Membrane differential pressure estimation device and air dispersion amount control device

Abstract

Not bound to the general idea of air blow amount control, a temporal change in the differential pressure between the membranes is appropriately estimated in order to perform proper membrane filtration operation. The estimation device (2) has: an input data acquisition section (21) that acquires input data measured in the membrane filtration operation, derived from operation data including the membrane filtration pressure and the air blow amount; and an estimation section (23) that performs M (M is an integer of 2 or more) times of processing that is a specific processing of estimating the differential pressure between the membranes after a prescribed time using a regression model (31), and updating the input data by changing the differential pressure between the membranes included in the input data to the estimated differential pressure between the membranes, with a part of the data included in the input data being different, and performs N (N is an integer of 2 or more) times of processing, thereby obtaining M estimation results of the temporal change of N times the prescribed time or more for the differential pressure between the membranes.

Description

Technical Field

[0001] The present invention relates to a membrane differential pressure estimation device, etc., used in membrane filtration treatment to aerate the membrane surface of a separation membrane disposed in water to be treated and to obtain treated water that has permeated through the separation membrane. Background Technology

[0002] Patent Document 1 discloses a technique for controlling the amount of gas released from a separation membrane used in membrane filtration. This technique involves comparing the amount, rate of change, or rate of increase of the inter-membrane differential pressure from a past point in time with a pre-set threshold and a target rate of increase selected based on the concentration of organic matter, to determine the amount of gas released. Specifically, it discloses a technique for suppressing the rate of increase of the inter-membrane differential pressure by increasing the amount of gas released when the calculated rate of increase of the inter-membrane differential pressure exceeds the target rate of increase.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent No. 6342101 Summary of the Invention

[0006] The problem that the invention aims to solve

[0007] In controlling the aeration rate during membrane filtration operation, it is generally believed that "when the membrane is in poor condition, increasing the aeration rate can suppress the rate of increase in inter-membrane differential pressure." However, the applicant of this application has found that, depending on the operating conditions of the membrane filtration, sometimes reducing the aeration rate can suppress the rate of increase in inter-membrane differential pressure. Furthermore, increasing the aeration rate can lead to an increase in inter-membrane differential pressure or fouling. In other words, the applicant has found that the general view regarding the relationship between aeration rate and inter-membrane differential pressure is not always correct.

[0008] One objective of the present invention is to provide a membrane differential pressure estimation device, which is not limited to the general viewpoint of gas dissipation control, and appropriately estimates the time-dependent changes in membrane differential pressure in order to perform proper membrane filtration operation.

[0009] Technical means to solve the problem

[0010] To address the aforementioned problems, one aspect of the present invention provides a membrane differential pressure estimation device comprising: an input data acquisition unit that acquires input data derived from operating data including membrane filtration pressure and aeration rate, wherein the operating data is data measured during membrane filtration operation in a membrane separation device, wherein the membrane separation device includes a separation membrane impregnated in the water to be treated and an aeration device for aeration of the membrane surface of the separation membrane, and treated water permeated through the separation membrane is obtained while aeration of the membrane surface is performed by the aeration device; and an estimation unit that performs estimation processing, wherein the estimation processing is performed by making a portion of the data included in the input data different to execute M(M The process involves performing N (N being an integer greater than or equal to 2) processing steps. The process involves using a regression model that uses the input data as the independent variable and the inter-membrane differential pressure correlation data (corresponding to the input data and established at a predetermined time) as the dependent variable to infer inter-membrane differential pressure correlation data after a predetermined time. Furthermore, the process updates the input data by changing the inter-membrane differential pressure correlation data contained in the input data to the inferred inter-membrane differential pressure correlation data. This process is performed N times, resulting in M ​​inferred results regarding the time-dependent changes in the inter-membrane differential pressure up to N × the predetermined time.

[0011] In addition, in one aspect of the gas dissipation control device of the present invention, the gas dissipation control device has a gas dissipation acquisition unit, which acquires gas dissipation correlation data included in the input data used in the prediction of the prediction result determined by the inter-membrane differential pressure prediction device. The input data includes the gas dissipation correlation data related to the gas dissipation of the gas dissipation device, and controls the gas dissipation of the gas dissipation device based on the acquired gas dissipation correlation data.

[0012] The membrane differential pressure estimation device and the gas dissipation control device of various embodiments of the present invention can be implemented by a computer. In this case, by making the computer operate as the various parts (software elements) of the membrane differential pressure estimation device and the gas dissipation control device, the control program of the membrane differential pressure estimation device and the gas dissipation control device, as well as the computer-readable recording medium recording thereon, are also within the scope of the present invention.

[0013] Invention Effects

[0014] One aspect of the invention is not limited to the general viewpoint of gas dissipation control, but appropriately estimates the time-dependent changes in inter-membrane differential pressure in order to achieve proper membrane filtration operation. Attached Figure Description

[0015] Figure 1This is a diagram illustrating an outline of a membrane differential pressure estimation system according to one embodiment of the present invention.

[0016] Figure 2 This is a diagram representing a specific example of input data derived from runtime data.

[0017] Figure 3 It means Figure 1 A diagram illustrating the general outline of the speculation process performed by the speculation device shown.

[0018] Figure 4 It means in Figure 1 The graph shows the time-varying changes in the measured membrane filtration pressure during membrane filtration operation performed by the membrane separation device shown.

[0019] Figure 5 It means Figure 1 The diagram shows an example of the main structural components of the regression model generation device, the prediction device, and the gas dissipation control device.

[0020] Figure 6 It means Figure 5 The speculation process performed by the speculation device shown and Figure 5 The flowchart shows an example of the process of performing air volume control by the air volume control device.

[0021] Figure 7 This is a diagram illustrating a specific example of aeration volume control treatment. Detailed Implementation

[0022] <Overview of the differential pressure estimation system between separated membranes>

[0023] Figure 1 This is a diagram showing an outline of the intermembrane differential pressure estimation system 100 for the separation membrane of this embodiment. The intermembrane differential pressure estimation system 100 is a system that uses a regression model generated through machine learning to estimate the time-dependent change in the intermembrane differential pressure of the separation membrane 93 used for membrane filtration operation, and controls the amount of gas supplied to the separation membrane 93 based on the estimation results. Details regarding the intermembrane differential pressure will be explained later.

[0024] The inter-membrane differential pressure estimation system 100 includes a regression model generation device 1, an estimation device 2 (inter-membrane differential pressure estimation device), a storage device 3, an operating data acquisition device 4, an input data calculation device 5, an air dissipation control device 8, and a membrane separation device 90. In addition, it may also include a storage device 7.

[0025] Furthermore, there are no limitations on the installation method and location of the regression model generation device 1, the inference device 2, the storage device 3, the running data acquisition device 4, the input data calculation device 5, and the storage device 7. However, as a preferred example, the running data acquisition device 4 and the air volume control device 8 are installed as PLCs (Programmable Logic Controllers), the inference device 2, the input data calculation device 5, and the storage device 7 are edge computing, and the regression model generation device 1 and the storage device 3 are cloud computing.

