Filtration basin cleaning scheduling device, filtration basin cleaning scheduling system, and filtration basin cleaning scheduling method

Abstract

To provide a filter basin washing scheduling device, a filter basin washing scheduling system, and a filter basin washing scheduling method for more accurately determining washing time of each filter basin.SOLUTION: A filter basin washing scheduling device 1 executes the processes of: measuring a filtration amount measurement value and head loss measurement value of each filter basin in a predetermined period in a water purifying treatment facility with a plurality of filter basins; calculating a future prediction value of the filtration amount from predetermined measurement data; predicting an ascending speed equivalent amount of the head loss for each filter basin from a time series of the measurement values; predicting a future head loss for each filter basin based on the ascending speed equivalent amount prediction value and the future prediction value of the filtration amount and calculating the time when the prediction value reaches a specified value; and determining washing time for each filter basin based on the calculated arrival time of each filter basin.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a filter basin cleaning scheduling device, a filter basin cleaning scheduling system, and a filter basin cleaning scheduling method for scheduling cleaning of filter basins in a water purification plant having a large number of filter basins. [Background technology]

[0002] In Patent Document 1, head loss is used as one of the input information for determining the timing of cleaning. The future value of head loss is linearly predicted, the time when it will reach a predetermined value is determined, and the timing of cleaning is determined based on this information. In Patent Document 2, a neural network is used to calculate the rate of head loss increase from head loss, turbidity of sedimentation-treated water, raw water temperature, etc. This information is used as one of the inputs to determine the optimal injection conditions for filter aid. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 10-286408 [Patent Document 2] Japanese Patent Application Laid-Open No. 2014-50773 Summary of the Invention [Problem to be solved by the invention]

[0004] In Patent Document 1, a simple linear prediction is used, making it impossible to predict head loss with high accuracy. In particular, changes in filtration volume and turbidity cause prediction accuracy to deteriorate, and this deterioration is prolonged. If the cleaning timing of each filter basin is determined based on this prediction error, problems arise, such as determining that cleaning times overlap when they do not, or determining that cleaning times overlap when they do not. In the former case, cleaning times are brought forward, leading to an increase in the number of cleanings, which in turn leads to increased costs. In the latter case, cleaning of one basin is delayed, resulting in the problem of exceeding the head loss upper limit.

[0005] Patent Document 2 utilizes a neural network, but to ensure accuracy in various situations, it is necessary to build a model in advance using a large amount of data. There is a possibility that prediction accuracy will deteriorate if an operating region not included in the training data appears. In addition, because it is a black box model, there is also the problem that the basis for the prediction cannot be explained. It is difficult to notice prediction errors, and there is a possibility that an incorrect predicted value will be used. Due to these issues, it is difficult to use this technology to predict future values ​​of head loss.

[0006] The present invention has been made to solve the above-mentioned problems. That is, one of the objects of the present invention is to provide a filter basin cleaning scheduling device, a filter basin cleaning scheduling system, and a filter basin cleaning scheduling method that can more accurately determine the cleaning timing of each filter basin. [Means for solving the problem]

[0007] In order to solve the above problems, the filter basin cleaning scheduling device of the present invention is applied to a water purification facility having a plurality of filtration basins, and includes an information processing device to which measurement values ​​including filtration volume measurement values ​​and head loss measurement values ​​of each of the filtration basins measured at a predetermined cycle in each of the filtration basins are input, and the information processing device is configured to predict an amount equivalent to the rate of rise of head loss for each of the filtration basins from the measurement values, predict future head losses for each of the filtration basins based on the predicted amount equivalent to the rate of rise of head loss and the future predicted value of filtration volume, calculate the time when the predicted head loss will reach a predetermined value, and determine the timing of cleaning each of the filtration basins based on the calculated arrival time.

[0008] The filter basin cleaning scheduling system of the present invention is applied to a water purification facility having a plurality of filtration basins, and includes an information processing device to which measurement values ​​including filtration volume measurement values ​​and head loss measurement values ​​of each of the filtration basins measured at a predetermined cycle in each of the filtration basins are input. The information processing device is configured to predict an amount equivalent to the rate of rise of head loss for each filtration basin from the measurement values, predict future head losses for each of the filtration basins based on the predicted amount equivalent to the rate of rise of head loss and the future predicted value of filtration volume, calculate the time when the predicted head loss will reach a predetermined value, and determine the timing of cleaning each of the filtration basins based on the calculated arrival time.

