activated sludge take up dissolved oxygen from the water and oxidize and decompose the organic pollutants in the water that they have taken in as food. The water to be treated that has been treated in the treatment tank 21 is sent to the settling tank 22 through a water supply line 23b.

[0018] The settling tank 22 allows the water to be treated, delivered from the treatment tank 21, to stand for a predetermined time, allowing the activated sludge to settle naturally. This separates the water into a supernatant and activated sludge. The supernatant is discharged as treated water via a water delivery line 23c. A portion of the settled activated sludge is returned to the treatment tank 21 via a return line 23d and recycled. The remainder of the activated sludge is withdrawn from the system as excess sludge via a line 23e. The excess sludge is dehydrated, dried, etc., and then incinerated for disposal, or recycled into soil conditioner, civil engineering and construction materials, etc. Note that the settling tank 22 is not limited to a settling tank type and may be, for example, a membrane separation device.

[0019] The treatment tank 21 is provided with a camera 25 that captures images of the wastewater containing activated sludge in the tank. The camera 25 is an example of an imaging device. The camera 25 may be a high-magnification camera that can capture images of various microorganisms contained in the activated sludge at high magnification. The camera 25 generates a sludge image that shows the state of the activated sludge. The sludge image is a still image or a moving image composed of a plurality of frames of still images. The installation position of the camera 25 is not limited to the treatment tank 21, and may be, for example, the settling tank 22. Multiple cameras 25 may be provided. The camera 25 is connected to the control device 3 by wire or wirelessly, and outputs the captured sludge image to the control device 3.

[0020] The wastewater treatment device 20 may further be provided with a sensor 26 for detecting the state of the wastewater treatment device 20 or the state of the wastewater and activated sludge within the wastewater treatment device 20. Examples of the sensor 26 include a thermometer for detecting temperature, a flow meter for detecting flow rate, a pH meter for detecting pH, an ORP meter for detecting ORP (Oxidation Reduction Potential), a DO meter for detecting DO (Dissolved Oxygen) concentration, an MLSS meter for detecting MLSS (Mixed Liquor Suspended Solids: activated sludge suspended solids) concentration, a TOC meter for detecting TOC (Total Organic Carbon), an ammonia meter for measuring ammonia concentration, a nitrogen meter for measuring nitrogen concentration, and a phosphorus meter for measuring phosphorus concentration. A plurality of sensors 26 may be provided. The sensor 26 is connected to the control device 3 by wire or wirelessly and outputs detection data obtained by detection to the control device 3.

[0021] The wastewater treatment device 20 is further provided with a tank 24 for storing a microbial agent containing microorganisms as added bacteria. The microbial agent stored in the tank 24 is continuously added in predetermined amounts to the treatment tank 21, for example, using a pump (not shown). The form of the microbial agent is not limited to liquid, and may be, for example, a formulation or powder. The microbial agent may be added in the settling tank 22, the water supply lines 23a and 23b, or the return line 23d. The microbial agent may be added manually.

[0022] Examples of added bacteria contained in the microbial agent include sludge-decomposing bacteria, settling property-improving bacteria, malodorous substance-decomposing bacteria, and persistent substance-decomposing bacteria. Sludge-decomposing bacteria refer to bacteria that have the ability to decompose activated sludge that leads to excess sludge. The addition of sludge-decomposing bacteria is expected to reduce the rate of excess sludge production and suppress the generation of excess sludge (reduce excess sludge). Settling property-improving bacteria refer to bacteria that have the effect of improving or maintaining the settling property of activated sludge. The addition of settling property-improving bacteria is expected to increase the active settling property, promote solid-liquid separation in the settling tank 22, and improve the concentration of activated sludge returned to the treatment tank 21. Malodorous substance-decomposing bacteria refer to bacteria that have the ability to decompose malodorous substances in wastewater. The addition of malodorous substance-decomposing bacteria is expected to reduce the malodorous odor of wastewater (treated water) and improve water quality. Difficult substance-decomposing bacteria refer to bacteria that have the ability to decompose persistent substances in wastewater. The addition of the persistent substance-degrading bacteria is expected to reduce the persistent substances in the wastewater (treated water) and improve the water quality. The added bacteria may be used alone or in combination of two or more. The microbial agent may contain, for example, only the sludge-degrading bacteria X1, or may contain the sludge-degrading bacteria X1, the sludge-degrading bacteria X2, and the sedimentation-improving bacteria Y1.

[0023] The configuration of the wastewater treatment device 20 is an example and is not limited thereto. The wastewater treatment device 20 may include, for example, a pretreatment tank and a pre-treatment settling tank provided before the treatment tank 21. The treatment tank 21 may be configured with multiple stages.

[0024] The control device 3 is, for example, a local computer provided in a factory including the wastewater treatment device 20. The control device 3 controls the operation of the wastewater treatment device 20 in accordance with operating conditions set for the wastewater treatment device 20. The control device 3 also receives sludge images captured by a camera 25 and detection data detected by a sensor 26, and transmits the received sludge images and detection data to the information processing device 1. The control device 3 may be communicatively connected to the information processing device 1 via a relay device (not shown), such as a router or a gateway.

[0025] In this specification, the operating conditions include not only conditions directly set for the wastewater treatment device 20, but also conditions indirectly set in accordance with these directly set items. Examples of the operating conditions include the dilution amount of raw water, the flow rate of wastewater flowing into the wastewater treatment device 20 (treatment flow rate), the total organic carbon content of the wastewater, the aeration rate in the treatment tank 21, the concentration of suspended solids in activated sludge (hereinafter also referred to as sludge concentration), the air flow rate, the blower rotation speed, the return flow rate in the settling tank 22, the interface position, and the amount of sludge extracted.

