JP7807311B2 - Systems and methods for assisting in wastewater treatment - Google Patents

Description

The present invention relates to computer technology for supporting sewage treatment.

In sewage treatment plants that treat sewage based on the standard activated sludge process, ammonia (NH ₃ ) in wastewater is nitrified by utilizing the metabolic activity of various nitrifying bacteria that live in the activated sludge in the aeration tank. Examples of nitrifying bacteria used in this process include ammonia oxidizing bacteria (AOB) and ammonia oxidizing archaea (AOA), which oxidize ammonia to produce nitrite (HNO ₂ ), and nitrite oxidizing bacteria (NOB), which oxidize nitrite to produce nitrate (HNO ₃ ).

The strength of activated sludge's nitrification ability depends on the viable cell count and cell concentration of various nitrifying bacteria in the activated sludge. Furthermore, the viable cell count and cell concentration of various nitrifying bacteria in the activated sludge are generally positively correlated with the dissolved oxygen concentration (DO) in the activated sludge. Therefore, sewage treatment plants that treat sewage based on the standard activated sludge process often include a means for aerating the aeration tank in order to maintain a constant dissolved oxygen concentration in the activated sludge in the aeration tank (see, for example, Patent Document 1).

Patent No. 4552393

Aeration in the aeration tank is typically performed using a blower attached to the aeration tank. This blower is typically controlled to maintain a constant dissolved oxygen concentration in the activated sludge in the aeration tank and a constant inflow ratio, which represents the ratio of the amount of air discharged by the blower (hereinafter referred to as the "air discharge volume") to the amount of sewage inflow into the sewage treatment plant (hereinafter referred to as the "sewage inflow volume"), so that the sewage treatment plant where it is installed meets statutory water quality standards. However, indicators such as the dissolved oxygen concentration in the activated sludge and the inflow ratio are not definitive and are difficult to uniquely determine because they change with fluctuations in the amount of sewage inflow into the sewage treatment plant and changes in the state of the activated sludge in the aeration tank. Therefore, when attempting to set these indicators as blower control target values and maintain them constant, skilled operators must adjust the blower's operation in a timely manner, which places a burden on the individual operator and on the management of the sewage treatment plant.

Furthermore, reducing power consumption is an urgent issue in the operation of sewage treatment plants today. For this reason, it is preferable that the blower control target value be one that enables sewage treatment plants to meet effluent water quality standards while also achieving energy conservation.

In consideration of the above issues, the present invention aims to provide technology that can automatically determine an optimal blower control target value.

The sewage treatment support system according to the present invention comprises a clustering unit, a transition condition learning unit, and a control target value determination unit. The clustering unit performs clustering of status data for a predetermined period representing the status of activated sludge in an aeration tank of a sewage treatment plant that performs sewage treatment based on a standard activated sludge process. The transition condition learning unit learns the conditions for transitions between clusters in multiple clusters generated by clustering. The control target value determination unit determines the control target value of a blower that discharges air to aerate the aeration tank based on the learned transition conditions and a time series of new status data representing the status of activated sludge in the aeration tank. The status data includes water quality data representing the quality of secondary effluent from the sewage treatment plant and operating data representing the operating status of the sewage treatment plant.

Furthermore, the problems and solutions disclosed in this application will be made clear in the detailed description of the invention and the drawings.

This invention makes it possible to automatically determine optimal blower control target values.

1 shows an example of the configuration of a sewage treatment support system according to an embodiment. 1 shows an example of the overall flow of processing performed in an embodiment. 10 shows an example of the flow of a transition condition learning process. An example of the water quality table configuration is shown below. 10 shows an example of the configuration of an operation table. 10 shows an example of the configuration of a state table. 10 shows an example of the configuration of a clustering state table. 1 shows an example of transition between clusters. An example of a transition condition learning matrix is shown below. 10 shows an example of the flow of a control target value determination process. 10 shows an example of the flow of a transition destination estimation process. An example of a transition destination estimation matrix is shown below. 1 shows an example of a display screen. 1 shows an example of a display screen. 10 shows an example of the flow of a transition condition learning process when learning transitions over multiple periods. 10 shows an example of a mean matrix for transition condition learning.

In the following description, an "interface apparatus" may refer to one or more interface devices. The one or more interface devices may be at least one of the following:
One or more I/O (Input/Output) interface devices. The I/O (Input/Output) interface devices are interface devices for at least one of an I/O device and a remote display computer. The I/O interface device for the display computer may be a communications interface device. The at least one I/O device may be a user interface device, for example, either an input device such as a keyboard and a pointing device, or an output device such as a display device.
One or more communication interface devices. The one or more communication interface devices may be one or more homogeneous communication interface devices (e.g., one or more NICs (Network Interface Cards)) or two or more heterogeneous communication interface devices (e.g., a NIC and an HBA (Host Bus Adapter)).

In the following description, "memory" refers to one or more memory devices, which are an example of one or more storage devices, and may typically be a primary storage device. At least one memory device in the memory may be a volatile memory device or a non-volatile memory device.

In the following description, a "persistent storage device" may refer to one or more persistent storage devices, which are an example of one or more storage devices. A persistent storage device may typically be a non-volatile storage device (e.g., an auxiliary storage device), and more specifically, may be, for example, a hard disk drive (HDD), a solid state drive (SSD), a non-volatile memory express (NVME) drive, or a storage class memory (SCM).

Also, in the following description, "storage device" may refer to at least memory, including memory and persistent storage devices.

In the following description, a "processor" may refer to one or more processor devices. The at least one processor device may typically be a microprocessor device such as a CPU (Central Processing Unit), but may also be another type of processor device such as a GPU (Graphics Processing Unit). The at least one processor device may be single-core or multi-core. The at least one processor device may also be a processor core. The at least one processor device may also be a broader processor device such as a circuit that is a collection of gate arrays written in a hardware description language that performs some or all of the processing (e.g., an FPGA (Field-Programmable Gate Array), a CPLD (Complex Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit)).

In the following description, functions may be described using the expression "yyy unit," but the functions may be realized by one or more computer programs being executed by a processor, by one or more hardware circuits (e.g., FPGAs or ASICs), or by a combination of these. When a function is realized by a program being executed by a processor, the specified processing is performed using storage devices and/or interface devices, etc., as appropriate, and therefore the function may be considered to be at least part of the processor. Processing described using a function as the subject may be processing performed by a processor or a device having that processor. A program may be installed from program source. The program source may be, for example, a program distribution computer or a computer-readable recording medium (e.g., a non-transitory recording medium). The description of each function is an example, and multiple functions may be combined into a single function, or a single function may be divided into multiple functions.