[0009] The filter basin cleaning scheduling method of the present invention is applied to a water purification facility having a plurality of filtration basins, and is a filter basin cleaning scheduling method using an information processing device to which measurement values ​​including filtration volume measurement values ​​and head loss measurement values ​​of each of the filtration basins measured at a predetermined cycle are input, and the information processing device predicts an amount equivalent to the rate of rise of head loss for each filtration basin from the measurement values, predicts future head losses for each of the filtration basins based on the predicted amount equivalent to the rate of rise of head loss and the future predicted value of filtration volume, calculates the time when the predicted head loss will reach a predetermined value, and determines the timing of cleaning each of the filtration basins based on the calculated arrival time. [Effects of the Invention]

[0010] According to the present invention, the timing for cleaning each filter basin can be determined more accurately. [Brief explanation of the drawings]

[0011] [Figure 1] Figure 1 shows the overall configuration of the water supply system. [Figure 2] Figure 2 is a diagram of the water supply facilities. [Figure 3A] FIG. 3A is a configuration diagram of the cleaning scheduling process executed by the filter basin cleaning scheduling system according to the first embodiment. [Figure 3B] FIG. 3B is a diagram showing the calculation formula. [Figure 4] Figure 4 is a graph showing a comparison between the predicted values ​​by the head loss prediction model and the measured head loss values. [Figure 5] FIG. 5 is a flowchart showing the processing flow of a first example of a program for constructing a prediction model. [Figure 6] FIG. 6 is a diagram for explaining a second example of the program 30 for constructing a head loss prediction model. [Figure 7] Figure 7 is a graph showing the head loss prediction results. [Figure 8] FIG. 8 is a diagram for explaining a method for setting the forgetting factor. [Figure 9] FIG. 9 is a flowchart showing the processing flow of a second example of the program for constructing a prediction model. [Figure 10A] FIG. 10A is a diagram for explaining a third example of a program for performing processing to construct a head loss prediction model. [Figure 10B] FIG. 10B is a diagram showing the calculation formula. [Figure 11] Figure 11 is a graph showing the head loss prediction results. [Figure 12] FIG. 12 is a flowchart showing the processing flow of a third example of the program for constructing a prediction model. [Figure 13] FIG. 13 is a diagram for explaining the calculation formula. [Figure 14] FIG. 14 is a graph showing the cleaning timing of each filter basin determined by solving the optimization problem. [Figure 15] FIG. 15 is a flowchart showing the processing flow of a program for performing cleaning scheduling (determining the cleaning time). [Figure 16A] FIG. 16A is a configuration diagram of the cleaning scheduling process executed by the filter basin cleaning scheduling system according to the second embodiment. [Figure 16B] FIG. 16B is a diagram showing the calculation formula. [Figure 16C] FIG. 16C is a diagram for explaining the calculation formula. [Figure 16D]FIG. 16D is a diagram for explaining the calculation formula. [Figure 17] FIG. 17 is a flowchart showing the processing flow of a program for performing cleaning scheduling (cleaning timing determination) according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] <<First Embodiment>> A filter basin cleaning scheduling device 1 according to a first embodiment of the present invention will be described. Fig. 1 is a configuration diagram of an entire water supply system including a filter basin cleaning scheduling device 1 according to the first embodiment of the present invention. The filter basin cleaning scheduling device 1 is a system that can optimize the cleaning schedule in response to environmental changes such as the occurrence of high turbidity or changes in the operating environment such as a power shortage.

[0013] As shown in Figure 1, the entire water supply system includes a filter basin cleaning scheduling device 1, a monitoring and control device 2, a water supply facility 3, and a water operation planning system 4. These are connected to each other via a network 5 so that they can send and receive data. That is, the filter basin cleaning scheduling device 1, the monitoring and control device 2, the water supply facility 3, and the water operation planning system 4 are connected to the network 5, and they can exchange data with each other through communication.