[0026] 2 is a block diagram showing the configuration of the information processing device 1. The information processing device 1 is a computer and includes a control unit 11, a storage unit 12, and a communication unit 13. The information processing device 1 may be a multi-computer consisting of multiple computers, or may be a virtual machine virtually constructed by software.

[0027] The control unit 11 includes one or more arithmetic processing devices such as a central processing unit (CPU) or a graphics processing unit (GPU). The control unit 11 controls each component and executes processing using built-in memories such as a read-only memory (ROM) or a random access memory (RAM), a clock, a counter, etc. The functions of the information processing device 1 may be realized by software, or some or all of them may be realized by hardware such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0028] The storage unit 12 includes a non-volatile memory such as a hard disk, flash memory, or SSD (Solid State Drive). The storage unit 12 may be an external storage device connected to the information processing device 1. The storage unit 12 stores various computer programs and data referenced by the control unit 11. The storage unit 12 stores a computer program 1P for causing a computer to execute processing related to the estimation of a microbial agent according to sludge image information, and a learning model 121 required for executing the computer program 1P. The learning model 121 is a model generated by machine learning. The learning model 121 is expected to be used as a program module that constitutes part of artificial intelligence software.

[0029] A computer program (computer program product) including computer program 1P may be provided on a non-transitory recording medium 1A on which the computer program is readably recorded. Storage unit 12 stores the computer program read from recording medium 1A by a reading device (not shown). Recording medium 1A may be, for example, a magnetic disk, optical disk, or semiconductor memory. Alternatively, the computer program may be downloaded from an external server connected to a communications network and stored in storage unit 12. Computer program 1P may be a single computer program or may be composed of multiple computer programs, and may be executed on a single computer or on multiple computers interconnected by a communications network.

[0030] The communication unit 13 includes a communication device that realizes communication via the network N. The control unit 11 transmits and receives data to and from the control device 3 via the communication unit 13.

[0031] The configuration of the information processing device 1 is not limited to the above example, and may include, for example, an operation unit for accepting user operations, a display unit for displaying images, and the like.

[0032] 3 is a block diagram showing the configuration of the control device 3. The control device 3 includes a control unit 31, a storage unit 32, a communication unit 33, a display unit 34, and an operation unit 35.

[0033] The control unit 31 includes one or more arithmetic processing units such as a CPU, an MPU, a GPU, etc. The control unit 31 controls each component and executes processing using built-in memory such as a ROM or a RAM, a clock, a counter, etc. The functions of the control device 3 may be realized by software, or some or all of them may be realized by hardware such as an ASIC or an FPGA.

[0034] The storage unit 32 includes a non-volatile memory such as a hard disk, flash memory, or SSD. The storage unit 32 stores various computer programs and data referenced by the control unit 31. The storage unit 32 stores a computer program 3P for causing a computer to execute processing related to receiving the estimation results of a microbial agent.

[0035] A computer program (computer program product) including the computer program 3P may be provided on a non-transitory recording medium 3A on which the computer program is readably recorded. The storage unit 32 stores the computer program read from the recording medium 3A by a reading device (not shown). The recording medium 3A may be, for example, a magnetic disk, an optical disk, or a semiconductor memory. Alternatively, the computer program may be downloaded from an external server connected to a communications network and stored in the storage unit 32. The computer program 3P may be a single computer program or may be composed of multiple computer programs, and may be executed on a single computer or on multiple computers interconnected by a communications network.

[0036] The communication unit 33 includes a communication device that realizes communication via the network N. The control unit 31 transmits and receives data to and from the information processing device 1 via the communication unit 33.

[0037] The display unit 34 includes a display device such as a liquid crystal display, an organic EL (Electro Luminescence) display, etc. The display unit 34 displays the estimation results received from the information processing device 1 in accordance with instructions from the control unit 31.

[0038] The operation unit 35 is an interface that accepts user operations. The operation unit 35 includes, for example, a keyboard, a mouse, a touch panel device with a built-in display, a speaker, a microphone, etc. The operation unit 35 accepts operation input from the user and sends a control signal according to the operation content to the control unit 31.

[0039] 4 is an explanatory diagram showing an overview of the learning model 121. The learning model 121 receives sludge image information related to sludge images of the wastewater treatment system 2 as input, and outputs microbial agent information related to the microbial agent to be used in the wastewater treatment system 2. The learning model 121 is, for example, a neural network (NN). The learning model 121 of this embodiment outputs the type and amount of the microbial agent to be added as the microbial agent information.

[0040] The learning model 121 includes an input layer to which sludge image information is input, an output layer that outputs information indicating the type and amount of the microbial agent, and an intermediate layer (hidden layer) that extracts features. The intermediate layer may include a convolutional layer, a pooling layer, a fully connected layer, etc. The intermediate layer has multiple nodes that extract features of the input data and passes the extracted features using various parameters to the output layer. When sludge image information is input to the input layer, calculations are performed in the intermediate layer using the learned parameters, and output information indicating the type and amount of the microbial agent is output from the output layer.

[0041] The sludge image information input to the input layer of the learning model 121 is information related to a sludge image. The sludge image information of this embodiment includes various feature quantities that represent the state of microorganisms in the sludge image. Here, the term "microorganisms" refers not only to microorganisms themselves, such as bacteria and protozoa, but also to aggregates of microorganisms, such as flocs.

[0042] In the example shown in FIG. 4 , the sludge image information includes information indicating the state of protozoa, the state of filamentous bacteria, and the state of flocs. The sludge image information may also include color information. The state of protozoa includes the type and number of protozoa included in the sludge image. The state of filamentous bacteria includes the number of filamentous bacteria included in the sludge image. The state of flocs includes the size and / or color of the flocs included in the sludge image.