In addition, in the following explanation, processing may be described using a "program" as the subject, but processing described using a program as the subject may also be processing performed by a processor or a device having that processor. Furthermore, two or more programs may be realized as a single program, or one program may be realized as two or more programs.

In the following explanation, information that produces an output in response to an input may be described using expressions such as "xxx table," but this information may be a table of any structure, or a learning model such as a neural network that generates an output in response to an input, a genetic algorithm, or a random forest. Therefore, "xxx table" can also be referred to as "xxx information." In the following explanation, the structure of each table is an example, and one table may be divided into two or more tables, or all or part of two or more tables may be one table.

Also, in the following explanation, "UI" stands for User Interface, but typically refers to GUI (Graphical User Interface).

In the following description, a system that supports sewage treatment will be referred to as a "sewage treatment support system." A sewage treatment support system may be one or more physical computers, a software-defined system implemented by at least one physical computer executing specified software, or a system implemented on a cloud platform (typically multiple types of computing resources including processors and storage devices). For example, if a computer has a display device and displays information on its own display device, that computer may be a sewage treatment support system. Also, for example, if a first computer (e.g., a server) transmits output information to a remote second computer (a display computer (e.g., a user terminal described below)) and the display computer displays that information (the first computer displays information on the second computer), at least the first of the first and second computers may be a sewage treatment support system. In other words, when the sewage treatment support system "displays output information," it may mean that the output information is displayed on a display device possessed by the computer, or that the computer sends the output information to a display computer (in the latter case, the output information is displayed by the display computer).

This embodiment is described in detail below.

The following embodiment will be described using as an example a sewage treatment support system that determines the control target value for a blower that is installed in an aeration tank of a sewage treatment plant that performs sewage treatment based on the standard activated sludge process and that discharges air to aerate the aeration tank. Furthermore, this blower is controlled using constant dissolved oxygen concentration control, which aims to maintain a constant dissolved oxygen concentration in the activated sludge in the aeration tank (hereinafter simply referred to as "dissolved oxygen concentration"). Therefore, the control target value for the blower determined by the sewage treatment support system of this embodiment is the dissolved oxygen concentration value in the activated sludge in the aeration tank.

This blower control can also be performed using constant inflow ratio control, which aims to maintain a constant inflow ratio, which represents the ratio of the blower's discharge air volume to the amount of sewage inflow into the sewage treatment plant. In this case, the blower control target value determined by the sewage treatment support system is the inflow ratio.

The sewage treatment support system of this embodiment determines the blower control target value based on weekly data (hereinafter referred to as "status data") representing the condition of activated sludge in the aeration tank. This status data includes data (hereinafter referred to as "water quality data") representing the weekly water quality of water after secondary treatment at the sewage treatment plant (hereinafter referred to as "secondary treated water") and data (hereinafter referred to as "operating data") representing the hourly operating status of various equipment at the sewage treatment plant, including the blower. Here, the quality of the secondary treated water reflects the condition of the activated sludge in the aeration tank, such as the level of nitrification activity. As mentioned above, the level of nitrification activity of the activated sludge is influenced by indicators such as viable cell count and bacterial cell concentration, which represent the growth status of various nitrifying bacteria in the activated sludge. Therefore, water quality testing to obtain new water quality data is typically conducted weekly, taking into account the growth rate of microorganisms in the activated sludge, including nitrifying bacteria. The water quality data in this embodiment also represents the results of water quality tests conducted weekly.

When setting the blower's control target value, it is necessary to take into account a time lag of about one week after obtaining the results of weekly water quality tests and set the control target value for the following week's operation. However, because the water quality of the secondary effluent changes during that time, deriving the control target value directly from water quality data or status data including water quality data makes it difficult to set an appropriate control target value for the activated sludge state during operation. Furthermore, since the measured values contained in the water quality data and operational data change each time they are measured and are never exactly the same, it is difficult to extract general trends from these measured values and build a learning model. Therefore, the sewage treatment support system of this embodiment clusters the weekly status data to generate clusters containing multiple status data. Here, multiple status data contained in the same cluster each represent the activated sludge state for the week the blower was operated under the same or similar conditions. In other words, each cluster represents the activated sludge state for a period when the same or similar control target values were set. The sewage treatment support system of this embodiment focuses on the transitions between clusters generated by clustering that occur as the blower's operating conditions change, and by learning the conditions for transitions between clusters when favorable cluster transitions occur (hereinafter referred to as "transition conditions"), it determines the blower's control target values appropriate for the activated sludge state during operation. This makes it possible to set control target values appropriate for the activated sludge state during blower operation based on the results of water quality testing conducted approximately one week before the blower starts operating.

Figure 1 shows an example configuration of a sewage treatment support system related to this embodiment.

Each component of the sewage treatment support system 100 shown in Figure 1 is realized by hardware including processor devices such as a CPU (Central Processing Unit) and various co-processors (hereinafter simply referred to as "processors"), storage devices such as memory and storage, wired or wireless communication lines and interface devices that connect them, and software that is stored in the storage devices and supplies processing instructions to the computing units.

The storage device stores at least a sewage treatment support program. The sewage treatment support program is a computer program for supporting sewage treatment. When the sewage treatment support program is executed by the processor, processes such as transition condition learning processing, control target value determination processing, and transition destination estimation processing are performed. Details of these processes are described below.

The sewage treatment support program may also be composed of device drivers, an operating system, various application programs located at higher levels, and libraries that provide common functions to these programs. Each block described below represents a functional block, rather than a hardware configuration.

The sewage treatment support system 100 has the following functional blocks: a data processing unit 10, a data storage unit 30, a user interface unit (not shown), and a communication unit (not shown).

The data processing unit 10 performs various data processing operations based on user input detected by the user interface unit, data acquired by the communication unit, and programs and data stored in the data storage unit 30. The data processing unit 10 also functions as an interface between the user interface unit, communication unit, and data storage unit 30.

The data processing unit 10 has the functional blocks of a condition learning unit 120 and an operation planning unit 140. The condition learning unit 120 performs various processes for building a machine learning model. The operation planning unit 140 performs various processes for controlling the blower.

The condition learning unit 120 includes a clustering unit 220 and a transition condition learning unit 240. The operation planning unit 140 also includes a control target value determination unit 420 and a power consumption estimation unit 440.

The clustering unit 220 performs clustering processing on status data for a specified period that represents the status of activated sludge in the aeration tank of a sewage treatment plant that performs sewage treatment based on the standard activated sludge method.

The transition condition learning unit 240 executes a process to learn the transition conditions between multiple clusters generated by clustering. This process executed by the transition condition learning unit 240 is referred to as the transition condition learning process. In this transition condition learning process, the transition condition learning unit 240 outputs a model that learns the transition of clusters from one week to the next when one or more values including a control target value among multiple values are input. This model uses a pair of cluster numbers that transitioned between one week and the next week as the objective variable, and one or more values as the explanatory variable. Details of the transition condition learning process will be described later in relation to Figures 2 and 3.