[0014] Figure 2 shows details (one example) of a water supply facility 3. As shown in Figure 2, one example of the water supply facility 3 includes a water intake plant 201, a water intake pump 202, a receiving well 203, a sedimentation tank 204, a filter 205, a filter 206, a filter 207, a purified water well 208, a distribution tank 209, an intake water transmission pipe connecting them, and a water distribution pipe network 210 for distributing water to consumers. Note that while Figure 2 shows a case where there are three filters, the number of filters may be more or less than this number.

[0015] Raw water taken from the river is sent to a settling tank 204 after its volume is adjusted in a receiving well 203. Here, a coagulant is injected to settle impurities. The settled water is sent to filters 205 to 207, where fine suspended solids are removed from the treated water by passing the treated water through a filter medium (sand layer). The filtered water (purified water) is sent to a distribution tank 209 via a purified water well 208. Purified water is supplied to consumers from the distribution tank 209 through a distribution pipe network 210. The filters 205 to 207 require regular cleaning to maintain their filtering performance. Treated water is filtered from the top of the filter medium downwards, but cleaning is done in the opposite direction (backflow cleaning).

[0016] The intake plant 201 and the sedimentation tank 204 are equipped with sensors that measure the turbidity of the raw water and treated water, the filtration tanks 205 to 207 are each equipped with a water level meter to measure head loss and a turbidity meter to measure the turbidity of the filtered water, each pipeline is equipped with a flow meter to measure the amount of water intake, filtration volume, etc., and the distribution tank 209 is equipped with a water level meter and a flow meter to measure the amount of water distributed, and the measurement data is sent to the monitoring and control device 2 at a predetermined interval.

[0017] Based on this data, the monitoring and control device 2 controls and manages the flow rate of each pipeline, pump operation, and water level in the distribution reservoir. The water operation planning system 4 obtains information such as water level and distribution volume required for future planning of water volume from the monitoring and control device 2, and calculates future planned values ​​for each pipeline flow rate and pump operation (ON / OFF information) based on future predictions of distribution volume. The planned values ​​are sent to the monitoring and control device 2, and the pipeline flow rate and pump operation are controlled based on the planned values. The filter basin cleaning scheduling device 1 obtains and uses information on filtration volume measurement values, head loss measurement values, turbidity measurement values, and future planned filtration volume values ​​required for cleaning scheduling (determining cleaning timing) from the monitoring and control device 2 and the water operation planning system 4.

[0018] The filter basin cleaning scheduling device 1 determines the optimal cleaning timing for each filter basin. The filter basin cleaning scheduling device 1 handles the head loss of each filter basin (the difference in head before and after the filter material) and the continuous operation time of the filter basin after cleaning as factors for requesting cleaning. For the former, the timing for cleaning is determined to be the timing when the head loss reaches a predetermined value (e.g., 1.1 m), and for the latter, the timing for cleaning is determined to be the timing when the continuous operation time reaches a predetermined value (e.g., 60 hours). Of the two timings, the earlier one is determined to be the cleaning timing.

[0019] If the cleaning timing for each filter basin is determined simply in accordance with this requirement, there may be cases where cleaning is repeated or occurs during peak power hours. Because a large amount of water is required for cleaning, repeated cleaning can cause problems such as a drop in the water level in the purified water well 208 and the distribution reservoir 209. Furthermore, because pumps are used to transport the cleaning water, which consumes electricity, it is desirable to perform cleaning outside of peak power hours. From these perspectives, it is necessary to determine the cleaning timing taking into account the water level (or cleaning interval) and power constraints. To achieve this, rather than simply determining the cleaning timing for the single filter basin that needs to be cleaned next, for example, if there are eight basins, it is necessary to determine the cleaning timing for all eight basins at the current time (i.e., perform cleaning timing scheduling), and then perform cleaning according to that plan. This is done by the filter basin cleaning scheduling device 1.

[0020] The filter basin cleaning scheduling device 1 will be described in detail below. As shown in FIG. 1, the system 1 includes a storage device 13 including RAM, a hard disk, etc., a communication I / F 14 for exchanging data via an external network, a control unit 15 including a CPU, etc., a data input device 16 including a man-machine interface such as a keyboard and a mouse, and a display device 17 such as a monitor. The control unit 15 calls and executes a cleaning schedule formulation program 11 stored in the storage device 13, stores the calculation results in a database 12, and displays them on the display device 17. During program execution, data imported from the data input device 16 and various information stored in the database 12 in the storage device are referenced as needed and used in the calculation process. For convenience, the device (computer) including the control unit 15 (CPU), storage device 13, communication I / F 14, etc. may also be referred to as an "information processing device." Therefore, the filter basin cleaning scheduling device 1 can also be said to include an information processing device.