[0043] The method for deriving the type and number of protozoa, the number of filamentous bacteria, and the size and color of flocs based on a sludge image is not particularly limited, and various image analysis techniques can be used, for example. By performing image analysis on the sludge image, the protozoa, filamentous bacteria, and flocs contained in the sludge image can be extracted, and the type and number of protozoa, the number of filamentous bacteria, and the size of the flocs can be identified based on the extraction results.

[0044] When extracting various microorganisms from sludge images, different analysis methods may be used for each extraction target. For example, if the target is a relatively large protozoan, the number of each type of protozoan may be derived by detecting each type of protozoan in the sludge image using an object detection model that inputs a sludge image and outputs the position and type of object (protozoan) in the sludge image. If the target is a relatively small protozoan, the number of protozoans as moving objects in the sludge image may be determined by detecting moving objects using a moving object detection method from a time series of sludge images as moving images.

[0045] When the target is filamentous bacteria, the number of filamentous bacteria in the sludge image may be determined by performing binarization processing and blob analysis on the sludge image. Specifically, blob regions in the binarized sludge image whose feature quantities (e.g., area, circularity, height-to-width ratio, maximum height, maximum width, etc.) satisfy preset extraction conditions are detected as filamentous bacteria, and the total number of detected filamentous bacteria is determined.

[0046] When the target is a floc, the size (area) of the floc may be determined by performing a binarization process on the sludge image using a threshold calculated by a predetermined method and extracting blob regions in the binarized image as flocs. The size of the floc may be determined using a statistical value (e.g., average area, total area, median area, etc.) obtained from multiple flocs included in the sludge image. Alternatively, the color of the extracted floc region may be determined using a known color analysis method. Sludge image information may be derived using different sludge images for each type of target microorganism.

[0047] Time-series data representing the types and numbers of protozoa, the number of filamentous bacteria, and the size of flocs may be input to the learning model 121. The sludge image information may include the amount of change in the number of protozoa, the amount of change in the number of filamentous bacteria, and the amount of change in floc size obtained from the time-series data of the number of protozoa, the number of filamentous bacteria, and the size of flocs, and may also include the trend in change in the number of protozoa, the trend in change in the number of filamentous bacteria, and the trend in change in floc size (for example, increase or decrease).

[0048] The above sludge image information is an example, and various feature quantities obtained from the sludge image can be used as the sludge image information. Furthermore, the sludge image information is not limited to information including feature quantities based on the sludge image, but may be the sludge image itself. The sludge image information may include at least one of the state of protozoa, the state of filamentous bacteria, and the state of flocs.

[0049] The output layer of the learning model 121 outputs the estimation results of the type of microbial agent and the amount of the microbial agent to be added. The type of microbial agent is represented, for example, by the name of the microbial agent. The amount of the microbial agent to be added may be the amount to be added after a change, or may be the changed amount to be added. The learning model 121 may also output the estimation results of the amount to be added for each of multiple types of microbial agents. The configuration of the output layer is not particularly limited as long as information regarding the type and amount of the microbial agent to be estimated can be obtained.

[0050] The learning model 121 can be generated by preparing training data in which sludge image information obtained from past operational records is associated with the type and amount of microbial agent used, and then using the training data to train an untrained neural network. The correct values ​​for the type and amount of microbial agent are, for example, addition conditions selected by an experienced operator.

[0051] An example of training data is shown below. An increase in colpoda, a relatively large protozoan, is assumed to be due to an increase in ammonia concentration. Therefore, when the number of colpoda is high, a microbial agent containing persistent substance-degrading bacteria is associated with the type of microbial agent. An increase in filamentous bacteria is assumed to be due to a decrease in dissolved oxygen concentration. A decrease in dissolved oxygen concentration can cause odors to be emitted from the treatment tank or sedimentation tank. Therefore, when the number of filamentous bacteria is high, a microbial agent containing malodorous substance-degrading bacteria is associated with the type of microbial agent. A decrease in floc size indicates the dispersion or disintegration of flocs. Therefore, when the floc size is small, a microbial agent containing sedimentation-improving bacteria is associated with the type of microbial agent. The amount of microbial agent added can be appropriately set depending on the volume and treatment flow rate of the treatment tank 21 and the sedimentation tank 22.

[0052] The information processing device 1 inputs sludge image information contained in the training data into the input layer of the learning model 121, undergoes arithmetic processing in the intermediate layer, and obtains the type and amount of the microbial agent output from the output layer. The information processing device 1 compares the type and amount of the microbial agent output from the output layer with the type and amount of the microbial agent contained in the training data, and optimizes parameters such as the weights (coupling coefficients) between neurons so that the type and amount of the microbial agent output from the output layer approach the correct values. The parameter optimization method is not particularly limited, and the information processing device 1, for example, optimizes various parameters using an error backpropagation algorithm. Upon completion of learning, the information processing device 1 stores definition information related to the trained learning model 121 in the memory unit 12 as the trained learning model 121. In this way, a learning model 121 trained to appropriately estimate the type and amount of the microbial agent from sludge image information is constructed.

[0053] The learning model 121 is not limited to one that is generated and learned by the information processing device 1. The learning model 121 may be a model that has been trained by an external server and transmitted to the information processing device 1 and stored in the storage unit 12.

[0054] The configuration of the learning model 121 is not limited to the example shown in Fig. 4. The learning model 121 may be any model that can identify the type and amount of the microbial agent from the sludge image information. The learning model 121 may be a model based on other learning algorithms, such as a transformer, a convolution neural network (CNN), a recurrent neural network (RNN), a support vector machine (SVM), or a decision tree.

[0055] 5 is a flowchart showing an example of a processing procedure executed by the information processing device 1. The processing in the flowchart below is executed by the control unit 11 in accordance with a computer program 1P stored in the storage unit 12 of the information processing device 1. The control unit 11 repeatedly executes the following processing every time a new sludge image is captured by the camera 25, for example.