The control target value determination unit 420 executes a process to determine a control target value for the blower that discharges air to aerate the aeration tank, based on the learned transition conditions and a time series of new status data representing the status of activated sludge in the aeration tank. This process executed by the control target value determination unit 420 is referred to as the control target value determination process. Details of the control target value determination process will be described later in relation to Figure 10.

Note that the above-described control target value determination process includes a transition destination estimation process. Details of the transition destination estimation process will be described later in relation to Figure 11.

The power consumption estimation unit 440 executes a process to estimate the power consumption of the blower based on the amount of air discharged by the blower. This allows the reduction of power consumption to be taken into consideration when determining the blower's control target value, without having to separately measure the amount of power consumed by the blower. Note that there is a roughly positive correlation between the amount of air discharged by the blower and the amount of power consumed by the blower.

The data processing unit 10 can realize these functional blocks by executing a specified program.

The data storage unit 30 is configured using a storage device such as RAM or flash memory, and stores in the storage device programs that supply various processing commands to the data processing unit 10, as well as data representing various information used in the processing performed by the data processing unit 10. For example, water quality data stored in the water quality data storage unit 320 and operating data stored in the operating data storage unit 340 are stored in the storage device by the data storage unit 30. By reading and writing this information to the data storage unit 30, the data processing unit 10 can realize the functional blocks of the clustering unit 220, transition condition learning unit 240, control target value determination unit 420, and power consumption estimation unit 440 described above.

The data storage unit 30 has the functional blocks of a water quality data storage unit 320 and an operating data storage unit 340.

The water quality data storage unit 320 mainly stores water quality data represented by the water quality table illustrated in Figure 4. The water quality table is a table for managing water quality data. Details of the water quality data and the water quality table will be described later in connection with Figures 3 and 4.

The driving data storage unit 340 mainly stores driving data represented by the driving table illustrated in Figure 5. The driving table is a table for managing driving data. Details of the driving data and driving table will be described later with reference to Figures 3 and 5.

The user interface unit accepts input operations from the user and is responsible for user interface-related processing such as image display and audio output. The user interface unit has the functional blocks of an input unit and an output unit. The input unit detects various operations from the user. The input unit is configured using, for example, a keyboard, pointing device, touch panel, etc. The output unit displays the determined blower control target values on the display device 50 and also displays various screens and outputs audio on the display device 50. The display device 50 is configured using, for example, an LCD display, touch screen, etc.

The communications unit is responsible for processing communications via the Internet (an example of a communications network) with user terminals owned by each user of the sewage treatment support system 100 and other devices, such as the server devices that make up the sewage treatment plant monitoring and control system 70. The monitoring and control system 70 is a system that monitors and controls the operating status of various pieces of equipment at the sewage treatment plant, including blowers. The communications unit is configured using, for example, a NIC (Network Interface Card) or an HBA (Host Bus Adapter).

In this embodiment, the functions of the sewage treatment support system 100 have been described as being implemented integrally by a single computer device. However, these functions may also be implemented by multiple interconnected computer devices or server devices. Furthermore, the sewage treatment support system 100 may be configured to include a general-purpose computer device such as a laptop PC and a web browser installed on it, or may be configured to include a web server and various mobile devices.

Next, we will explain the flow of each process performed by the sewage treatment support system 100.

Figure 2 is a flowchart showing an example of the overall processing flow performed in this embodiment.

In step S210, the data processing unit 10 executes a process in which the transition condition learning unit 240 determines whether or not re-learning of the transition conditions between clusters is necessary. In this embodiment, the sewage treatment support system 100 executes each process from step S220 onward when a water quality test is conducted on the secondary treated water and new water quality data representing the test results is stored in the water quality data storage unit 320—in other words, when a new record representing the test results is recorded in the water quality table described below. Therefore, the determination of whether or not re-learning is necessary is made, for example, by determining whether or not the water quality data storage unit 320 has stored new water quality data. If it is determined that re-learning is necessary (step S210: Y), the process proceeds to step S220 to re-learn the transition conditions between clusters. If it is determined that re-learning is not necessary (step S210: N), the process proceeds directly to step S240.

In step S220, the data processing unit 10 executes a transition condition learning process using the transition condition learning unit 240. This allows the control target value determination process in step S230 to be executed with high accuracy. When the transition condition learning process is complete, the data processing unit 10 proceeds to step S230.

In step S230, the data processing unit 10 executes a control target value determination process using the control target value determination unit 420. This control target value determination process is performed based on the results of the transition condition learning process in step S220. As a result, an appropriate blower control target value is automatically determined. When the control target value determination process is complete, the data processing unit 10 proceeds to step S240.

In step S240, if the data processing unit 10 determines in step S210 that re-learning of the transition conditions is necessary (step S210: Y), it causes the display device 50 to display a screen showing the results of the control target value determination process currently executed in step S230 via the output unit. On the other hand, if the data processing unit 10 determines in step S210 that re-learning of the transition conditions is not necessary (step S210: N), it causes the display device 50 to display a screen showing the results of the control target value determination process previously executed in step S230 via the output unit. The output unit displays the determined blower control target value, which is the result of the control target value determination process, on the display device 50 in the manner of the display screens exemplified in Figures 13A and 13B. When the data processing unit 10 completes the processing in step S240, it terminates the processing shown in the flowchart of Figure 2.

In this way, the sewage treatment support system 100 of this embodiment executes the processes of steps S210 to S240 in FIG. 2 to determine favorable blower control target values and displays them on the display device 50. The processes for favorably determining blower control target values can be broadly divided into a transition condition learning process executed by the transition condition learning unit 240 in step S220, and a control target value determination process executed by the control target value determination unit 420 in step S230. Therefore, details of the transition condition learning process executed in step S220 will first be explained with reference to the flowchart in FIG. 3. Details of the control target value determination process executed in step S230 will be explained later with reference to the flowchart in FIG. 10.

Figure 3 shows an example of the flow of the transition condition learning process.