[0021] 3A shows the processing configuration of the cleaning schedule formulation program 11. The cleaning schedule formulation program 11 is made up of a program 30 that performs head loss prediction model construction processing, a program 31 that performs predetermined head loss arrival time prediction processing, a program 32 that performs standard cleaning timing determination processing, and a program 33 that performs cleaning timing determination processing. The control unit 15 (CPU) can execute (realize) each process (each function) by executing each of these programs.

[0022] Program 30 constructs a model by identifying the model coefficients of the head loss prediction model. Model construction is performed for each filter basin. Program 31 uses the head loss prediction model to predict the time when the future head loss of each filter basin (e.g., eight filters) will reach a predetermined value (e.g., 1.1 m). Program 32 sets the reference cleaning time TB as the earlier of the time T1 when the predetermined head loss is reached for each filter basin determined by Program 31 and the time T2 when the continuous operation time since the previous cleaning for each filter basin reaches a predetermined value (e.g., 60 hours). Program 33 determines the cleaning time for each filter basin that satisfies the water level constraints and power constraints while satisfying the reference cleaning time as much as possible by solving a planning problem (optimization problem), and provides guidance to the user (the operator of the water facility) based on the determined time.

[0023] <First example of Program 30> First, we will explain the details of the processing of the first example of the program 30. In this program 30, the model coefficients k and F(0) of the head loss prediction model are identified in equations (1) and (2) in Figure 3B and stored in a predetermined area of ​​the database 12.

[0024] In equations (1) and (2), t is time (the time when cleaning is completed is set to 0), P is head loss, Q is filtration volume, F is resistance coefficient, and k is coefficient (corresponding to the head loss rise speed). In equations (1) and (2), coefficients k and F(0) can be estimated by calculating F(t) = P(t) / Q(t) using the past filtration volume measurement value Q(t) and head loss measurement value P(t), and then applying the time series of F(t) and Q(t) to equation (2) using the least squares method. The identified model coefficients are stored in database 12 of the storage means. This process is performed every 30 minutes using the latest measurement data. Figure 4 shows a comparison of the head loss prediction value 43 (dotted line) and the actual value 42 (solid line) using the model. It can be seen that the two values ​​match with high accuracy.

[0025] 5 is a flowchart showing the processing flow of a first example of the program 30 for constructing a prediction model. The control unit 15 (CPU) executes this processing flow at a predetermined cycle (for example, every 30 minutes) by executing the first example of the program 30.

[0026] The control unit 15 starts the process from step 1200, and executes the processes of steps 1201 and 1202 described below in order, and then proceeds to step 1295 to temporarily end this processing flow.

[0027] Step 1201: The control unit 15 reads out from the database 12 information on the measured filtration volume and the measured head loss required for model construction.

[0028] Step 1202: The control unit 15 uses the measurement data to identify the model coefficients k and F(0) of equations (1) and (2) and stores them in the database 12. The stored model coefficients are used to predict head loss for determining the timing of cleaning.

[0029] <Second example of Program 30> A second example of program 30 for constructing a head loss prediction model will be described with reference to Figure 6. In this second example of program 30, a model with the same structure as the previously described model is constructed and utilized, as shown in block 51 in Figure 6, but in this example a different method is used to estimate the model coefficient k. The model coefficient k is identified by applying the least squares method with a forgetting factor to the measurement data of the filtration volume and head loss after cleaning. Because the model coefficient k is considered a parameter that depends on the turbidity of the raw water and treated water, the accuracy of head loss prediction cannot be maintained unless an appropriate value is determined and utilized in response to external environmental changes such as heavy rain. The least squares method with a forgetting factor is used to identify an appropriate model coefficient k that can adapt to environmental changes.

[0030] This calculation formula is shown in block 52 in Figure 6. Among the variables in block 52, t represents time, time 0 represents the time when the most recent cleaning was completed, and time T0 represents the current time. The evaluation function JN is the sum of the squares of the differences between the estimated and actual values ​​of the resistance coefficient F (the actual value is found by dividing the measured head loss value by the measured filtration volume value) from immediately after the completion of the most recent cleaning until the present, weighted by a constant ρ.