[0026] (Membrane separation device 90)

[0027] The membrane separation unit 90 is a device that performs membrane filtration to filter water being treated using a separation membrane and obtains treated water that has passed through the separation membrane. This treated water can also be characterized as water in which impurities have been removed through filtration.

[0028] The membrane separation device 90 includes a membrane separation tank 91, a separation membrane 93, an air diffuser 94, an air diffuser 95, a filtered water piping 96, and a filter pump 97. The membrane separation tank 91 stores the water to be treated 92. The separation membrane 93 is immersed in the water to be treated 92 and filters it. The filtered water piping 96 is connected to the membrane separation tank 91 via the separation membrane 93, allowing the treated water 92 to flow through the filtered water. The filter pump 97 is connected to the separation membrane 93 via the filtered water piping 96, allowing the treated water to flow out. The air diffuser 95 supplies air for stripping impurities adhering to the separation membrane 93. In other words, the air diffuser 95 diffuses air onto the membrane surface of the separation membrane 93. The air diffuser 94 is positioned directly below the separation membrane 93 and supplies air bubbles flowing upwards from below the separation membrane 93, supplied by the air diffuser 95.

[0029] The membrane separation tank 91 only needs to be able to accept and store the treated water 92 flowing into it, and can be made of a watertight material such as concrete, stainless steel, or resin. In addition, the structure of the membrane separation tank 91 only needs to be watertight.

[0030] The separation membrane 93 can be any membrane capable of separating solids and liquids, such as hollow fiber membranes or flat membranes. Examples of separation membranes 93 include, but are not limited to, reverse osmosis (RO) membranes, nanofiltration (NF) membranes, ultrafiltration (UF) membranes, and microfiltration (MF) membranes.

[0031] The air diffuser 94 only needs to be capable of supplying air bubbles, and its material can be glass, stainless steel, sintered metal, or resin, etc. The air diffuser 95 only needs to be a device such as a blower that can compress air.

[0032] (Data acquisition device 4)

[0033] The operation data acquisition device 4 uses various sensors to acquire operation data measured during membrane filtration operation and sends this operation data to the input data calculation device 5. The operation data in this embodiment includes at least membrane filtration pressure, aeration rate, water temperature, and elapsed time. Membrane filtration pressure is obtained, for example, from a pressure gauge located on the filter water piping 96 between the separation membrane 93 and the filter pump 97. Aeration rate is the amount of air supplied from the aeration device 95, obtained directly from the aeration device 95. Water temperature is the temperature of the treated water 92, obtained from a thermometer located within the treated water 92 stored in the membrane separation tank 91. Elapsed time is the elapsed time since the chemical washing of the separation membrane 93, obtained from a timer. This timer resets the elapsed time each time chemical washing is performed, and its location is not particularly limited. This timer can also be installed as an application program on the operation data acquisition device 4. Furthermore, this timer can be connected to the input data calculation device 5 in a communicable manner, or it can be installed as an application program on the input data calculation device 5. In this example, the input data calculation device 5 acquires the elapsed time without going through the operation data acquisition device 4. In addition, the above reset can also be performed manually by the user of the inter-membrane differential pressure estimation system 100.

[0034] In addition, chemical washing refers to the use of chemicals to clean the separation membrane 93 that has been contaminated by membrane filtration.

[0035] Furthermore, the operating data is not limited to this example. For instance, the operating data may also include membrane filtration flow rate. The membrane filtration flow rate is obtained, for example, from a flow meter configured in the filtered water piping 96.

[0036] (Input data calculation device 5)

[0037] The input data calculation device 5 derives input data for input to the regression model generation device 1 and the inference device 2 based on the received operating data. The input data represents characteristic quantities of the operating data and can be either the operating data itself or data obtained through calculations performed on the operating data. Then, in the regression model generation stage, the input data calculation device 5 directly sends the calculated input data to the regression model generation device 1 or to the storage device 7 for storing the input data. Conversely, in the inference stage of the inter-membrane differential pressure, the input data calculation device 5 sends the calculated input data to the inference device 2. Further details regarding the input data will be explained later.

[0038] (Regression Model Generation Device 1)

[0039] The regression model generation device 1 generates a regression model for estimating the inter-membrane differential pressure through machine learning using the received input data, and stores it in the storage device 3. Details regarding the generation of the regression model will be explained later.

[0040] (Speculation device 2)

[0041] The estimation device 2 accesses the regression model stored in the storage device 3 and uses it to estimate the time-dependent change in the inter-membrane differential pressure based on the input data received from the input data calculation device 5. Details regarding this time-dependent change estimation will be explained later.

[0042] (Gas dissipation control device 8)

[0043] The air dissipation control device 8 determines the air dissipation level of the air dissipation device 95 (hereinafter referred to as "air dissipation level") based on the prediction result of the prediction device 2, and controls the air dissipation device 95 to perform air dissipation in a manner that determines the air dissipation level. Details regarding air dissipation control will be explained later.

[0044] (Specific example of input data)

[0045] Figure 2 This is a diagram illustrating a specific example of input data derived from operational data. The input data calculation device 5 calculates the dispersion of the membrane filtration pressure, for example, based on the membrane filtration pressure, which is the operational data. The dispersion of the membrane filtration pressure is the dispersion of the membrane filtration pressure during a specific cycle of membrane filtration operation (hereinafter referred to as the "cycle of interest"). Furthermore, details regarding the cycle of membrane filtration operation will be explained later.

[0046] Transmembrane pressure (TMP) is the difference between the pressure applied to the treated water side and the pressure applied to the treated water side within the separation membrane 93. The rate of change of the transmembrane pressure (hereinafter sometimes simply referred to as the "rate of change") is calculated as the slope (ΔTMP / ΔT) of the transmembrane pressure over a specified period (hereinafter referred to as "P") starting from a specified time point in the cycle of concern. It should be noted that P is appropriately chosen between several hours and several days. As an example, the rate of change can be calculated as the slope of a regression model (linear regression) of the time-dependent change in the transmembrane pressure at P. In this case, the rate of change may not be negative.

[0047] In addition, the input data calculation device 5 calculates the cumulative value of the air dissipation based on the air dissipation volume, which is used as operating data, as an example. The cumulative value of the air dissipation volume (hereinafter referred to as the cumulative air dissipation volume) is the cumulative value of the air dissipation volume in P, and as an example, it is calculated as the integral value of the air dissipation volume in P.

[0048] like Figure 2 As shown, the input data can include multiple variation rates of P and the cumulative gas dissipation volume. Figure 2 In the example, P represents the most recent 1 hour, the most recent 3 hours, and the most recent 24 hours, respectively.

[0049] Additionally, the input data calculation device 5 calculates the average water temperature as an example, based on the water temperature used as operating data. The average water temperature is the average water temperature during the observation cycle. The elapsed time is the elapsed time itself included in the operating data.

[0050] In addition, the input data is not limited to Figure 2 Examples of input data include, for instance, data calculated based on membrane filtration pressure, and may also include the maximum value of membrane filtration pressure, the minimum value of membrane filtration pressure, the standard deviation of membrane filtration pressure, the average value of membrane filtration pressure, the inter-membrane differential pressure, the rate of change of the inter-membrane differential pressure, the amount of change of the inter-membrane differential pressure, and the rate of change of the inter-membrane differential pressure.