In step S301, the data processing unit 10 acquires one of the most recent pieces of water quality data stored in the water quality data storage unit 320 in the water quality table. An example of the configuration of this water quality table is shown in Figure 4. As mentioned above, water quality tests to acquire water quality data are conducted weekly. Therefore, the water quality table 400 has a record for each week on which a water quality test was conducted. The record includes as items at least the date on which the water quality test was conducted (hereinafter referred to as the "measurement date") and the biochemical oxygen demand (BOD) of the secondary treated water on that measurement date. The records may also include measurements of water quality test items in effluent quality standards that set standards for the water quality of effluent from sewage treatment plants (hereinafter simply referred to as "effluent quality"), such as the chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) of the secondary treated water on the measurement date. That is, the water quality table 400 represents the results of weekly water quality tests by linking the measurement date with each measurement value representing the water quality of the secondary treated water and the effluent water on that measurement date for each record. In the example shown in FIG. 4 , the water quality table 400 records that the biochemical oxygen demand (BOD) of the secondary treated water was 4.2 mg/L and the chemical oxygen demand was 7.2 mg/L in a water quality test conducted on April 8, 2020. When the processing in step S301 is completed, the data processing unit 10 proceeds to step S302.

In step S302, the data processing unit 10 acquires the operation data stored by the operation data storage unit 340 in the operation table. An example of the configuration of this operation table is shown in FIG. 5 . As described above, the operation data is the hourly operating results of various pieces of equipment in the sewage treatment plant, including the blower, recorded by the monitoring and control system 70. Therefore, the operation table 500 has a record for each hour. The record includes, as items, at least the date and time when the monitoring and control system 70 recorded the operation data, the amount of air discharged from the blower from the previous recording date and time of the operation data to the current date and time, the dissolved oxygen concentration in the activated sludge in the aeration tank at the current date and time, the amount of sludge returned to the aeration tank at the current date and time (hereinafter referred to as the "returned sludge amount"), and the amount of sewage inflow into the sewage treatment plant at the current date and time. The records may also include values such as the concentration of activated sludge (mixed liquor suspended solids; MLSS) in the aeration tank, measured values of water quality test items such as total phosphorus and total nitrogen, the amount of polyaluminum chloride (PAC) injected, and the inflow amount of nitrification liquid. That is, the operation table 500 records, for each record, the date and time and the values of each indicator representing the operating performance of the sewage treatment plant at that date and time, thereby representing the hourly operating status of various equipment of the sewage treatment plant, including the blower. According to the example shown in FIG. 5 , the operation table 500 records that at "00:00:00 on April 1, 2020," the blower discharge air volume was "20,000 m ³ ," and the dissolved oxygen concentration in the activated sludge in the aeration tank was "1.3 mg/L." In order to match the time unit of the water quality data in the processing of step S303 described below, the data processing unit 10 acquires the operating data for the most recent week and converts it into a weekly average in the processing of step S302. After completing the processing of step S302, the data processing unit 10 proceeds to step S303.

In step S303, the data processing unit 10 generates status data based on the latest water quality data acquired in step S301 and the operating data representing the weekly average for the most recent week acquired in step S302. The generated status data is stored in the data storage unit 30 in a status table. An example of the configuration of this status table is shown in FIG. 6. As described above, the status data includes water quality data representing the results of weekly water quality tests. Therefore, the status table 600 has a record for each period equivalent to one week. The record represents the period equivalent to one week, the average daily blower discharge air volume for the week represented by the period (hereinafter referred to as the "week"), the average hourly dissolved oxygen concentration in the activated sludge for the week, the average daily volume of returned sludge to the aeration tank for the week, the average daily volume of sewage inflow to the sewage treatment plant for the week, and the average daily biochemical oxygen demand of the secondary effluent for the week. Of these, the daily average value of the biochemical oxygen demand of the secondary effluent is derived from water quality data, while the other values are derived from operational data. That is, the status table 600 represents the weekly state of the activated sludge in the aeration tank by linking a one-week period with the water quality data and operational data values for that week for each record. For example, the data processing unit 10 generates new status data by storing water quality data representing the results of a water quality test conducted on April 8 in a record that stores operational data representing the weekly average for the week from April 1 to April 7. In the example shown in FIG. 6 , the status table 600 records that, for the week from April 1 to April 7, 2020, the daily average value of the blower's discharge air volume was 520,000 ^{m 3} and the hourly average value of the dissolved oxygen concentration in the activated sludge was 1.5 mg/L. After completing the processing in step S303, the data processing unit 10 proceeds to step S304.

In step S304, the data processing unit 10 causes the clustering unit 220 to perform clustering processing on the status data generated in step S303. The clustering processing is performed by generating a matrix with the values of each row of the status data as elements and applying various clustering methods to this matrix as input. This results in clustering of the weekly status data. Status data classified into one of the multiple clusters generated by clustering is referred to as post-clustered status data. In this embodiment, clustering is performed using the k-means method. The status data is classified into one of eight clusters. For convenience, eight clusters are used in this embodiment, but the user can specify any number of clusters in advance. The post-clustered status data is stored in the data storage unit 30 in a post-clustered status table. An example of the configuration of this post-clustered status table is shown in FIG. 7. Like the status table 600 shown in FIG. 6, the post-clustered status table 700 shown in FIG. 7 also has a record for each period equivalent to one week. Each record represents a period equivalent to one week, the daily average value of the blower's discharge air volume for that week, the hourly average value of the dissolved oxygen concentration in the activated sludge for that week, the daily average value of the return sludge volume to the aeration tank for that week, the daily average value of the sewage inflow volume to the sewage treatment plant for that week, the daily average value of the biochemical oxygen demand of the secondary effluent for that week, and a cluster number for uniquely identifying the cluster into which the status data is classified. That is, the clustered status table 700 is a table in which a cluster number is assigned to each record of the status table 600 illustrated in FIG. 6. According to the example illustrated in FIG. 7, the clustered status table 700 records that the status data for the "week from April 1 to April 7, 2020," which indicates that the daily average value of the blower's discharge air volume was "520,000 ^m3 " and the hourly average value of the dissolved oxygen concentration in the activated sludge was "1.5 mg/L," is classified into a cluster assigned a cluster number of "1." It should be noted that the clusters may be identified using other identification codes such as alphabets, symbols, etc., instead of the cluster numbers described above. After completing the process in step S304, the data processing unit 10 proceeds to step S305.