[0031] The constant ρ, known as the forgetting factor, is set to a value less than 1 but close to 1, such as 0.97 to 0.99. Block 52 uses the bottom equation in the block to find the model parameters (k, F(0)) that minimize the evaluation function JN. This means that the model parameter (k, F(0)) estimation method places greater weight on the most recent data, emphasizing the most recent fitting. When model parameters fluctuate in response to changes in the external environment, this method is effective for finding appropriate parameters that adapt to the latest environment. The second example of program 30 applies the most recent estimated values ​​k and F(0) to the model in block 51 to make a prediction, enabling highly accurate head loss prediction even during high turbidity conditions. The calculated model parameters are stored in database 12 and can be referenced and used as needed.

[0032] Figure 7 shows the head loss prediction results using this prediction method. ρ was set to 0.97, and a prediction model was constructed using measurements of filtration rate and head loss at one-minute intervals. For the model, predictions were made using the model. In Figure 7, time 40 is the current time. Each hour corresponds to one hour. Figure 7 shows the filtration rate (601) (actual value before time 40, planned value after 40), the actual head loss (measured value) (602), the head loss estimate by the model (603), the head loss prediction by the model (604), the estimated coefficient k (605), and the measured turbidity of the sedimentation treatment water (606). While the model estimate (603) and the actual value (602) differ in the past, they are consistent for the most recent data, demonstrating that parameter estimation (model identification) emphasizes fitting to the most recent data. The estimated coefficient k (605) also changes in response to turbidity changes. Using the most recent estimated coefficient k for predictions is expected to improve accuracy. The coefficient k can be estimated stably by varying the value of ρ in the prediction method of FIG. 6 according to the turbidity change over the most recent (past) hour, as shown in FIG. 8. The absolute value of the difference between the most recent turbidity measurement value and the turbidity measurement value from one hour earlier is calculated, and if this difference is less than 0.5, ρ is set to 0.99; if it is 0.5 or greater but less than 1, ρ is set to 0.98; and if it is 1.0 or greater, ρ is set to 0.97. The smaller the fluctuation, the larger the ρ value to be set (in other words, the larger the fluctuation, the smaller the ρ value to be set). If estimation is performed using a fixed value of ρ = 0.97, the estimated value of coefficient k may fluctuate slightly (become unstable) when the turbidity is constant. To avoid this, the ρ setting shown in FIG. 8 is effective.

[0033] 9 is a flowchart showing the processing flow of a second example of the program 30 for constructing a prediction model. The control unit 15 (CPU) executes this processing flow at a predetermined cycle (for example, every 30 minutes) by executing the second example of the program 30. The control unit 15 starts processing from step 1300 and sequentially executes the processing of steps 1301 to 1303 described below, and then proceeds to step 1395, where this processing flow is temporarily terminated.

[0034] Step 1301: The control unit 15 reads out from the database 12 information on the filtration rate measurement value, head loss measurement value, and turbidity measurement value required for model construction.

[0035] Step 1302: The control unit 15 sets the forgetting factor ρ based on the amount of change in turbidity over the past hour according to FIG.

[0036] Step 1303: The control unit 15 uses the set forgetting factor to identify the model coefficient k and F(0) based on the arithmetic expression in block 52, and stores them in the database 12. The stored model coefficient k is used to predict head loss for determining the timing of cleaning.

[0037] <Third example of Program 30> A third example of the program 30 for carrying out the process of constructing a head loss prediction model will be described with reference to Fig. 10A. The third example performs more advanced head loss prediction by also using a learning function. In this method, a prediction model 81 with a similar structure to that described above is constructed and utilized. The method is characterized in that the characteristics of the model parameter coefficient k are obtained by learning, and these characteristics are incorporated into the prediction model to perform predictions, and in that a correction coefficient k1 is provided to compensate for model errors.

[0038] First, we will explain how to model the characteristics of the coefficient k. The modeling is performed by the processes of blocks 83, 84, and 85 in Fig. 10A. In block 83, the model coefficient k in the above-mentioned equations (1) and (2) is estimated (identified) using measurement data of the filtration rate and head loss at one-minute intervals over a predetermined past time period (for example, the past 30 minutes).