[0051] The maximum membrane filtration pressure is the maximum membrane filtration pressure in the focus cycle. The minimum membrane filtration pressure is the minimum membrane filtration pressure in the focus cycle. The standard deviation of the membrane filtration pressure is the standard deviation of the membrane filtration pressure in the focus cycle. The average membrane filtration pressure is the average membrane filtration pressure in the focus cycle.

[0052] Additionally, input data may include, for example, the amount and rate of change of the intermembrane differential pressure. The amount of change in the intermembrane differential pressure (hereinafter referred to as "the amount of change") refers to the change in P. For example, the amount of change is calculated as the difference between the value of TMP at a specified time point and the value of TMP at a time point after P. The rate of change in the intermembrane differential pressure (hereinafter referred to as "the rate of change") refers to the rate of change in P. For example, the rate of change (ΔTMP / (TMP×ΔT)) is calculated by dividing the rate of change by the intermembrane differential pressure.

[0053] Alternatively, the input data may include, for example, the average value of the air volume calculated based on the air volume. The average value of the air volume is the average value of the air volume in the cycle of interest.

[0054] Alternatively, the input data may include, for example, the average membrane filtration flow rate and the cumulative membrane filtration flow rate, which are calculated based on the membrane filtration flow rate. The average membrane filtration flow rate is the average membrane filtration flow rate in the focus cycle. The cumulative membrane filtration flow rate is the cumulative value of the membrane filtration flow rate in P, and as an example, it is calculated as the integral value of the average membrane filtration flow rate in P.

[0055] In addition, although not illustrated, a correspondence can be established between the input data and the time information, which indicates the time when the running data, which is the source of the input data, was obtained.

[0056] In addition, the input data, which is calculated based on various operational data and P, may include data other than the rate of change and the cumulative aeration. For example, the input data may also include the average water temperature in P.

[0057] <Summary of Speculative Processing>

[0058] In the extrapolation processing of inter-membrane differential pressure in this embodiment (hereinafter referred to as "extrapolation processing"), long-term extrapolation based on regression analysis is performed by executing a simulation with changing parameters and performing iterative steps simultaneously.

[0059] [Process 1] Figure 3 This is a schematic diagram representing an overview of speculative processing. (See reference...) Figure 3 First, we will provide an overview of long-term predictions based on regression analysis.

[0060] In this embodiment, the regression analysis uses the input data obtained from the input data calculation device 5 or the storage device 7 as the independent variable, and the data related to the intermembrane differential pressure of the separation membrane 93 n hours after the time corresponding to the input data is established (n is a positive integer) as the dependent variable (hereinafter referred to as "intermembrane differential pressure correlation data"). As an example, the intermembrane differential pressure correlation data is at least one of the above-mentioned intermembrane differential pressure itself, the rate of change of the intermembrane differential pressure, the amount of change of the intermembrane differential pressure, and the rate of change of the intermembrane differential pressure. It should be noted that in this embodiment, the intermembrane differential pressure correlation data is the rate of change. Through this regression analysis, the intermembrane differential pressure n hours after which the current value of the gas dissipation volume is estimated is processed. Then, using the estimated rate of change n hours later, data updating the rate of change in the input data (hereinafter referred to as "updated data") is generated, and regression analysis is performed again on the updated data. This process is repeated N times (N is an integer of 2 or more). That is, perform the following process N times: "Infer the rate of change after n hours through regression analysis, and update the input data by changing the rate of change contained in the input data to the inter-membrane differential pressure correlation data after n hours".

[0061] Specifically, as the first iteration, a regression analysis is performed using the input data as the independent variable. Based on the predicted rate of change n hours after the corresponding moment with the input data, this rate of change is used to update the rate of change in the input data, generating updated data U. (1，1) .

[0062] As the Xth iteration (X is an integer greater than 2 and less than N), the data U is updated. (1，X-1) In regression analysis, the rate of change of the input data is estimated based on the time interval X×n hours prior to the initial time. This rate of change is then used to update the updated data U.(1，X-1) The rate of change in the data generates updated data U. (1，X) .

[0063] As the Nth iteration, by updating the data U (1，N-1) Regression analysis, using the input data as the independent variable, infers the rate of change N×n hours after the corresponding time point. This rate of change can also be used to update the update data U. (1，N-1) The rate of change in the data generates updated data U. (1， N ) .

[0064] Based on the above, a total of N inter-membrane differential pressures were established corresponding to the input data after n hours, 2n hours, ..., N×n hours. Based on the above, the temporal change of the inter-membrane differential pressure after N×n hours while maintaining the current value of the gas dissipation rate was predicted.

[0065] [Process 2] Next, using data (hereinafter referred to as "simulated data") with altered data related to the gas dissipation volume in the input data (a portion of the data, hereinafter referred to as "gas dissipation volume associated data"), the above-described Process 1 is performed. The gas dissipation volume associated data refers to at least one of the above-described gas dissipation volume, the average value of the gas dissipation volume, and the cumulative value of the gas dissipation volume. It should be noted that in this embodiment, the gas dissipation volume associated data is set as the gas dissipation volume. The gas dissipation volume is changed and this process is performed M times (M is an integer of 2 or more). That is, Process 1 is performed on M different data (input data and M-1 simulated data). That is, in Process 2, by making a portion of the data contained in the input data different and performing Process 1 M times, M inferred results are obtained for the time-varying change of the inter-membrane differential pressure up to N×n hours.

[0066] The first of the M processes is the aforementioned process 1 for the input data, specifically as described above.

[0067] As the Y-th processing in the M-th iteration (Y is an integer greater than 2 and less than M), the simulated data S in the input data with the altered gas dissipation volume is processed. Y Perform the above process 1. Specifically, as the first iteration in process 1, the simulation data S is... Y Regression analysis with the inferred and simulated data S as the independent variable Y After establishing the rate of change n hours from the corresponding time point, the simulation data S is updated using this rate of change. Y The rate of change in the data generates updated data U. (Y，1) As the Xth iteration in process 1 above, the updated data U is... (Y，X-1) Regression analysis with the inferred and simulated data S as the independent variable YAfter establishing the rate of change X×n hours from the corresponding time point, use this rate of change to update the update data U. (Y，X-1) The rate of change in the data generates updated data U. (Y，X) As the Nth iteration in process 1 above, the data U is updated... (Y，N-1) Regression analysis with the inferred and simulated data S as the independent variable Y The rate of change was established N×n hours after the corresponding time point.

[0068] By performing the above processing 1 and processing 2, it is possible to predict the time-dependent change of the inter-membrane differential pressure for each of the M gas dissipation volumes up to N×n hours while maintaining the current value of the gas dissipation volume.

[0069] <The Cycle of Membrane Filtration Operation>

[0070] Figure 4 This is a graph showing the change in membrane filtration pressure over time, measured during membrane filtration operation performed by the membrane separation unit 90. (Using...) Figure 4 The operation cycle of membrane filtration is described below. A membrane filtration operation cycle consists of an operating period during which membrane filtration is performed (e.g., approximately 5 minutes) and a rest period after this operating period during which membrane filtration is not performed (e.g., approximately 1 minute). Membrane filtration operation is an intermittent operation consisting of repeating this cycle. It should be noted that in this specification, the membrane filtration operation cycle is sometimes described as a "unit period".