In step S305, the data processing unit 10, using the transition condition learning unit 240, generates a transition condition learning matrix (hereinafter simply referred to as the "learning matrix") as illustrated in Figure 9 based on the clustered state data. This learning matrix represents the explanatory variables and target variables in the transition condition learning process as a matrix. Among these, the matrix representing the explanatory variables represents the cluster number of the cluster to which the state data belongs, the average daily blower discharge air volume for the previous week, the average hourly dissolved oxygen concentration in the activated sludge for the previous week, the average daily returned sludge volume to the aeration tank for the previous week, the average daily sewage inflow volume to the sewage treatment plant for the previous week, the average daily biochemical oxygen demand of the secondary effluent for the previous week, and the average hourly dissolved oxygen concentration in the activated sludge for the week following the current week (DO at the next state). The matrix representing the dependent variable is composed of pairs of cluster numbers (equivalent to the combined value of each number treated as a string) that correspond to the cluster number belonging to a certain week and the cluster number belonging to the following week. For example, if data belonged to cluster i in the previous week and cluster j in the following week, the value constituting the dependent variable matrix is ij. The learning matrix illustrated in FIG. 9 represents the actual cluster transitions that occurred when the measured values obtained during a certain week of operation were recorded. In this way, the transition condition learning unit 240 constructs explanatory variables as a learning matrix including the amount of air discharged by the blower in the previous week, the dissolved oxygen concentration in the activated sludge in the aeration tank in the previous week, the amount of returned sludge to the aeration tank in the previous week, the amount of sewage inflow to the sewage treatment plant in the previous week, the biochemical oxygen demand of the secondary effluent from the sewage treatment plant, and the predicted value of the dissolved oxygen concentration in the activated sludge in the aeration tank in the following week. When the processing in step S305 is completed, the data processing unit 10 proceeds to step S306.

In step S306, the data processing unit 10 executes a process for learning transition conditions between clusters using the transition condition learning unit 240. The transition conditions are learned by calculating the angle of the centroid vector representing the transition direction and the time series before and after for two clusters with adjacent time series, and learning the conditions under which the angle of the centroid vector is equal to or less than a predetermined threshold. If no training data selection is performed, the angle threshold can be set arbitrarily. The centroid of each cluster may be calculated, for example, when applying the k-means method, and stored in memory for use. In this embodiment, this process is performed using a decision tree. However, the transition condition learning process in step S306 may also be performed using other machine learning methods capable of classification, such as a support vector machine (SVM). After completing the process in step S306, the data processing unit 10 proceeds to step S307.

In step S307, the data processing unit 10 outputs a learned model using the transition condition learning unit 240. The learned model (projection of the input matrix x onto y) is saved in memory. When the processing in step S307 is completed, the data processing unit 10 ends the transition condition learning process shown in the flowchart in Figure 3.

In the transition condition learning process, the transition condition learning unit 240 may calculate the angle between the center of gravity vector of the transitioned path and a vector of a good transition derived in advance, and learn only transitions represented by an angle below a predetermined threshold. In addition, at this time, the transition condition learning unit 240 may exclude data that shows a bad transition. This makes it possible to select states taking into account medium-term transitions and avoid learning transitions that are good in the short term but take a bad turn in the medium term. Details of the transition condition learning process for such cases will be explained later with reference to the flowchart in Figure 14.

In this embodiment, the sewage treatment support system 100 inputs newly generated status data at the time of water quality testing. It then calculates the cluster with the smallest distance between the center of gravity and the status data, identifying the current cluster representing the sewage treatment plant's current treatment performance. It then searches for a dissolved oxygen concentration value that will allow the plant to either move to a better cluster or remain in the current cluster, determines it as the blower control target value, and presents it to the user. In this process, explanatory variables for the transition conditions are constructed from the newly generated status data and the dissolved oxygen concentration provisionally set during the search process, and estimates which cluster the plant will transition to based on the provisionally set dissolved oxygen concentration (i.e., the control target value for the next week). The optimal control target value for the next week is determined while confirming whether the destination cluster satisfies at least one of minimizing the blower's air volume or improving water quality. This makes it possible to estimate the sewage treatment performance of a sewage treatment plant and predict whether the next water quality improvement will be necessary, based on a variety of easily available, general measurement results contained in water quality data and operational data, without using activated sludge models (ASMs), which are vulnerable to external disturbances, or without procuring a large number of expensive measuring instruments. In other words, each cluster represents the sewage treatment performance of the sewage treatment plant.

FIG. 8 shows an example of transitions between clusters. FIG. 8 shows post-clustering status data for each week, along with cluster numbers, on a scatter plot with the average daily value of the blower's discharge air volume for each week (m ³ /day/week) on the horizontal axis and the average daily value of the secondary treated water's biochemical oxygen demand for that week (mg/L) on the vertical axis. The arrows indicate movement from a measurement point in one week to a measurement point in the following week. The points representing post-clustering status data on the scatter plot transition depending on the blower's discharge air volume. Essentially, increasing the blower's discharge air volume causes the points representing the post-clustering status data to transition to a point where the secondary treated water's biochemical oxygen demand value becomes lower. On the other hand, decreasing the blower's discharge air volume causes the points representing the post-clustering status data to transition to a point where the secondary treated water's biochemical oxygen demand value becomes higher. Therefore, the sewage treatment support system 100 assumes that the weekly status data represents the status of the activated sludge in the aeration tank for that week, and generates a transition destination estimator by learning an operating method that will improve the status or maintain the current status through a transition condition learning process.Then, in the control target value determination process, the generated transition destination estimator is used to estimate a dissolved oxygen concentration value that will move toward a better cluster, and this is determined as the control target value for the blower and presented to a user of the sewage treatment support system 100 (e.g., a blower operator).

Figure 10 shows an example of the flow of the control target value determination process.

In step S1001, the data processing unit 10 generates post-clustering state data using the control target value determination unit 420. In this process in step S1001, the control target value determination unit 420 performs the same processing as steps S1101 to S1104, described below, to identify the cluster of the latest value. When the data processing unit 10 completes the processing in step S1001, it proceeds to step S1002.

In step S1002, the data processing unit 10 constructs an optimization model for determining the optimal control target value using the control target value determination unit 420. This optimization model uses the dissolved oxygen concentration value, which is the blower control target value for the following week, as the decision variable. To search for this decision variable, an optimization algorithm that can be applied without using derivatives, such as the Nelder-Mead method, is used. The objective variable of this optimization model is expressed by the following (Equation 1).

minimize w1・a1+w2(max(0,b1-b0))...(Formula 1)
w1, w2: weights (any real number)
a1: Value of air volume at the center of gravity of the cluster after the transition b1: Value of biochemical oxygen demand at the center of gravity of the cluster after the transition b0: Latest value of biochemical oxygen demand

In the above (Equation 1), if b1 is greater than b0, it is considered that the water quality has deteriorated, so b0 is the penalty term. After completing the processing in step S1002, the data processing unit 10 proceeds to step S1003.

In step S1003, the data processing unit 10 sets the dissolved oxygen concentration value to be the blower control target value for the following week as the decision variable using the control target value determination unit 420. After completing the processing in step S1003, the data processing unit 10 proceeds to step S1004.