[0039] As mentioned above, the resistance coefficient F(t) is calculated by dividing the measured head loss P(t) by the measured filtration rate Q(t). This calculation is performed for each time period over the past 30 minutes to obtain a time series of the resistance coefficient F(t) in one-minute increments. This time series and the measured filtration rate time series Q(t) are applied to equation (2) to estimate the value of the coefficient k. The least squares method is used for the estimation. The estimated value of the coefficient k can be considered as the average value over the past 30 minutes.

[0040] The processing in block 83 is executed every 30 minutes. In block 84, the estimated coefficient k is paired with the average value of the turbidity measurement values ​​during the 30-minute period for which the coefficient was calculated, and stored in the database 12. In block 85, accumulated data from a predetermined period in the past (e.g., the past one or two months) is used to model the characteristics of the model coefficient k, i.e., the relationship between the coefficient k and the turbidity x. A nonlinear regression technique such as Gaussian process regression is used for the modeling. The processing in block 85 is executed every 30 minutes. The model coefficient k is modeled as a function k(x) of the turbidity x, and the characteristics are incorporated into the prediction model 81. In this way, by incorporating the characteristic equation for turbidity into the model coefficient, the responsiveness of head loss prediction to turbidity changes can be improved. The latest calculated characteristic function k(x) is stored in the database 12 and can be referenced and used as needed.

[0041] The error compensation coefficient k1 is determined by processing in block 82. Even if the characteristics of coefficient k are learned and incorporated into the prediction model, the model as a whole may still contain an error, and this error is compensated for by coefficient k1. The least squares method with a forgetting factor, similar to that described above, is applied to estimate k1 in the equation in block 82. In this method, Θ, y, and z are set in block 52 as shown in the equations in block BR10 in FIG. 10B, and coefficient k1 and F(0) are estimated using the equations at the bottom of block 52 in FIG. 6.

[0042] The processing of block 82 is executed every 30 minutes. The latest values ​​of k1 and F(0) are applied to the prediction model 81 to make a prediction. The latest values ​​of k1 and F(0) are stored in the database 12 and are referenced and used as needed.

[0043] Figure 11 shows the head loss prediction results using this prediction method. In Figure 11, time 40 is the current time. In Figure 11, the filtration rate is 901 (actual values ​​before time 40, planned values ​​after time 40), the actual head loss (measured value) is 902, the head loss estimate by the model is 903, the head loss prediction by the model is 904, the estimated coefficient k1 × k(x) is 905, and the turbidity of the sedimentation treated water is 906. It can be seen that the model-estimated value 903 and the actual value 902 are generally consistent from the past to the present, maintaining the accuracy of the model. Furthermore, the response speed of coefficient 905 matches that of turbidity 906, demonstrating improved responsiveness of head loss prediction compared to the adaptive prediction method in Figure 6. It is expected that the accuracy of head loss prediction will be maintained even in the face of sudden changes in turbidity.

[0044] 12 is a flowchart showing the processing flow of a third example of the program 30 for constructing a prediction model. The control unit 15 (CPU) executes this processing flow at a predetermined cycle (for example, every 30 minutes) by executing the third example of the program 30. The control unit 15 starts processing from step 1400 and sequentially executes the processing of steps 1401 to 1406 described below, and then proceeds to step 1495 to temporarily end this processing flow.

[0045] Step 1401: The control unit 15 reads out from the database 12 information on the measured filtration volume and the measured head loss required for model construction. Step 1402: The control unit 15 identifies the model coefficient k using the processing of block 83.

[0046] Step 1403: The control unit 15 reads out the turbidity measurement value from the database and calculates the average turbidity value for the most recent 30 minutes.

[0047] Step 1404: The control unit 15 stores the model coefficient k and the average turbidity value in pairs in a predetermined area of ​​the database 12 (corresponding to the process of block 84).

[0048] Step 1405: The control unit 15 performs the process of block 85, models the characteristics of the model coefficient k as a function k(x) of the turbidity x, and stores it in the database.

[0049] Step 1406: The control unit 15 identifies the model error compensation coefficient k1 and stores it in the database using the processing of block 82. The stored model coefficient k1 and characteristic function k(x) are used to predict head loss for determining the timing of cleaning.