[0056] The control unit 11 of the information processing device 1 acquires the sludge image by receiving the sludge image of the wastewater treatment system 2 from the control device 3 (step S11). The control unit 11 stores the received sludge image in the memory unit 12. The sludge image is associated with identification information that identifies the control device 3 or the wastewater treatment system 2. In step S11, the control unit 11 may acquire the sludge image by reading out the sludge image stored in the memory unit 12.

[0057] The control unit 11 derives sludge image information by using various image analysis techniques based on the acquired sludge image (step S12). The sludge image information includes, for example, the type and number of protozoa, the number of filamentous bacteria, and the size of flocs in the sludge image.

[0058] The control unit 11 inputs the derived sludge image information into the learning model 121 (step S13) and acquires microbial agent information output from the learning model 121 (step S14). The microbial agent information includes the name of the microbial agent, which indicates the type of microbial agent, and the amount of the microbial agent added.

[0059] The control unit 11 outputs the acquired microbial agent information to the control device 3 (step S15). The control unit 11 generates screen information showing, for example, the estimated results of the microbial agent information, and transmits the generated screen information to the control device 3 identified by the identification information associated with the sludge image. The screen information may include, in addition to the microbial agent information, sludge image information obtained from the sludge image, i.e., the type and number of protozoa, the number of filamentous bacteria, and the size of flocs.

[0060] The control unit 31 of the control device 3 receives the screen information transmitted from the information processing device 1 and displays the received screen information on the display unit 34. The display unit 34 displays the names of one or more microbial agents to be added and the amounts of the microbial agents to be added. The operator determines the conditions for adding the microbial agents based on the microbial agent information displayed on the display unit 34 and sets the addition conditions in the control device 3 using the operation unit 35. The control device 3 controls the wastewater treatment device 20 in accordance with the set addition conditions to add the microbial agent.

[0061] In the above-described processing, the information processing device 1 may output the microbial agent information to a device (for example, a personal computer used by an operator) other than the control device 3 from which the sludge image was acquired.

[0062] The information processing device 1 may also transmit an instruction to add a microbial agent according to the acquired microbial agent information to the control device 3. The control device 3 adds the microbial agent to the wastewater treatment device 20 in accordance with the instruction to add the microbial agent received from the information processing device 1. According to the above configuration, the process from estimating the microbial agent information to adding the microbial agent can be automatically performed as a series of processes.

[0063] The information processing device 1 preferably performs the above-described processing in real time, thereby enabling the conditions for adding the microbial agent to be adjusted as needed, but may also be configured to perform the estimation processing collectively at predetermined intervals.

[0064] According to this embodiment, appropriate microbial agent information can be estimated and presented to the user based on sludge images captured in the wastewater treatment system 2. Because the quality and quantity of wastewater in the wastewater treatment system 2 change daily, changes also occur in the ecosystem of microorganisms that make up the activated sludge contained in the wastewater. It is not easy to grasp the state of activated sludge, which is a collection of many types of microorganisms, and appropriately adjust the conditions for adding a microbial agent. According to this embodiment, by using the learning model 121, microbial agent information can be accurately estimated. Appropriate control of the wastewater treatment system 2 is possible without relying on the skill of the operator, improving the treatability of the wastewater treatment system 2 and stabilizing its operation.

[0065] The learning model 121 uses sludge image information obtained from sludge images as input elements, and can accurately estimate microbial agent information by taking into account the state of various microorganisms that have a significant impact on water quality.

[0066] Second Embodiment In the second embodiment, the details of the explanatory variables and the objective variables of the learning model 121 are different from those in the first embodiment. In the following embodiments, differences from the first embodiment will be mainly described, and the same reference numerals will be used to designate components that are common to the first embodiment, and detailed descriptions thereof will be omitted.

[0067] 6 is an explanatory diagram showing an overview of the learning model 121 of the second embodiment. The learning model 121 of the second embodiment receives sludge image information and operating data in the wastewater treatment system 2 as input, and outputs microbial agent information and operating conditions for the wastewater treatment system 2. Except for differences in the content of the input information and output information, the configuration of the learning model 121 of the second embodiment is the same as that of the learning model 121 of the first embodiment, and therefore detailed description thereof will be omitted.

[0068] The operating data input to the learning model 121 is information representing the operating state of the wastewater treatment system 2. The operating data includes, for example, detection data from the sensors 26 and set values ​​of operating conditions set in the wastewater treatment device 20. The operating data may include values ​​calculated from the detection data and / or operating conditions. Examples of the operating data include the temperature, flow rate, pH, oxidation-reduction potential, dissolved oxygen concentration, sludge concentration, total organic carbon, ammonia concentration, nitrogen concentration, phosphorus concentration, aeration volume of the treatment tank 21, and sludge load of the wastewater. The sludge load is obtained by multiplying the BOD or TOC of the wastewater by the flow rate and dividing the result by the sludge concentration of the treatment tank 21. In the example shown in FIG. 6 , the operating data input to the learning model 121 includes the dissolved oxygen concentration, ammonia concentration, pH, aeration volume, and sludge load.

[0069] The operating conditions output from the learning model 121 include one or more of the operating conditions described in the first embodiment and change values ​​for the operating conditions. The change value may be the set value of the changed operating condition or the change amount of the set value. In the example shown in Figure 6, the operating conditions output from the learning model 121 include the aeration volume, treatment flow rate, sludge concentration, and dilution volume of raw water. Note that the output from the learning model 121 in the second embodiment may be only microbial agent information.

[0070] As in the first embodiment, the learning model 121 is constructed by machine learning using training data in which sludge image information and operating data obtained from past operating records are associated with microbial agent information and operating conditions.

[0071] FIG. 7 is a flowchart illustrating an example of a processing procedure executed by the information processing apparatus 1 according to the second embodiment.