[0071] In the intermembrane differential pressure estimation system 100, as an example, preferably during a rest period, the operating data acquired by the operating data acquisition device 4 during operation is used, the input data calculation device 5 derives the input data, and then preferably the estimation device 2 estimates the time-dependent change in the intermembrane differential pressure, and the gas dissipation device 95 is controlled based on the estimation result. This series of processes is performed periodically in the intermembrane differential pressure estimation system 100. Specifically, the estimation device 2 acquires input data from the input data calculation device 5 at a period of L (L is an integer greater than or equal to 1) times per unit period (hereinafter referred to as "unit period"), and performs the above-described estimation process each time input data is acquired. Thus, in the intermembrane differential pressure estimation system 100, after estimating the time-dependent change in the intermembrane differential pressure during the rest period of each unit period, the gas dissipation amount of the gas dissipation device 95 can be appropriately controlled. It should be noted that the typical value of L is "1". That is, preferably, the estimation device 2 acquires input data from the input data calculation device 5 and performs the above-described estimation process during each rest period.

[0072] <Main Structure of Each Device>

[0073] Figure 5This is a block diagram illustrating an example of the main structural components of the regression model generation device 1, the prediction device 2, and the gas dissipation control device 8 in this embodiment.

[0074] (Regression Model Generation Device 1)

[0075] The regression model generation apparatus 1 includes a control unit 10. The control unit 10 uniformly controls all parts of the regression model generation apparatus 1, and, for example, is implemented by a processor and a memory. In this example, the processor accesses a memory space (not shown), loads a program (not shown) stored in the memory space into the memory, and executes a series of commands included in the program. Thus, the various parts constituting the control unit 10 are described.

[0076] The control unit 10 includes an input data acquisition unit 11, a correspondence establishment unit 12, and a regression model generation unit 13.

[0077] The input data acquisition unit 11 acquires input data from the input data calculation device 5 or the storage device 7, and outputs the acquired input data to the correspondence establishment unit 12.

[0078] The correspondence establishment unit 12 establishes a correspondence between the rate of change n hours after the time when the input data is established and each establishment of the input data. The value of N is, for example, 12 or 24, but is not limited to this example.

[0079] The correspondence establishment unit 12 outputs the input data, after establishing a correspondence with the rate of change after n hours, to the regression model generation unit 13. It should be noted that the rate of change after n hours does not yet exist. Therefore, input data that has not yet established a correspondence with the rate of change can simply remain in the correspondence establishment unit 12 until the rate of change can be obtained.

[0080] The regression model generation unit 13 takes the input data as independent variables, generates a regression model 31 with the rate of change after n hours as the dependent variable, and stores it in the storage device 3.

[0081] (Speculation device 2)

[0082] The speculation device 2 includes a control unit 20 and an output unit 27. The control unit 20 uniformly controls all parts of the speculation device 2, and for example, is implemented by a processor and a memory. In this example, the processor accesses a memory space (not shown), loads a program (not shown) stored in the memory space into the memory, and executes a series of commands contained in the program. Thus, the various parts of the control unit 20 are constituted.

[0083] As these components, the control unit 20 includes an input data acquisition unit 21, an access unit 22, a lifespan determination unit 24 (period determination unit), a cost calculation unit 25, and a prediction result selection unit 26 (determination unit).

[0084] The input data acquisition unit 21 acquires input data from the input data calculation device 5 every unit cycle and outputs it to the access unit 22. Preferably, the input data is data derived from the most recent operating data measured during membrane filtration operation.

[0085] Access unit 22 accesses regression model 31 stored in storage device 3. Access unit 22 includes prediction unit 23.

[0086] The inference unit 23 uses the regression model 31 accessed by the access unit 22 to perform inference processing using the input data obtained from the input data acquisition unit 21. Specifically, the inference unit 23 first performs the aforementioned processing 1. Specifically, the inference unit 23 inputs the input data obtained by the input data acquisition unit 21 into the regression model 31 accessed by the access unit 22, thereby obtaining the rate of change after n hours from the regression model 31. Next, the inference unit 23 inputs updated data, which is updated with the obtained rate of change after n hours, into the regression model 31, thereby obtaining the rate of change after 2n hours from the regression model 31. This process is repeated N times to obtain the rate of change up to N×n hours.

[0087] Then, the estimation unit 23 performs the aforementioned processing 2. Specifically, the estimation unit 23 performs estimations based on the regression model 31 of processing 1 for the input data and simulated data, obtaining the rate of change up to N×n hours. Thus, the estimation unit 23 can obtain M estimation results for the time-dependent change in inter-membrane differential pressure up to N×n hours. The estimation unit 23 outputs the input data or simulated data from the estimation source and the M estimation results respectively to the lifetime determination unit 24. Thereafter, the combination of the estimation results and the input data or simulated data is described as "estimated data".

[0088] The lifespan determination unit 24 determines the lifespan of the separation membrane 93 based on the M predicted results (i.e., the time-dependent changes in the intermembrane differential pressure) derived by the prediction unit 23. Here, lifespan refers to the period until the intermembrane differential pressure of the separation membrane 93 reaches a predetermined upper limit value; in other words, it is the period during which the separation membrane 93 requires chemical washing. This upper limit value is, for example, 12 kPa, but is not limited to it. As an example, the lifespan determination unit 24 determines the time point when the intermembrane differential pressure reaches the upper limit value from the time-dependent changes in the intermembrane differential pressure included in each predicted data set as the lifespan of the separation membrane 93. The lifespan determination unit 24 outputs the determined lifespans to the cost calculation unit 25, establishing a correspondence between each lifespan and the predicted data from the determination source.

[0089] The cost calculation unit 25 calculates the operating costs incurred during membrane filtration operation up to the end of the lifespan of the separation membrane 93. That is, the cost calculation unit 25 calculates the aforementioned operating costs based on M estimated data points.

[0090] Here, operating cost refers to the total cost of energy spent on gas dispersing by the gas dispersing device 95 until it reaches the aforementioned lifespan, and the total cost of chemical washing of the separation membrane 93 until it reaches the aforementioned lifespan. The cost of chemical washing includes the purchase cost of the chemicals used for chemical washing and the labor cost of the personnel performing the chemical washing operation.

[0091] For each of the M projected data points, the cost calculation unit 25 calculates the amount of gas dissipated per unit of membrane filtration flow rate up to the end of the separation membrane 93's lifespan. As an example, the cost calculation unit 25 calculates the total amount of gas dissipated at the point in time when the separation membrane 93 reaches its lifespan, based on the amount of gas dissipated in the input data or simulation data included in each projected data point, the cumulative amount of gas dissipated, and the corresponding lifespan established for each projected data point. Next, the cost calculation unit 25 calculates the amount of gas dissipated per unit of membrane filtration flow rate by dividing the calculated totals by the total membrane filtration flow rate up to the end of the separation membrane 93's lifespan. This total membrane filtration flow rate can be calculated, for example, by pre-determining the membrane filtration flow rate per unit time and multiplying that membrane filtration flow rate by the time until the separation membrane 93 reaches its lifespan.