[0050] <Processing of Program 31> Next, we will explain the details of the processing of program 31, which predicts the time when the future head loss of each filter basin (e.g., eight filter basins) will reach a predetermined value (e.g., 1.1 m) using the head loss prediction model constructed by program 30. In order to predict the future head loss P using the head loss prediction model described above (see equations (1), (2), Figures 6 and 10A), the future filtration volume Q is required. As the future filtration volume Q, the planned value of the filtration volume of each filter basin calculated and output by the water operation planning system 4 is used. Water operation planning systems generally calculate planned values ​​up to 24 hours in advance, so values ​​up to 24 hours in advance can be obtained.

[0051] The predicted filtration rate Qp(t) for 24 hours or more (t≧24) is assumed to be maintained at the value Qp(24) for 24 hours from now. That is, Qp(t) = Qp(24)(t≧24). Using this future filtration rate and the head loss prediction model described above, it is possible to predict the time when the future head loss will reach a predetermined value (e.g., 1.1 m). Specifically, the time is increased by one hour at a time, such as one hour, two hours, etc., and the predicted head loss value for each future time is calculated. The time immediately before the predicted head loss value exceeds the predetermined value is defined as the arrival time. The above process is performed for each filter basin, and the time when the head loss of each filter basin will reach the predetermined value is predicted. The latest arrival time for each filter basin is stored in database 12.

[0052] If the water operation planning system only calculates the planned value of the total filtration volume, rather than the filtration volume of each individual filter basin, the latest filtration volume measurement value of each filter basin is used, and the total filtration volume is decomposed using each measurement value as a weight to calculate the filtration volume of each filter basin.If the latest filtration volume measurement values ​​of the four filter basins are Qm1, Qm2, Qm3, and Qm4, and the planned total filtration volume is Q(t) (t: time, current time is 0), the future filtration volume Qi(t) (i=1, 2, 3, 4) of each filter basin is calculated using the following formula. Qi(t)=Qmi / (Qm1+Qm2+Qm3+Qm4)·Q(t)

[0053] If there is no planned value for the filtration rate, the most recent measured value of each filter basin is assumed to continue into the future, and the future filtration rate of each filter basin is set and used. That is, the filtration rate is set using the following formula: Qi(t)=Qmi

[0054] <Program 32> Next, we will explain the process for determining the reference cleaning timing performed by program 32. As described above, this process sets the reference cleaning timing TB to the earlier of the time T1 at which each filter basin reaches the predetermined head loss determined by program 31 and the time T2 at which the continuous operation time since the previous cleaning of each filter basin reaches a predetermined value (for example, 60 hours).

[0055] <Program 33> Next, we will explain the process of determining the cleaning timing for each filter basin, which is performed by program 33. Here, the cleaning timing for each filter basin is determined by solving a mathematical programming problem. The mathematical programming problem will be explained assuming that there are eight filters. The mathematical programming problem consists of decision variables, an objective function, and constraints. The decision variables are the cleaning times Ti (i = 1, 2, ..., 8) for the eight filters (filter basin No. 1 to 8). Ti is the cleaning time for filter basin No. i. The reference cleaning time determined by the processing of program 32 is Tbi ​​(i = 1, 2, ..., 8). Tbi is the reference cleaning time for filter basin No. i.

[0056] Here, for simplicity, we assume that Tb1≦Tb2≦···≦Tb8. This assumption does not impair generality. The objective function J is defined by equation (3) in Figure 13.

[0057] The objective function is called a penalty function, and it becomes larger as the cleaning time deviates from the reference time. Any other function may be used as long as it has this characteristic. Equations (4) and (5) in Figure 13 are used as constraints.