[0072] The control unit 11 of the information processing device 1 acquires (step S21) sludge images and operation data of the wastewater treatment system 2. The control unit 11 acquires, via the control device 3, the sludge images captured by the camera 25, the detection data detected by the various sensors 26, and the setting values ​​of the operation conditions held by the control device 3.

[0073] The control unit 11 derives sludge image information based on the acquired sludge image (step S22), similar to step S12 in the first embodiment.

[0074] The control unit 11 inputs the derived sludge image information and the acquired operating data into the learning model 121 (step S23), and acquires the microbial agent information and operating conditions output from the learning model 121 (step S24). The control unit 11 outputs the acquired microbial agent information and operating conditions to the control device 3 (step S25), and the series of processes ends.

[0075] The control unit 31 of the control device 3 displays the microbial agent information and operating conditions received from the information processing device 1 on the display unit 34. The control unit 31 may automatically set the microbial agent addition conditions and operating conditions in accordance with the microbial agent information and operating conditions received from the information processing device 1, and control the operation of the wastewater treatment device 20.

[0076] According to this embodiment, in addition to sludge image information, it is possible to estimate microbial agent information and operating conditions taking into account the operating state of the wastewater treatment system 2. It is possible to accurately estimate microbial agent information and operating conditions that are in line with the operating state of the wastewater treatment system 2, thereby further improving the treatability of the wastewater treatment system 2. Furthermore, by presenting the operating conditions in addition to the microbial agent, it is possible to more appropriately support the operation management of the wastewater treatment system 2.

[0077] Third Embodiment In a third embodiment, a configuration for learning a learning model 121 by reinforcement learning will be described.

[0078] FIG. 8 is an explanatory diagram showing an overview of the learning model 121 of the third embodiment. The learning model 121 of the third embodiment learns microbial agent information corresponding to sludge image information through reinforcement learning. In the example shown in FIG. 8, the learning model 121 is a reinforcement learning model having a deep neural network, and is a DQN (Deep Q Network). The learning model 121 is a neural network model that outputs the action value of each of multiple actions a when a state s is input. Actions a with a high action value represent appropriate microbial agent information.

[0079] The learning model 121 includes an input layer to which a state s is input, an output layer that outputs the action value Q(s, a) of an action a in the input state s, and an intermediate layer that extracts the features of the state s.

[0080] The state s input to the learning model 121 includes sludge image information including the types and numbers of protozoa, the number of filamentous bacteria, and the size of flocs, as described above.

[0081] The output layer of the learning model 121 includes nodes corresponding to multiple actions a. Each node outputs an action value Q(s, a) for the action a corresponding to the node. The action a corresponds to a combination of the type and amount of microbial agent added. By selecting the action node with the largest action value based on the action values ​​output from the learning model 121, the type and amount of microbial agent corresponding to the sludge image information can be obtained.

[0082] The information processing device 1 generates a learning model 121 through deep reinforcement learning using a predetermined data group. In reinforcement learning, the sludge image information is used as the "state," the type and amount of the microbial agent are used as the "action," a required "reward" is calculated, and the value of the Q value or Q function (action value function) is learned.

[0083] The update formula for the action value function is expressed by the following formula (1): Q(s, a) ← Q(s, a) + α(R + γmaxQ(s_next, a_next) - Q(s, a) ... (1) where, s: state a: action selected in state s α: learning coefficient R: reward obtained as a result of the action γ: discount rate maxQ(s_next, a_next): maximum value of action value Q for the action that can be taken in the next state

[0084] The information processing device 1 learns the learning model 121 by optimizing the parameters of the learning model 121, for example, by backpropagation so that the second term on the right side of the above formula (1) approaches zero. In other words, the parameters of the learning model 121 are updated so that the error between the expected value of the reward and the current behavioral evaluation approaches zero.

[0085] The information processing device 1 performs reward-based reinforcement learning by setting the sludge image information before the change in the microbial agent addition conditions as state s and the type and amount of microbial agent after the change as action a. The reward is automatically calculated by the information processing device 1 based on, for example, evaluation information indicating the evaluation results of the operating status of the wastewater treatment system 2.

[0086] The evaluation information may include evaluation results for one or more evaluation items selected from the amount of excess sludge reduction, the amount of greenhouse gas reduction, the degree of improvement in settling property, the degree of suppression of bulking, the amount of chemicals reduced, the degree of improvement in treatability, the degree of reduction in offensive odors, the degree of improvement in water quality, and the stability of operation in the wastewater treatment system 2. The information processing device 1 calculates the reward so that the smaller the deviation between the target value of each of the above evaluation items and the actual evaluation value of the wastewater treatment system 2, the larger the value. If the deviation between the target value and the actual evaluation value is large, a reward of zero or a negative reward is given. The calculation formula for calculating the reward is not particularly limited.

[0087] The information processing device 1 trains the learning model 121 based on the sludge image information input to the learning model 121, the action value Q output when each state variable is input, and the calculated reward. The information processing device 1 uses the above formula (1) to train the parameters of the learning model 121 so as to maximize the reward.

[0088] The reinforcement learning method of the learning model 121 is not limited. The learning model 121 may be configured to adjust the microbial agent information using a map that associates state quantities with action values. In addition, in the reinforcement learning of the learning model 121, the parameters of the learning model 121 may be adjusted using a small device (lab-scale) such as a simulator or laboratory, and then the learning model 121 after parameter adjustment may be deployed in a real environment.

[0089] FIG. 9 is a flowchart illustrating an example of a processing procedure related to reinforcement learning executed by the information processing device 1 according to the third embodiment.

[0090] The control unit 11 of the information processing device 1 sets the parameters of the learning model 121 to initial values ​​(step S31).