[0092] Next, the cost calculation unit 25 calculates the power consumption per unit membrane filtration flow rate based on the calculated air dissipation volume per unit membrane filtration flow rate (hereinafter referred to as "air dissipation volume"). As an example, the cost calculation unit 25 calculates the power consumption per unit membrane filtration flow rate by multiplying the power consumption per unit air dissipation volume, which is a predetermined value, by the calculated air dissipation volume.

[0093] Next, the cost calculation unit 25 calculates the electricity cost per unit of membrane filtration flow based on the calculated power consumption per unit of membrane filtration flow (hereinafter referred to as "power consumption"). This electricity cost is the cost of the aforementioned energy. As an example, the cost calculation unit 25 calculates the electricity cost per unit of membrane filtration flow by multiplying the electricity cost per unit of power consumption, which is a predetermined value, by the calculated power consumption.

[0094] Next, the cost calculation unit 25 calculates the operating cost by adding the cost of chemical washing per unit of membrane filtration flow, which is a predetermined value, to the electricity cost per unit of membrane filtration flow. The cost calculation unit 25 outputs the calculated operating cost, along with the lifespan of the calculation source and the estimated data, to the prediction result selection unit 26.

[0095] The prediction result selection unit 26 selects prediction data that meets preset conditions from the acquired M prediction data. These conditions are at least one of a period condition related to the lifespan after the prediction data is correlated and a cost condition related to operating costs. In this embodiment, an example of a cost condition will be described. For example, the cost condition is "minimum operating cost". The prediction result selection unit 26 outputs the gas dissipation volume of the input data or simulation data contained in the selected prediction data, i.e., the prediction data with the minimum operating cost after correlation, to the output unit 27.

[0096] The output unit 27 is a communication device that outputs (sends) the air volume obtained from the prediction result selection unit 26 to the air volume control device 8.

[0097] (Gas dissipation control device 8)

[0098] The air volume control device 8 includes an air volume acquisition unit 81 and an air volume control unit 82. The air volume acquisition unit 81 acquires the air volume received from the estimation device 2 and outputs it to the air volume control unit 82.

[0099] The air dissipation control unit 82 controls the air dissipation device 95 in a manner that dissipates air based on the acquired air dissipation volume.

[0100] <Flowchart for Predictive Processing and Gas Dispersion Control>

[0101] Figure 6 This is a flowchart illustrating an example of the process of prediction processing performed by prediction device 2 and the process of air volume control processing performed by air volume control device 8.

[0102] The input data acquisition unit 21 remains in standby mode until a unit cycle has elapsed (S1). If a unit cycle has elapsed (S1 is "yes"), the input data acquisition unit 21 acquires input data from the input data calculation device 5 (S2). The input data is then output to the access unit 22.

[0103] When input data is acquired, the access unit 22 accesses the regression model 31 stored in the storage device 3. Next, the inference unit 23 performs regression analysis processing (S3). Specifically, the inference unit 23 obtains the rate of change after n hours from the regression model 31 by inputting the input data acquired by the input data acquisition unit 21 into the regression model 31 accessed by the access unit 22.

[0104] Next, the prediction unit 23 determines whether the number of regression analyses performed as the number of times regression analysis processing has reached N (S4). If it has not reached N (S4 is "No"), the prediction unit 23 performs data update processing (S5). Specifically, the prediction unit 23 uses the obtained rate of change after n hours to generate updated data that updates the rate of change in the input data. Then, the prediction unit 23 uses the updated data to perform the processing of S3 again. It should be noted that the execution object of the subsequent processing of S3 is the updated data most recently generated in the processing of S5. That is, the prediction unit 23 inputs the generated updated data into the regression model 31. The prediction unit 23 repeatedly performs the processing of S5 and the subsequent processing of S3 until it is determined in the processing of S4 that the number of regression analyses has reached N. When the number of regression analyses reaches N, the prediction unit 23 obtains the rate of change of the current value of the gas dispersion up to N×n hours from the input data.

[0105] If the number of regression analyses reaches N (S4 is "Yes"), the prediction unit 23 determines whether the number of times the simulation data has been generated has reached M-1 (S6). It should be noted that when S6 is reached for the first time, the number of generation is 0, therefore, the prediction unit 23 determines that the number of generation has not reached M-1.

[0106] If the number of times the simulation data has been generated has not reached M-1 (S6 is "No"), the prediction unit 23 performs simulation data generation processing (S7). Specifically, the prediction unit 23 changes the gas dissipation amount in the input data to generate simulation data. Then, the prediction unit 23 performs processing S3 to S5 on the generated simulation data. As a result, the prediction unit 23 obtains the rate of change after N×n hours while maintaining the current value of the gas dissipation amount for the generated simulation data. In addition, the prediction unit 23 repeatedly performs processing S7 and the following processing S3 to S5 until it is determined in processing S6 that the number of times the simulation data has been generated is M-1. When the number of generation reaches M-1, the prediction unit 23 obtains M rates of change up to N×n hours later.

[0107] When the number of simulation data generation times reaches M-1 (S6 is "Yes"), the estimation unit 23 generates the time-dependent change in intermembrane differential pressure based on each rate of change up to N×n hours. Then, the estimation unit 23 combines (A) the input data, (B) the time-dependent change in intermembrane differential pressure up to N×n hours estimated based on the input data, and (C) the simulation data S2 to S... M (D) Based on simulation data S2~S M Each of the predicted time-varying differential pressures between membranes up to N×n hours is output to the lifetime determination unit 24. The combination of (A) and (B), and the combination of (C) and (D) are the predicted data mentioned above.

[0108] Next, the lifetime determination unit 24 performs lifetime determination processing (S8). Specifically, the lifetime determination unit 24 determines the lifetime of the separation membrane 93 based on the time variation estimated from the input data obtained from the estimation unit 23 and the time variation estimated from the simulation data. The lifetime determination unit 24 outputs each of the determined M lifetimes to the cost calculation unit 25, establishing a correspondence between each and the estimated time variation including the determined source.

[0109] Next, the cost calculation unit 25 performs operating cost calculation processing (S9). Specifically, the cost calculation unit 25 calculates the operating cost based on the corresponding lifespan established with each estimated data and the cumulative gas dissipation included in the input data or simulation data contained in each estimated data. The cost calculation unit 25 outputs each of the calculated M operating costs to the estimation result selection unit 26, corresponding to the estimated data including the input data or simulation data from the calculation source and the lifespan.

[0110] Next, the prediction result selection unit 26 performs prediction result selection processing (S10). Specifically, the prediction result selection unit 26 selects the prediction data with the lowest corresponding operating cost from the M prediction data obtained. The prediction result selection unit 26 outputs the gas dissipation amount of the input data or simulation data included in the selected prediction data to the output unit 27.

[0111] Next, the output unit 27 performs the air volume output processing (S11). Specifically, the output unit 27 sends the air volume selected by the prediction result selection unit 26 to the air volume control device 8.