In step S1004, the data processing unit 10 executes the transition destination estimation process, described later in relation to FIG. 11, using the control target value determination unit 420. The results of the transition destination estimation process are stored in memory, for example, in the table format shown in FIG. 10. In the table shown in FIG. 10, x is an explanatory variable in the transition destination estimation process and represents multiple values contained in the clustered state data for each specified period, represented as a transition destination estimation matrix (hereinafter referred to as the "transition destination estimation matrix") shown in FIG. 12. These multiple values include the latest dissolved oxygen concentration value, which is the control target value immediately prior to the specified period, and the predicted dissolved oxygen concentration value, which is the control target value for the specified period. y represents the estimation result as a two-digit number consisting of the cluster number of the transition source cluster and the cluster number of the transition destination cluster. In addition, the table also records the loop count, which indicates the number of times the optimization process has been repeated. Among the components of x, the value of the dissolved oxygen concentration for the next week is a decision variable and therefore changes each time the loop process is performed. The amount of change in the dissolved oxygen concentration value at this time depends on the applied optimization algorithm. Note that the effluent water quality does not necessarily need to be improved or maintained. As mentioned above, the blower's discharge air volume is generally positively correlated with power consumption. Therefore, for example, if the goal is to reduce power consumption, the blower's discharge air volume may be reduced even if it worsens the effluent water quality. The effluent water quality needs to be within the effluent water quality standards; specifically, the biochemical oxygen demand value does not need to exceed the standard value. The priority between effluent water quality and power consumption reduction is determined by how the values of w1 and w2 are set in the above (Equation 1). After completing the transition destination estimation process in step S1004, the data processing unit 10 proceeds to step S1005.

In step S1005, the data processing unit 10 calculates the objective function using the control target value determination unit 420. At this time, the data processing unit 10 may use the power consumption estimation unit 440 to estimate the power consumption of the blower from the air discharge volume of the blower based on the correlation between the air discharge volume of the blower and the power consumption. This allows power consumption reduction to be taken into consideration when determining the control target value of the blower, without having to separately measure the power consumption of the blower. When the data processing unit 10 completes the processing in step S1005, it proceeds to step S1006.

In step S1006, the data processing unit 10 determines whether the optimization has converged using the control target value determination unit 420. If it is determined that the optimization has converged (step S1006: Y), the process proceeds directly to step S1007. If it is determined that the optimization has not converged (step S1006: N), the process returns to step S1003 to reset the decision variables.

In step S1007, the data processing unit 10 causes the control target value determination unit 420 to output the optimized decision variables as the dissolved oxygen concentration value that will become the blower's control target value for the following week. At this time, the control target value determination unit 420 also outputs the improved values for the blower discharge air volume, which represents the center of gravity of the cluster after the transition, the blower's power consumption, and the biochemical oxygen supply volume, which is the distance between the centers of gravity of the clusters, as shown in FIG. 10. The output unit displays these values output by the control target value determination unit 420 on the display device 50, for example, in the manner of the display screen shown in FIG. 13B. When the data processing unit 10 completes the processing in step S1007, it terminates the control target value determination processing shown in the flowchart of FIG. 10.

Figure 11 shows an example of the flow of the transition destination estimation process.

In step S1101, the data processing unit 10 acquires the water quality data stored in the water quality table 400 by the water quality data storage unit 320. Note that this processing in step S1101 is the same as the processing in step S301 in Figure 3. When the processing in step S1101 is completed, the data processing unit 10 proceeds to step S1102.

In step S1102, the data processing unit 10 acquires the driving data stored by the driving data storage unit 340 in the driving table 500. Note that this processing in step S1102 is the same as the processing in step S302 in Figure 3. When the processing in step S1102 is completed, the data processing unit 10 proceeds to step S1103.

In step S1103, the data processing unit 10 generates status data as a one-row matrix consisting of the latest values based on the water quality data acquired in step S1101 and the operating data acquired in step S1102. The generated status data is stored by the data storage unit 30 as a new record in the status table 600. Note that this processing in step S1103 is the same as the processing in step S303 in Figure 3. However, when step S1004 is called, the value provisionally set in the optimization process is substituted for the control target value for the next week, which is one of the explanatory variables used to estimate the transition destination. When the processing in step S1103 is completed, the data processing unit 10 proceeds to step S1104.

In step S1104, the data processing unit 10 uses the control target value determination unit 420 to identify the cluster after the transition. This process is performed by estimating that the cluster whose center of gravity, determined in the process of step S306 in FIG. 3, is closest to the latest value vector is the cluster after the transition. The control target value determination unit 420 assigns a cluster number representing the cluster identified as the destination of the transition, and generates post-clustering state data as a one-row matrix consisting of the latest value and cluster number. When the process of step S1104 is completed, the data processing unit 10 proceeds to step S1105.

In step S1105, the data processing unit 10 causes the control target value determination unit 420 to generate a transition destination estimation matrix, as illustrated in FIG. 12, based on the post-clustering state data generated in step S1104. This matrix represents the explanatory variable X in the transition destination estimation process. Note that the arrangement of the elements of the transition destination estimation matrix is the same as the arrangement of the elements of the learning matrix generated in step S305 of FIG. 3. When the processing in step S1105 is completed, the data processing unit 10 proceeds to step S1106.

In step S1106, the data processing unit 10 causes the control target value determination unit 420 to execute a process of estimating the cluster after the transition. This process is executed by providing the transition destination estimation matrix generated in step S1105 as input to the learning model output in step S307 of FIG. 3. When the process in step S1106 is completed, the data processing unit 10 proceeds to step S1107.

In step S1107, the data processing unit 10 outputs the estimated post-transition cluster number to memory using the control target value determination unit 420. Upon completing the processing in step S1107, the data processing unit 10 terminates the transition destination estimation processing shown in the flowchart of FIG. 11.

As mentioned above, the output unit displays the determined control target value for the blower, which is the result of the control target value determination process, on the display device 50 in the manner of the display screens exemplified in Figures 13A and 13B. In this case, the output unit may, for example, display on the display device 50 past history including the most recent measurement results, cluster transition history, and the process for determining the dissolved oxygen concentration that is the control target value, as shown in Figures 13A and 13B. The output unit may also, for example, illustrate the final dissolved oxygen concentration setting value and the resulting cluster to which it transitions, as shown in Figure 13B.

The transition condition learning unit 240 may also generate the learning matrix as an average vector for the past period T. This allows the sewage treatment support system 100 to learn transitions over multiple periods.

Figure 14 shows an example of the flow of the transition condition learning process performed by the transition condition learning unit 240 in such a case.

In step S1401, the data processing unit 10 generates status data based on the water quality data acquired from the water quality data storage unit 320 and the operating data acquired from the operating data storage unit 340. The generated status data is stored by the data storage unit 30 in the status table 600. Note that this processing in step S1401 is the same as the processing in steps S301 to S303 in Figure 3 and the processing in steps S1101 to S1103 in Figure 11. When the processing in step S1401 is completed, the data processing unit 10 proceeds to step S1402.