[0091] The control unit 11 acquires sludge image information including the type and number of protozoa, the number of filamentous bacteria, and the size of flocs obtained from the sludge image via the control device 3 (step S32). The sludge image information corresponds to state s.

[0092] The control unit 11 determines microbial agent information including the type and amount of the microbial agent based on the acquired sludge image information (step S33). The microbial agent information corresponds to action a. The control unit 11 selects action a that can be taken in state s.

[0093] The control unit 11 acquires sludge image information based on the state resulting from action a, i.e., the sludge image acquired after the microbial agent has been added in accordance with the determined microbial agent information (step S34). Specifically, the control unit 11 outputs the microbial agent information determined in step S33 to the control device 3, and acquires the sludge image after the microbial agent has been added to the wastewater treatment system 2 in accordance with the output microbial agent information, thereby acquiring the sludge image information based on the sludge image.

[0094] The control unit 11 acquires evaluation information on the wastewater treatment system 2 after the microbial agent has been added in accordance with the determined microbial agent information (step S35). The evaluation information is obtained by measuring or evaluating each evaluation item in the wastewater treatment system 2.

[0095] The control unit 11 executes reinforcement learning processing for the learning model 121 (step S36). Specifically, the control unit 11 calculates a reward R based on the acquired evaluation information, and uses the above formula (1) to learn (update) the parameters of the learning model 121 so that the current Q value becomes the maximum reward obtainable among the actions that can be taken in the next state. The information processing device 1 executes a loop process that repeats steps S32 to S35 as necessary. When learning is completed, the information processing device 1 stores definition information related to the trained learning model 121 in the storage unit 12 as the trained learning model 121. The information processing device 1 may terminate learning, for example, when it has updated the parameters a predetermined number of times.

[0096] The information processing device 1 of the third embodiment estimates microbial agent information corresponding to sludge image information using the above-described learning model 121 learned by reinforcement learning. Details of the estimation process using the learning model 121 are the same as those of the first embodiment.

[0097] According to this embodiment, the learning model 121 can be constructed by reinforcement learning, eliminating the need to generate training data and reducing the burden required to construct the learning model 121. In a wastewater treatment system 2 whose state changes due to a variety of factors, it is not easy to define an evaluation scale for the effect of adding a microbial agent. According to this embodiment, the reward used for learning can be efficiently calculated by using evaluation information indicating an evaluation of the operating state of the wastewater treatment system 2.

[0098] Fourth Embodiment In the fourth embodiment, the details of the explanatory variables and the objective variables of the learning model 121 that undergoes reinforcement learning are different from those in the third embodiment.

[0099] FIG. 10 is an explanatory diagram showing an overview of the learning model 121 of the fourth embodiment. The learning model 121 of the fourth embodiment is a DQN, as in the third embodiment. The state s input to the learning model 121 of the fourth embodiment includes sludge image information and operating data. The action a in the output layer of the learning model 121 includes the type and amount of the microbial agent and the operating conditions. Note that the action a in the output layer of the learning model 121 of the fourth embodiment may be configured to include only the type and amount of the microbial agent (microbial agent information).

[0100] Other than the differences in the content of the input information and output information, the configuration and learning method of the learning model 121 of the fourth embodiment are the same as those of the learning model 121 of the third embodiment, so detailed explanations will be omitted. However, the information processing device 1 performs reinforcement learning on the learning model 121 so as to maximize the reward based on the reward calculated based on the above-mentioned evaluation information.

[0101] The information processing device 1 of the fourth embodiment estimates microbial agent information and operating conditions according to sludge image information and operating data using the above-mentioned learning model 121 learned by reinforcement learning. Details of the estimation process using the learning model 121 are the same as those of the second embodiment.

[0102] Fifth Embodiment In the fifth embodiment, a configuration for deploying a learning model to a control device 3 will be described.

[0103] Fig. 11 is a diagram showing the configuration of a management system 100 according to the fifth embodiment. An information processing device 1 according to the fifth embodiment is communicatively connected to control devices 3 in a plurality of wastewater treatment systems 2. In Fig. 11, the wastewater treatment device 20, and the control units of the information processing device 1 and the control device 3 are omitted for the sake of simplicity.

[0104] 11 , in the management system 100 of the fifth embodiment, a learning model 121 generated and trained by the information processing device 1 is deployed to each control device 3 and stored in the storage unit 32 of the control device 3. In the following, the learning model stored in the storage unit 32 of the control device 3 will be referred to as a learning model 321 to distinguish it from the learning model 121 stored in the information processing device 1. In the following, as an example, the learning model 121 will be described as inputting sludge image information and outputting microbial agent information, but the learning model 121 of this embodiment may be any of the learning models 121 described in the above embodiments.

[0105] The control device 3 can retrain the learning model 321 in each control device 3 using sludge images, etc. obtained from the wastewater treatment system 2 that the control device 3 controls.

[0106] 12 is a flowchart showing an example of a processing procedure executed by the management system 100 according to the fifth embodiment. The processing in the flowchart in FIG. 12 is executed by the control unit 11 in accordance with a computer program 1P stored in the storage unit 12 of the information processing device 1, and is also executed by the control unit 31 in accordance with a computer program 3P stored in the storage unit 32 of the control device 3.

[0107] The control unit 11 of the information processing device 1 transmits the constructed trained learning model 121 to the control device 3 (step S41).

[0108] The control unit 31 of the control device 3 receives the learning model 321 from the information processing device 1 and stores the received learning model 321 in the storage unit 32 (step S42). Through the above-mentioned processing, the learning model 321 is deployed to the control device 3.

[0109] The control unit 31 updates the learning model 321 by re-training it using sludge images and the like obtained from the wastewater treatment system 2 (step S43). When performing supervised learning, the control unit 31 acquires training data including sludge image information based on sludge images newly obtained in the wastewater treatment system 2 and microbial agent information related to the microbial agent to be used in the wastewater treatment system 2. The control unit 31 fine-tunes the learning model 321 based on the acquired training data. Details of the learning method for the learning model 321 are the same as those in the first embodiment.