[0112] Next, the air dissipation control device 8 performs air dissipation control processing (S12). Specifically, the air dissipation control unit 82 controls the air dissipation device 95 to dissipate air based on the air dissipation amount received by the air dissipation acquisition unit 81. After step S12, i.e., the air dissipation control processing, is completed, the process is expected to return to step S1.

[0113] <Specific examples of gas dissipation control>

[0114] Figure 7 This is a diagram illustrating a specific example of aeration volume control treatment. Specifically, Figure 7 This is a graph showing the time-varying operation amount of the most recent gas dissipation volume. Here, operation amount refers to the difference between the gas dissipation volume in the previous unit cycle and the gas dissipation volume in the current unit cycle; specifically, it is the value obtained by subtracting the gas dissipation volume of the previous cycle from the gas dissipation volume in the current unit cycle. Furthermore, the points on this graph represent the operation amount in each unit cycle.

[0115] exist Figure 7In the example, the above operating values ​​are -0.5, -0.1, 0, 0.1, 0.5, and 1 (L / min). Negative operating values ​​(-0.5 and -0.1) indicate a decrease in the amount of gas dissipated compared to the previous unit cycle. Conversely, positive operating values ​​(0.1, 0.5, and 1) indicate an increase in the amount of gas dissipated compared to the previous unit cycle. It should be noted that the values ​​of the operating values ​​are not limited to this example.

[0116] An operation value of 0 indicates that the gas dissipation volume was not changed from the previous unit cycle. Additionally, it indicates that the prediction result selection unit 26 of the prediction device 2 selected prediction data that includes the input data. Furthermore, operation values ​​other than 0 indicate that the prediction result selection unit 26 selected any one of the prediction data that includes the simulation data.

[0117] Thus, the gas dissipation control device 8 of this embodiment performs gas dissipation control based on the gas dissipation amount selected by the estimation device 2 for each unit cycle. As a result, in membrane filtration, gas dissipation control can be performed in a manner that satisfies preset conditions (in this embodiment, the conditions with the lowest operating cost).

[0118] <Effect>

[0119] As described above, the prediction device 2 of this embodiment includes: an input data acquisition unit 21 that acquires input data; and a prediction unit 23 that performs prediction processing on the time-dependent change of inter-membrane differential pressure to obtain M prediction results.

[0120] According to this structure, in the long-term prediction of inter-membrane differential pressure based on regression analysis, it is possible to perform iterative simulations while changing a portion of the input data, resulting in M ​​prediction results. Since a portion of the input data differs for each of the M prediction results, they represent predictions of different temporal changes in inter-membrane differential pressure. Therefore, the user of the prediction device 2 can select an appropriate prediction result from the M prediction results and perform membrane filtration operation based on that prediction result.

[0121] Furthermore, in speculative processing, since simulations can be performed in various modes of the input data, appropriate membrane filtration operation can be achieved, taking into account situations where the general viewpoint of "increasing the gas flow rate to suppress the rise in inter-membrane differential pressure if the separation membrane is in poor condition" is not followed.

[0122] In addition, the input data acquisition unit 21 acquires input data every unit cycle, and the estimation unit 23 performs estimation processing every time the input data acquisition unit 21 acquires input data.

[0123] Based on this structure, a prediction process is performed every unit cycle, thus M prediction results can be obtained per unit cycle. Therefore, the user of prediction device 2 can select the optimal prediction result at each unit cycle, thereby enabling long-term, continuous, and appropriate membrane filtration operation.

[0124] In addition, the estimation device 2 also includes a lifespan determination unit 24, which determines the lifespan of the separation membrane 93 for each time-related change estimated by the estimation unit 23.

[0125] Based on this structure, the lifespan of the separation membrane 93 can be determined for each of the M predicted results. Thus, the user of the prediction device 2 can select an appropriate prediction of the lifespan of the separation membrane 93 from the M predicted results and perform membrane filtration operation based on that prediction.

[0126] In addition, the estimation device 2 also includes a cost calculation unit 25, which calculates the operating cost of membrane filtration operation until the lifespan of the separation membrane 93.

[0127] Based on this structure, the operating cost up to the chemical washing stage can be calculated for each of the M predicted results. Therefore, the user of the prediction device 2 can select the prediction result with an appropriate operating cost from the M predicted results and perform membrane filtration operation based on that prediction result.

[0128] In addition, the prediction device 2 also includes a prediction result selection unit 26, which selects the prediction result from M prediction results whose operating cost calculated by the cost calculation unit 25 meets the preset cost conditions.

[0129] Based on this structure, the prediction result that meets the pre-set cost conditions is selected from M prediction results, thus reducing the time required for users to select the appropriate prediction result.

[0130] In addition, the cost calculation unit 25 calculates the total cost of the energy spent on the aeration device 95 until the separation membrane 93 reaches its lifespan, and the cost spent on the chemical washing of the separation membrane 93, as the operating cost.

[0131] Based on this structure, it is possible to calculate the operating cost, taking into account the cost of energy and the cost of chemical washing.

[0132] In addition, the gas dissipation control device 8 of this embodiment includes: a gas dissipation acquisition unit 81, which receives the gas dissipation contained in the input data or simulation data selected by the prediction device 2 as the prediction source; and a gas dissipation control unit 82, which controls the gas dissipation device 95 to perform gas dissipation based on the gas dissipation amount.

[0133] According to this structure, the amount of air dissipated by the air dissipation device 95 can be controlled based on the prediction result of the prediction device 2. In addition, since the amount of air dissipated is based on the amount of air dissipated contained in the input data or simulation data used to predict the prediction result that meets the conditions, it is possible to achieve the purpose of dissipating air at the amount of air dissipated as desired by the user.

[0134] The selection criteria used by the prediction result selection unit 26 for selecting prediction data can be limited to a period condition. This period condition could be, for example, "the separation membrane 93 has the longest lifespan." That is, in this example, the prediction result selection unit 26 selects the prediction data from M prediction data that has the longest lifespan of the separation membrane 93.

[0135] According to this structure, since the prediction result that satisfies the preset period conditions is determined from M prediction results, the effort required for the user to select the appropriate prediction result can be reduced. It should be noted that in this example, the control unit 20 may not include the cost calculation unit 25.

[0136] Furthermore, the conditions used by the prediction result selection unit 26 to select prediction data can also be period conditions and cost conditions. In this example, the prediction result selection unit 26 selects prediction data from the M acquired prediction data that establishes a corresponding lifespan that meets a preset period condition and establishes a corresponding operating cost that meets a preset cost condition. As an example, the period condition is "lifespan is within a specified numerical range," and the cost condition can be "operating cost is within a specified numerical range." Regarding these numerical ranges, the user of the prediction device 2 can set the desired numerical range.

[0137] In this example, if there are multiple estimated data points after establishing a correspondence between the lifespan that meets the period conditions and the operating cost that meets the cost conditions, the estimation result selection unit 26 selects any one of the estimated data points. As an example, the estimation result selection unit 26 selects the estimated data point with the smallest sum of the error between the specified target number of days and the lifespan, and the error between the specified target value and the operating cost.