In step S1402, the data processing unit 10 sets a period T for determining a medium-term direction using the transition condition learning unit 240. If this period T is, for example, one month, it is set to T = 4. The transition condition learning unit 240 may accept an input operation for the period T from the user via the communication unit or input unit. When the processing in step S1402 is completed, the data processing unit 10 proceeds to step S1403.

In step S1403, the data processing unit 10 uses the transition condition learning unit 240 to generate a mean matrix X for transition condition learning (hereinafter referred to as the "learning mean matrix") in the manner illustrated in FIG. 15. This learning mean matrix X is calculated by calculating the mean vector of a past period T based on a certain date, and uses these values as components. The learning mean matrix X illustrated in FIG. 15 is for T=4, which indicates that the period T is one month. When the data processing unit 10 completes the processing in step S1403, it proceeds to step S1404.

In step S1404, the data processing unit 10 performs clustering using the transition condition learning unit 240 to generate post-clustering state data. The processing performed here is the same as the processing performed in step S304 in Figure 3. When the processing in step S1404 is completed, the data processing unit 10 proceeds to step S1405.

In step S1405, the data processing unit 10 generates a matrix for medium-term learning using the transition condition learning unit 240. The elements of this matrix are the same as the elements of the learning matrix illustrated in FIG. 9. The values of this matrix are made up of the mean vector calculated in the processing of step S1403. When the processing of step S1405 is completed, the data processing unit 10 proceeds to step S1406.

In step S1406, the data processing unit 10 executes learning processing using the transition condition learning unit 240. The content of the processing executed here is the same as the processing executed in step S306 of Figure 3. In other words, this learns the transition destination for the next period T from the operating state of the blower during the past period T. When the data processing unit 10 completes the processing in step S1406, it proceeds to step S1407.

In step S1407, the data processing unit 10 outputs a model using the transition condition learning unit 240. The processing performed here is the same as the processing performed in step S307 in Figure 3. Upon completing the processing in step S1407, the data processing unit 10 terminates the transition condition learning processing shown in the flowchart in Figure 14.

In the above embodiment, the blower control method was described as constant dissolved oxygen concentration control, and the blower control target value determined by the sewage treatment support system 100 was the dissolved oxygen concentration value in the activated sludge in the aeration tank. However, as mentioned above, the blower can also be controlled by constant inflow ratio control. When constant inflow ratio control is used as the blower control method, the blower control target value determined by the sewage treatment support system 100 is the inflow ratio. In this case, the data processing unit 10 of the sewage treatment support system 100 calculates the inflow ratio as appropriate based on, for example, the value of the sewage inflow rate to the sewage treatment plant and the value of the blower discharge air volume contained in the operating data.

The above-described embodiment of the present invention provides the following advantages:

(1) The sewage treatment support system 100 includes a clustering unit 220, a transition condition learning unit 240, and a control target value determination unit 420. The clustering unit 220 performs clustering of status data for a predetermined period that represents the state of activated sludge in the aeration tank of a sewage treatment plant that performs sewage treatment based on a standard activated sludge process. The transition condition learning unit 240 learns the conditions for transitions between clusters among multiple clusters generated by clustering. The control target value determination unit 420 determines the control target value of a blower that discharges air to aerate the aeration tank based on the learned transition conditions and a time series of new status data that represents the state of activated sludge in the aeration tank. The status data includes water quality data that represents the water quality of the secondary effluent from the sewage treatment plant and operating data that represents the operating status of the sewage treatment plant. As a result, the sewage treatment support system 100 can automatically determine an optimal control target value for the blower.

(2) The control target value is at least one of the following:
- Dissolved oxygen concentration in activated sludge in the aeration tank,
- Inflow ratio, which represents the ratio of the blower's discharge air volume to the amount of sewage inflow into the sewage treatment plant.
This allows the blower to perform constant dissolved oxygen concentration control or constant inflow rate control.

(3) The water quality data includes the biochemical oxygen demand of the secondary effluent from the sewage treatment plant as an indicator. This allows the sewage treatment support system 100 to successfully execute the transition condition learning process and the control target value determination process.

(4) The operating data includes, as indicators, the amount of sewage inflow into the sewage treatment plant, the dissolved oxygen concentration in the activated sludge in the aeration tank, the amount of sludge returned to the aeration tank, and the amount of air discharged by the blower. This allows the sewage treatment support system 100 to successfully execute the transition condition learning process and the control target value determination process.

(5) The status data for each specified period includes multiple values, and the multiple values in the status data for each specified period include the control target value immediately before the specified period and the control target value for the specified period. The transition condition learning unit 240 outputs a combination of cluster transitions from one week to the next week and learns a model in which one or more values including the control target value from the multiple values are input. In this way, the sewage treatment support system 100 can estimate the transition destination when a certain control target value is assumed, based on the results of executing the transition condition learning process.

(6) The model uses the combination of cluster transitions from one week to the next as the objective variable and one or more values as the explanatory variables. As a result, the sewage treatment support system 100 can estimate possible cluster transitions based on the results of the transition condition learning process.

(7) The transition condition learning unit 240 calculates the angle of the centroid vector representing the transition direction and the time series before and after for two clusters with adjacent time series, and learns the transition condition when the centroid vector has an angle equal to or less than a predetermined threshold. In this way, the sewage treatment support system 100 can estimate the strength of the nitrification ability of the aeration tank, i.e., the sewage treatment performance of the sewage treatment plant, based on fragmentary information.

(8) The specified period is one week. This allows the sewage treatment support system 100 to effectively support sewage treatment by taking into account the growth rate of microorganisms in the aeration tank.

(9) The transition condition learning unit 240 constructs explanatory variables as a learning matrix including the amount of air discharged by the blower in the previous week, the dissolved oxygen concentration in the activated sludge in the aeration tank in the previous week, the amount of sludge returned to the aeration tank in the previous week, the amount of sewage inflow to the sewage treatment plant in the previous week, the biochemical oxygen demand of the secondary effluent from the sewage treatment plant, and the predicted value of the dissolved oxygen concentration in the activated sludge in the aeration tank in the following week. In this way, the sewage treatment support system 100 can successfully execute the transition condition learning process and determine an appropriate blower control target value based on the execution results.

(10) The transition condition learning unit 240 generates the learning matrix as the average vector of the past period T. In this way, the sewage treatment support system 100 can learn transitions over multiple periods.

(11) The system further includes an output unit that displays the determined blower control target value on the display device 50. In this way, the sewage treatment support system 100 can present the determined blower control target value to the user as visual information.

(12) The system further includes a power consumption estimation unit 440 that estimates the power consumption of the blower based on the amount of air discharged by the blower. As a result, the sewage treatment support system 100 can take power consumption reduction into consideration when determining the blower's control target value, without having to separately measure the amount of power consumed by the blower.