[0110] In steps S44 to S47, the control unit 31 executes the same processes as steps S11 to S14 in the first embodiment, thereby estimating microbial agent information using the updated learning model 321.

[0111] In step S43, when unsupervised learning is performed, the control unit 31 acquires sludge image information based on a sludge image newly obtained in the wastewater treatment system 2, microbial agent information related to the microbial agent used in the wastewater treatment system 2, and evaluation information for predetermined evaluation items. The control unit 31 uses a reward corresponding to the acquired evaluation information to perform reinforcement learning on the learning model 321 so as to increase the reward. Details of the learning method of the learning model 321 are the same as those in the third embodiment.

[0112] Each control device 3 in the management system 100 executes the above-described processing to update (adjust) various parameters of the learning model 321 deployed to the device itself.

[0113] In the above-described processing, the generation of training data and the calculation of rewards may be executed on the information processing device 1 side.

[0114] According to this embodiment, the learning model 321 can be updated using data from each individual wastewater treatment system 2, so that the learning model 321 can be customized for each wastewater treatment system 2, thereby improving the estimation accuracy of the learning model 321.

[0115] Sixth Embodiment In the sixth embodiment, a configuration will be described in which federated learning is performed based on update parameters of the learning model 321 deployed to each control device 3.

[0116] In the management system 100 of the sixth embodiment, as described in the fifth embodiment, the learning model 121 generated and learned by the information processing device 1 is deployed to each control device 3, and thereby stored in the storage unit 32 of the control device 3. Furthermore, fine tuning or reinforcement learning of the learning model 321 is performed in each control device 3, and thereby various parameters of the learning model 321 are updated.

[0117] FIG. 13 is a flowchart illustrating an example of a processing procedure executed by the information processing apparatus 1 according to the sixth embodiment.

[0118] The control unit 11 of the information processing device 1 acquires updated parameters of each learning model 321 from each of the multiple control devices 3 (step S51). The control unit 11 performs federated learning of the learning model 121 based on the parameters related to each learning model 321 (step S52), and updates the learning model 121. In federated learning, parameters learned by individual local computers are aggregated and learned in the information processing device 1 on the cloud, thereby improving the accuracy of the learning model 121. The information processing device 1 performs federated learning by, for example, calculating a simple average of multiple parameters acquired from each control device 3, or calculating a weighted average that assigns a weight to a learning model 321 with high accuracy or a learning amount that is large.

[0119] The control unit 11 transmits the updated learning model 121 to each of the multiple control devices 3 (step S53), and deploys the updated learning model 121 to each control device 3.

[0120] According to this embodiment, the estimation accuracy of the learning model 321 can be improved through the operation of the management system 100.

[0121] The following supplementary notes are further disclosed with respect to the above embodiments. (Supplementary Note 1) A computer program causing a computer to execute a process of acquiring sludge image information related to sludge images of a wastewater treatment system using an activated sludge process, and inputting the acquired sludge image information into a model that outputs microbial agent information related to a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system is input, and outputting the microbial agent information. (Supplementary Note 2) The computer program according to Supplementary Note 1, wherein the sludge image information includes at least one of the state of protozoa, the state of filamentous bacteria, and the state of flocs obtained from the sludge image. (Supplementary Note 3) The computer program according to Supplementary Note 1 or Supplementary Note 2, wherein the microbial agent information includes the type and amount of the microbial agent. (Supplementary Note 4) A computer program according to any one of Supplements 1 to 3, acquiring operating data of the wastewater treatment system, and inputting the acquired sludge image information and operating data into the model that outputs microbial agent information related to a microbial agent to be used in the wastewater treatment system when sludge image information and operating data of the wastewater treatment system are input, and outputting the microbial agent information. (Supplementary Note 5) The computer program according to any one of Supplementary Notes 1 to 4, wherein acquired sludge image information is input to the model, which outputs microbial agent information related to a microbial agent used in a wastewater treatment system and operating conditions in the wastewater treatment system when sludge image information of the wastewater treatment system is input, and outputs the microbial agent information and operating conditions. (Supplementary Note 6) The computer program according to any one of Supplementary Notes 1 to 5, wherein evaluation information indicating an evaluation result of the operating status of the wastewater treatment system is acquired, and a reward corresponding to the acquired evaluation information is used to reinforce learning the model so that the reward is increased. (Supplementary Note 7) The computer program according to Supplementary Note 6, wherein the sludge image information including at least one of a state of protozoa, a state of filamentous bacteria, and a state of flocs obtained from the sludge image is used as a state, and microbial agent information including the type and amount of the microbial agent is used as an action, and wherein reinforcement learning is performed on the model so that the reward corresponding to the evaluation information is increased.(Supplementary Note 8) An information processing device comprising a control unit that executes a process of acquiring sludge image information related to sludge images of a wastewater treatment system using an activated sludge method, and inputting the acquired sludge image information into a model that outputs microbial agent information related to a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system is input, to output the microbial agent information. (Supplementary Note 9) An information processing method in which a computer executes a process of acquiring sludge image information related to sludge images of a wastewater treatment system using an activated sludge method, and inputting the acquired sludge image information into a model that outputs microbial agent information related to a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system is input, to output the microbial agent information. (Supplementary Note 10) An information processing method that deploys a model that outputs microbial agent information related to a microbial agent to be used in the wastewater treatment system when sludge image information of a wastewater treatment system using an activated sludge method is input into a computer that controls the wastewater treatment system, acquires sludge image information related to sludge images of the wastewater treatment system, and inputs the acquired sludge image information into the deployed model to output the microbial agent information. (Supplementary Note 11) The information processing method according to Supplementary Note 10, comprising acquiring training data including sludge image information newly obtained in the wastewater treatment system and microbial agent information related to a microbial agent used in the wastewater treatment system, and adjusting the model based on the acquired training data. (Supplementary Note 12) The information processing method according to Supplementary Note 10 or Supplementary Note 11, comprising acquiring evaluation information indicating an evaluation result of an operating status of the wastewater treatment system, and performing reinforcement learning on the model using a reward according to the acquired evaluation information so as to increase the reward.(Supplementary Note 13) An information processing method comprising: deploying, on each of a plurality of computers controlling each wastewater treatment system, a model that outputs microbial agent information regarding a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system using an activated sludge method is input; collecting, from each of the plurality of computers, parameters of each model in each wastewater treatment system adjusted based on a data group including sludge image information newly obtained in each wastewater treatment system and microbial agent information regarding the microbial agent to be used in each wastewater treatment system; performing federated learning of the model based on the collected parameters of each model; and deploying the model after federated learning on each of the plurality of computers.