[0138] According to this structure, since the prediction results that satisfy the pre-set period conditions and cost conditions are determined from M prediction results, the effort required for users to select appropriate prediction results can be reduced.

[0139] The prediction device 2 may also omit the lifespan determination unit 24, the cost calculation unit 25, and the prediction result selection unit 26, and instead have a structure that displays M prediction data on a display device (not shown). This display device may be integrated with the prediction device 2 or may be separate from the prediction device 2.

[0140] In this example, the user of the prediction device 2 selects one from M predicted data. The output unit 27 sends the gas dissipation amount of the input data or analog data contained in the predicted data selected by the user to the gas dissipation amount control device 8.

[0141] In addition to the lifespan determination unit 24, the cost calculation unit 25, and the prediction result selection unit 26, the prediction device 2 may also include an output unit 27. In this example, the user of the prediction device 2 inputs the gas dissipation volume of the selected prediction data, which includes the input data or simulation data, to the gas dissipation volume control device 8.

[0142] Furthermore, the operating costs calculated by the cost calculation unit 25 are not limited to the examples mentioned above. Operating costs may, for example, be the cost of energy itself.

[0143] Alternatively, the input data and simulation data may include an operational quantity related to the gas dissipation volume, or in addition to the gas dissipation volume. In the example where this operational quantity is included instead of the gas dissipation volume, the operational quantity becomes data that is modified during the simulation data generation process. Furthermore, in the example where this operational quantity is included in addition to the gas dissipation volume, both the gas dissipation volume and the operational quantity become data that is modified during the simulation data generation process.

[0144] In the example where the input and simulation data include the operational quantity of the gas dissipation rate, this operational quantity in the input data is 0. Furthermore, this operational quantity in the simulation data takes a value other than 0. Figure 7 In the case of M=6, the operands in each simulation data are -0.5, -0.1, 0.1, 0.5, and 1.

[0145] Alternatively, a numerical range that can be obtained as the air dissipation volume can be preset. In other words, the air dissipation volume control device 8 can also control the air dissipation volume within this numerical range. In this example, when the air dissipation volume received by the air dissipation volume control device 8 exceeds the upper limit or falls below the lower limit of the numerical range, the air dissipation volume control device 8 can also maintain the current air dissipation volume (i.e., the upper or lower limit of the numerical range).

[0146] <Postscript>

[0147] One aspect of the present invention provides a membrane differential pressure estimation device, comprising: an input data acquisition unit that acquires input data derived from operating data including membrane filtration pressure and aeration rate, wherein the operating data is data measured during membrane filtration operation in a membrane separation device, wherein the membrane separation device includes a separation membrane impregnated in water to be treated and an aeration device for aeration of the membrane surface of the separation membrane, and treated water permeated through the separation membrane is obtained while aeration of the membrane surface is performed by the aeration device; and an estimation unit that performs estimation processing, wherein the estimation processing is performed by making a portion of the data included in the input data different to perform M (M is 2 or more). The process involves performing N (N being an integer greater than 2) processing operations, wherein the specific processing is as follows: for a regression model that uses the input data as the independent variable and the inter-membrane differential pressure correlation data corresponding to the input data and established a specified time after the inter-membrane differential pressure is as the dependent variable, the inter-membrane differential pressure correlation data after the specified time is inferred, and the input data is updated by changing the inter-membrane differential pressure correlation data contained in the input data to the inferred inter-membrane differential pressure correlation data, the specific processing is performed N (N being an integer greater than 2), thereby obtaining M inferred results of the time-dependent change of the inter-membrane differential pressure up to N × the specified time.

[0148] According to this structure, in the long-term prediction of inter-membrane differential pressure based on regression analysis, iterative simulations can be performed by changing a portion of the input data to obtain M prediction results. Each of the M prediction results represents a different portion of the input data, thus resulting in different predictions of the temporal variation of the inter-membrane differential pressure. Therefore, the user of the prediction device 2 can select an appropriate prediction result from the M prediction results and perform membrane filtration operation based on that prediction result.

[0149] Furthermore, in the above-mentioned speculative processing, since simulations can be performed in various modes of input data, appropriate membrane filtration operation can be achieved, taking into account situations where the general view of "increasing the gas flow rate to suppress the rise of inter-membrane differential pressure if the separation membrane is in poor condition" is not followed.

[0150] In addition, in one aspect of the membrane differential pressure estimation device of the present invention, the membrane filtration operation is intermittent, the input data is derived from the operating data within a unit period consisting of the operating period and the rest period after the operating period, the input data acquisition unit acquires the input data at a period of L (L is an integer multiple of 1 or more) of the unit period, and the estimation unit performs the estimation process whenever the input data acquisition unit acquires the input data.

[0151] According to the described structure, since the estimation process is performed at a cycle of L times the unit period, M estimation results can be obtained for each cycle. Therefore, the user of the intermembrane differential pressure estimation device can select the optimal estimation result for each cycle, thus enabling long-term, continuous, and appropriate membrane filtration operation.

[0152] In addition, one aspect of the membrane differential pressure estimation device of the present invention may also include a period determination unit, which determines the period until the membrane differential pressure reaches a preset upper limit value for each of the time-dependent changes estimated by the estimation unit.

[0153] Based on the aforementioned structure, the required period for washing the separation membrane (in other words, the membrane's lifespan) can be determined for each of the M predicted results. Thus, the user of the membrane differential pressure prediction device can select the prediction result with an appropriate membrane lifespan from the M predicted results and perform membrane filtration operation based on that prediction result.

[0154] In addition, one aspect of the membrane differential pressure estimation device of the present invention may also include a cost calculation unit that calculates the operating costs incurred by the membrane filtration operation up to a determined period.

[0155] Based on the aforementioned structure, the operating cost up to the time of chemical washing can be calculated for each of the M predicted results. Therefore, the user of the membrane differential pressure prediction device can select the prediction result with an appropriate operating cost from the M prediction results and perform membrane filtration operation based on that prediction result.

[0156] In addition, one aspect of the membrane differential pressure estimation device of the present invention may also include a determination unit, which determines the estimation result among the M estimation results in which the period determined by the period determination unit satisfies a preset period condition.

[0157] According to the structure, the prediction result that satisfies the preset period conditions is selected from M prediction results, thus reducing the time required for the user to select the appropriate prediction result.

[0158] In addition, one embodiment of the membrane differential pressure estimation device of the present invention may further include a determining unit, which determines the estimation result among the M estimation results in which the operating cost calculated by the cost calculation unit satisfies a preset cost condition.

[0159] According to the structure, the prediction result that satisfies the pre-set cost conditions is determined from M prediction results, thus reducing the effort required for users to select the appropriate prediction result.

[0160] In addition, one aspect of the membrane differential pressure estimation device of the present invention may further include a determination unit, which determines one of the M estimation results in which the period determined by the period determination unit satisfies a preset period condition and the operating cost calculated by the cost calculation unit satisfies a preset cost condition.

[0161] According to the structure, the prediction result that satisfies the pre-set period conditions and cost conditions is determined from M prediction results, thus reducing the time required for users to select appropriate prediction results.