(13) The control target value determination unit identifies a latest location cluster, which is a cluster in which the latest state data is classified, and estimates a cluster to which the latest location cluster will transition based on the latest location cluster, the latest state data, and the control target value for the next specified period. By repeating the process including the following, the control target value is searched for such that the cluster to which the latest location cluster will transition will be a cluster that reduces the amount of air discharged by the blower or improves the quality of secondary treated water from the sewage treatment plant.
- Changing the latest control target value for the next predetermined period based on the learned transition condition.
- Estimating a cluster to which the latest position cluster will transition based on the latest changed control target value for the next predetermined period;
The found control target value is output. Note that this found control target value is the control target value at the time of optimization convergence. What is achieved up to the transition condition learning stage is that when the current cluster, operating data, and next week's control target value are input, a model is obtained that outputs (can estimate) the destination cluster. In addition, by further incorporating this configuration, the sewage treatment support system 100 can specifically determine the blower control target value for next week that will allow transition to a good cluster.

Note that the present invention is not limited to the above-described embodiment, and can be implemented using any components within the scope of the present invention.

The above embodiments and modifications are merely examples, and the present invention is not limited to these details as long as the characteristics of the invention are not impaired. Furthermore, while various embodiments and modifications have been described above, the present invention is not limited to these details. Other aspects that can be considered within the technical scope of the present invention are also included within the scope of the present invention.

100: Sewage Treatment Support System

Claims (14)

a clustering unit that performs clustering of status data for each predetermined period that represents the status of activated sludge in an aeration tank of a sewage treatment plant that performs sewage treatment based on a standard activated sludge method;
a transition condition learning unit that learns conditions for transitions between clusters among the plurality of clusters generated by the clustering;
a control target value determination unit that determines a control target value of a blower that discharges air for aerating the inside of the aeration tank based on the learned transition condition and a time series of new status data that represents a status of the activated sludge in the aeration tank,
The status data includes water quality data representing the quality of secondary treated water from the sewage treatment plant and operation data representing the operating status of the sewage treatment plant,
The operating data includes, as indicators, an amount of sewage inflow into the sewage treatment plant, a dissolved oxygen concentration in the activated sludge in the aeration tank, an amount of sludge returned to the aeration tank, and an amount of air discharged by the blower.
Sewage treatment support system. The control target value is at least one of the following:
- Dissolved oxygen concentration (DO) in the activated sludge in the aeration tank,
- an inflow ratio representing the ratio of the discharge air volume of the blower to the sewage inflow volume into the sewage treatment plant;
The sewage treatment support system according to claim 1 . The sewage treatment support system of claim 1, wherein the water quality data includes the biochemical oxygen demand (BOD) of the secondary effluent of the sewage treatment plant as an indicator. The status data for each predetermined period includes a plurality of values,
In the state data for each predetermined period, the plurality of values include a control target value immediately before the predetermined period and a control target value for the predetermined period,
the transition condition learning unit learns a model in which a state transition from a certain record to a next record of the state data is an output, and one or more values including a control target value among the plurality of values is an input.
The sewage treatment support system according to claim 1 . The sewage treatment support system according to claim 4 , wherein the model is a model in which the state transition is a response variable and the one or more values are explanatory variables. The sewage treatment support system of claim 1, wherein the transition condition learning unit calculates the angle of the centroid vector representing the transition direction and the previous and next time series for two clusters with adjacent time series, and learns the transition condition when the centroid vector angle is less than a predetermined threshold. The sewage treatment support system of claim 1, wherein the specified period is a week. The sewage treatment support system of claim 7, wherein the transition condition learning unit constructs explanatory variables as a learning matrix including the amount of air discharged by the blower in the previous week, the dissolved oxygen concentration in the activated sludge in the aeration tank in the previous week, the amount of sludge returned to the aeration tank in the previous week, the amount of sewage inflow into the sewage treatment plant in the previous week, the biochemical oxygen demand of the secondary treatment water of the sewage treatment plant, and a predicted value of the dissolved oxygen concentration in the activated sludge in the aeration tank in the next week . The sewage treatment support system according to claim 8 , wherein the transition condition learning unit generates the learning matrix as an average vector of a past period T. The sewage treatment support system of claim 1 further comprises an output unit that displays the determined blower control target value on a display device. The sewage treatment support system of claim 1 further comprises a power consumption estimation unit that estimates the power consumption of the blower based on the amount of air discharged by the blower. The control target value determination unit
Identifying a latest location cluster, which is a cluster in which the latest state data is classified;
estimating a cluster to which the latest position cluster will transition based on the latest position cluster, the latest state data, and a control target value for a next predetermined period;
By repeating the process including the following, a control target value is searched for such that the cluster to which the latest position cluster is to be transitioned will be a cluster in which the discharge air volume of the blower is reduced or the quality of secondary effluent from the sewage treatment plant is improved;
- changing the latest control target value for the next predetermined period based on the learned transition condition;
- estimating a cluster to which the latest position cluster will transition based on the latest changed control target value for the next predetermined period;
Output the control target value found.
The sewage treatment support system according to claim 1 . The computer performs clustering of status data for each predetermined period that represents the status of activated sludge in an aeration tank of a sewage treatment plant that performs sewage treatment based on the standard activated sludge method,
a computer learning conditions for transitions between the clusters generated by the clustering;
a computer determines a control target value for a blower that discharges air for aerating the aeration tank based on the learned transition condition and a time series of new status data that represents the status of the activated sludge in the aeration tank;
The status data includes water quality data representing the quality of secondary treated water from the sewage treatment plant and operation data representing the operating status of the sewage treatment plant,
The operating data includes, as indicators, an amount of sewage inflow into the sewage treatment plant, a dissolved oxygen concentration in the activated sludge in the aeration tank, an amount of sludge returned to the aeration tank, and an amount of air discharged by the blower.
Sewage treatment support methods. Clustering is performed on status data for each predetermined period that represents the status of activated sludge in an aeration tank of a sewage treatment plant that performs sewage treatment based on the standard activated sludge method;
learning conditions for transitions between clusters among the plurality of clusters generated by the clustering;
causing a computer to determine a control target value for a blower that discharges air for aerating the inside of the aeration tank based on the learned transition condition and a time series of new status data that represents the status of the activated sludge in the aeration tank;
The status data includes water quality data representing the quality of secondary treated water from the sewage treatment plant and operation data representing the operating status of the sewage treatment plant,
The operating data includes, as indicators, an amount of sewage inflow into the sewage treatment plant, a dissolved oxygen concentration in the activated sludge in the aeration tank, an amount of sludge returned to the aeration tank, and an amount of air discharged by the blower.
Wastewater Treatment Assistance Program.