[0122] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The technical features described in each embodiment can be combined with each other, and the scope of the present invention is intended to include all modifications within the scope of the claims and equivalents thereto. The sequences shown in each embodiment are not limited, and within the scope of no contradiction, each processing step may be executed in a different order, or multiple processes may be executed in parallel. The entity that performs each process is not limited, and within the scope of no contradiction, the process of each device may be executed by another device.

[0123] The matters described in each embodiment can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations, regardless of the reference format. Furthermore, the claims use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used.

[0124] REFERENCE SIGNS LIST 100 Management system 1 Information processing device 11 Control unit 12 Memory unit 13 Communication unit 121 Learning model 1P Computer program 1A Recording medium 2 Wastewater treatment system 3 Control device 31 Control unit 32 Memory unit 33 Communication unit 34 Display unit 35 Operation unit 3P Computer program 3A Recording medium 321 Learning model

Claims

1. A computer program that causes a computer to execute a process of acquiring sludge image information relating to sludge images of a wastewater treatment system using an activated sludge method, inputting the acquired sludge image information into a model that outputs microbial agent information relating to a microbial agent to be used in the wastewater treatment system when the sludge image information of the wastewater treatment system is input, and outputting the microbial agent information.

2. The computer program according to claim 1, wherein the sludge image information includes at least one of a state of protozoa, a state of filamentous bacteria, and a state of flocs obtained from the sludge image.

3. The computer program according to claim 1 or 2, wherein the microbial agent information includes the type and amount of the microbial agent.

4. A computer program as claimed in any one of claims 1 to 3, which acquires operating data for the wastewater treatment system, and inputs the acquired sludge image information and operating data into the model which outputs microbial agent information regarding a microbial agent to be used in the wastewater treatment system when sludge image information and operating data of the wastewater treatment system are input, and outputs the microbial agent information.

5. A computer program according to any one of claims 1 to 4, which inputs acquired sludge image information into the model, which outputs microbial agent information regarding a microbial agent to be used in the wastewater treatment system and operating conditions in the wastewater treatment system when sludge image information of the wastewater treatment system is input, and outputs microbial agent information and operating conditions.

6. A computer program according to any one of claims 1 to 5, further comprising: acquiring evaluation information indicating an evaluation result of the operating status of the wastewater treatment system; and performing reinforcement learning on the model using a reward corresponding to the acquired evaluation information so as to increase the reward.

7. The computer program according to claim 6, wherein the sludge image information including at least one of the state of protozoa, the state of filamentous bacteria, and the state of flocs obtained from the sludge image is defined as a state, and microbial agent information including the type and amount of the microbial agent is defined as an action, and the model is subjected to reinforcement learning so as to increase the reward corresponding to the evaluation information.

8. An information processing device comprising a control unit that executes a process of acquiring sludge image information relating to sludge images of a wastewater treatment system using an activated sludge method, and inputting the acquired sludge image information into a model that outputs microbial agent information relating to a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system is input, and outputting microbial agent information.

9. An information processing method in which a computer executes a process of acquiring sludge image information relating to sludge images of a wastewater treatment system using an activated sludge method, inputting the acquired sludge image information into a model which outputs microbial agent information relating to a microbial agent to be used in the wastewater treatment system when sludge image information of the wastewater treatment system is input, and outputting microbial agent information.

10. An information processing method, comprising: deploying a model that outputs microbial agent information regarding a microbial agent to be used in a wastewater treatment system using an activated sludge method when sludge image information of the wastewater treatment system is inputted, on a computer that controls the wastewater treatment system; acquiring sludge image information regarding sludge images of the wastewater treatment system; and inputting the acquired sludge image information into the deployed model to output the microbial agent information.

11. The information processing method according to claim 10, further comprising acquiring training data including sludge image information newly obtained in the wastewater treatment system and microbial agent information regarding a microbial agent to be used in the wastewater treatment system, and adjusting the model based on the acquired training data.

12. The information processing method according to claim 10 or 11, further comprising: acquiring evaluation information indicating an evaluation result of the operating status of the wastewater treatment system; and performing reinforcement learning on the model using a reward according to the acquired evaluation information so as to increase the reward.

13. An information processing method comprising: deploying a model that outputs microbial agent information regarding a microbial agent to be used in a wastewater treatment system using an activated sludge method when sludge image information from the wastewater treatment system is input to each of a plurality of computers that control each wastewater treatment system; collecting parameters of each model in each wastewater treatment system from each of the plurality of computers, the parameters being adjusted based on a data group including sludge image information newly obtained in each wastewater treatment system and microbial agent information regarding the microbial agent to be used in each wastewater treatment system; associatively learning the model based on the collected parameters of each model; and deploying the model after associative learning to each of the plurality of computers.

Images