[0162] In addition, in one aspect of the membrane differential pressure estimation device of the present invention, the cost calculation unit may also calculate the total cost of energy spent on gas dissipation by the gas dissipation device and the cost spent on chemical washing of the separation membrane up to the determined period, as the operating cost.

[0163] Based on the structure described, it is possible to calculate the operating cost, taking into account the cost of energy and the cost required for chemical washing.

[0164] In addition, in one aspect of the gas dissipation control device of the present invention, the gas dissipation control device has a gas dissipation acquisition unit, which acquires gas dissipation correlation data included in the input data used in the prediction of the prediction result determined by the inter-membrane differential pressure prediction device. The input data includes the gas dissipation correlation data related to the gas dissipation of the gas dissipation device, and controls the gas dissipation of the gas dissipation device based on the acquired gas dissipation correlation data.

[0165] According to the structure described above, the gas dissipation amount of the gas dissipation device can be controlled based on the prediction result of the inter-membrane differential pressure prediction device. Furthermore, since this gas dissipation amount is based on the gas dissipation amount correlation data included in the input data used to predict the prediction result that meets the conditions, it is possible to achieve gas dissipation in a manner that meets the user's desired gas dissipation amount.

[0166] [Software-based implementation example]

[0167] The control modules (especially control units 10 and 20, and air volume control unit 82) of the regression model generation device 1, the prediction device 2, and the air volume control device 8 can be implemented by logic circuits (hardware) formed on integrated circuits (IC chips), or by software.

[0168] In the latter case, the regression model generation device 1, the prediction device 2, and the gas dissipation control device 8 are equipped with a computer that executes commands for software, i.e., programs, to achieve each function. This computer has, for example, one or more processors and a computer-readable recording medium storing the aforementioned program. Furthermore, in the computer, the processor reads the program from the recording medium and executes it, thereby achieving the object of the present invention. As the processor, for example, a CPU (Central Processing Unit) can be used. As the recording medium, a "non-temporary tangible medium," such as tape, disk, card, semiconductor memory, or programmable logic circuit, can be used, in addition to ROM (Read Only Memory). Additionally, RAM (Random Access Memory) for deploying the program may also be included. Furthermore, the program can be provided to the computer via any transmission medium capable of transmitting the program (communication network or broadcast wave, etc.). It should be noted that one aspect of the present invention can also be implemented by electronically transmitting the program as a data signal embedded with a carrier wave.

[0169] This invention is not limited to the above-described embodiments. Various modifications can be made within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included within the technical scope of this invention.

[0170] Figure 2 Reference Numerals 2 Prediction Device (Inter-membrane Differential Pressure Prediction Device) 8 Gas Dissipation Control Device

[0171] 21 Input Data Acquisition Department

[0172] 23. Speculation Section; 24. Lifespan Determination Section (Period Determination Section)

[0173] 25 Cost Calculation Department; 26 Predicted Result Selection Department (Determination Department)

[0174] 31 Regression Model

[0175] 90 membrane separation unit

[0176] 92 treated water

[0177] 93 Separation Membrane 95 Gas Dispersion Device

Claims

1. A membrane differential pressure estimation device, wherein has: an input data acquisition section that acquires input data derived from operation data including a membrane filtration pressure and a gas sparging amount, the operation data being data measured in a membrane filtration operation performed in a membrane separation device that has a separation membrane disposed immersed in treated water and a gas sparging device that performs membrane surface gas sparging of the separation membrane, the treated water that has passed through the separation membrane being obtained while the membrane surface gas sparging is performed by the gas sparging device; and an estimation section, performs an estimation process that performs long-term estimation based on regression analysis by performing simulation that changes parameters and simultaneously performs iteration, wherein the regression analysis uses a regression model that uses the input data as an independent variable and membrane differential pressure correlation data related to a membrane differential pressure of the separation membrane at an n hours later from a time at which the input data is established as a dependent variable to estimate the membrane differential pressure correlation data at the n hours later, and then updates the input data by changing the membrane differential pressure correlation data included in the input data to the estimated membrane differential pressure correlation data, and performs the regression analysis again on the updated data, and repeats this process N times, where n is a positive integer and N is an integer of 2 or more, the process of performing the process N times is a process 1, and by the process 1, a temporal change in the membrane differential pressure at N x n hours later while maintaining a current value of the gas sparging amount is estimated, a process of performing the process 1 on M pieces of mutually different data M times is a process 2, where the M pieces of data are the input data and M-1 pieces of simulation data that are data in which the input data related to the gas sparging amount is changed, and M is an integer of 2 or more, and by the process 2, M estimation results are obtained for a temporal change in the membrane differential pressure up to N x n hours later, by performing the process 1 and the process 2, a temporal change in the membrane differential pressure up to N x n hours later while maintaining a current value of the gas sparging amount can be estimated for each of M gas sparging amounts, the membrane differential pressure estimation device further has a period determination section, a determination section, and an output section, the estimation section outputs the input data or the simulation data of the estimation source and the M estimation results to the period determination section respectively in correspondence with each other, where a combination of the estimation result and the input data or the simulation data is taken as estimation data, the period determination section determines a period until the membrane differential pressure reaches a pre-set upper limit value for each of the temporal changes estimated by the estimation section, the determination section determines estimation data in which the period determined by the period determination section among the M estimation data satisfies a pre-set period condition, the output section outputs the gas sparging amount acquired from the determination section to a gas sparging amount control device.

2. The membrane differential pressure estimation device according to claim 1, wherein the membrane filtration operation is a batch operation, the input data is derived from the operation data in a unit period composed of an operation period and a rest period after the operation period, The input data acquisition section acquires the input data at a cycle of L times the unit period, where L is an integer of 1 or greater, The estimation section performs the estimation processing each time the input data acquisition section acquires the input data.

3. The membrane differential pressure estimation device according to claim 1, wherein Further comprising a cost calculation section that calculates an operation cost expended by the membrane filtration operation up to the determined period.

4. The membrane differential pressure estimation device according to claim 3, wherein The determination section further determines the estimation data in which the operation cost calculated by the cost calculation section satisfies a predetermined cost condition, among the M estimation data.

5. The membrane differential pressure estimation device according to claim 3, wherein The determination section further determines the estimation data in which the period determined by the period determination section satisfies a predetermined period condition and the operation cost calculated by the cost calculation section satisfies a predetermined cost condition, among the M estimation data.

6. The membrane differential pressure estimation device according to claim 3, wherein The cost calculation section calculates, as the operation cost, a total of a cost of energy expended by the air diffuser to perform air diffusion and a cost expended by a chemical cleaning of the separation membrane, up to the determined period.

7. An intermembrane differential pressure estimation system, wherein, having: the membrane differential pressure estimation device according to any one of claims 1 to 6; and an air diffuser amount control device, the input data includes air diffuser amount-related data related to an air diffuser amount of the air diffuser, the air diffuser amount control device has an air diffuser amount acquisition section that acquires the air diffuser amount-related data included in the input data used in the estimation of the estimation result determined by the membrane differential pressure estimation device, controls the air diffuser to perform air diffusion based on the acquired air diffuser amount-related data.

Images