[0058] In equation (5), Tmin is the minimum cleaning interval between the two filters, and Tmax is the maximum cleaning interval between the two filters. If the maximum continuous operation time is 60 hours, the average cleaning interval for the eight filters is 60 / 8 = 7.5 hours. Tmin = 7.5 - 2.5 = approximately 5 hours, and Tmax = 7.5 + 2.5 = approximately 10 hours. Equation (4) corresponds to the constraint for keeping the head loss below a predetermined level, while equation (5) is the constraint for ensuring the cleaning interval between filters to restore the water levels in the clean water reservoir and distribution reservoir that have dropped due to cleaning. Under the constraints of equations (4) and (5), the cleaning time Ti for each filter is determined to minimize the objective function of equation (3), and the information is stored in database 12. In the processing performed by program 33, the cleaning time for each filter is determined so as to be as close to the reference cleaning time as possible while ensuring the cleaning interval between each filter. This reduces the number of cleanings. The processes of the above programs 31 to 33 are executed continuously at a predetermined cycle (every 30 minutes or every hour).

[0059] FIG. 14 shows an example of determined cleaning times (cleaning periods). In FIG. 14, the current time is set to 0. The cleaning times (cleaning periods) for filtration basins No. 1 to No. 8 are shown in the upper part 100 of FIG. 14. The cleaning time for filtration basin No. 1 is 6:00 (6 hours later), the cleaning time for filtration basin No. 2 is 11:00 (11 hours later), ..., the cleaning time for filtration basin No. 8 is 52:00 (52 hours later). Starting from the current time 0, the predicted head loss up to the cleaning time for each filtration basin is shown as graphs (lines) 101 to 108. Line 101 shows the predicted head loss up to the cleaning time for filtration basin No. 1. Line 102 shows the predicted head loss up to the cleaning time for filtration basin No. 2. Line 103 shows the predicted head loss up to the cleaning time for filtration basin No. 3. Line 104 shows the predicted head loss up to the cleaning time for filtration basin No. 4. Line 105 shows the predicted head loss up to the cleaning time for filter No. 5. Line 106 shows the predicted head loss up to the cleaning time for filter No. 6. Line 107 shows the predicted head loss up to the cleaning time for filter No. 7. Line 108 shows the predicted head loss up to the cleaning time for filter No. 8. As mentioned above, these can be calculated using the head loss prediction model.

[0060] 15 is a flowchart showing the processing flow of programs (programs 31, 32, and 33) that perform cleaning scheduling (cleaning time determination). The control unit 15 (CPU) executes the programs to execute this processing flow at predetermined intervals (for example, every 30 minutes, every hour, etc.).

[0061] The control unit 15 starts the process from step 1500 and executes the processes from step 1501 to step 1504 described below in order, and then proceeds to step 1595 to temporarily end this processing flow.

[0062] Step 1501: The control unit 15 reads out from the database 12 information on the head loss prediction model to be used.

[0063] Step 1502: The control unit 15 executes the processing of the program 31 to calculate the time when the predetermined head loss is reached.

[0064] Step 1503: The control unit 15 executes the processing of the program 32 to calculate the reference cleaning time for each filtration basin.

[0065] Step 1504: The control unit 15 executes the processing of the program 33 to determine the cleaning timing of each filter basin.

[0066] <Effects> As explained above, the filter basin cleaning scheduling device 1 according to the first embodiment of the present invention can predict the equivalent of the head loss increase speed adapted to environmental changes from the measurement time series of the filtration rate and head loss, and further predicts future head loss by using the planned filtration rate (future predicted filtration rate) in combination, thereby improving the accuracy of head loss prediction. As a result, the filter basin cleaning scheduling device 1 according to the first embodiment of the present invention can more accurately determine the cleaning timing for each filter basin. This filter basin cleaning scheduling device 1 can determine the effective cleaning timing for each filter basin by utilizing a highly accurate head loss prediction model. As a result, this filter basin cleaning scheduling device 1 can contribute to cost reductions, such as by reducing the number of cleanings.

[0067] <<Second embodiment>> A filter basin cleaning scheduling device 1 according to a second embodiment of the present invention will be described. The filter basin cleaning scheduling device 1 according to the second embodiment differs from the filter basin cleaning scheduling device 1 according to the first embodiment only in that the cleaning plan formulation program 11 includes a program 110 instead of the program 33, as shown in Fig. 16A. The following description will focus on this difference.

[0068] The cleaning schedule formulation program 11 in the second embodiment will be described with reference to Figure 16A. In the second embodiment, a program 110 that performs water management schedule formulation processing is provided instead of the cleaning timing decision program 33 (see Figure 3A) of the first embodiment. The processing content of programs 30, 31, and 32 other than program 110 is the same as that already described with reference to Figure 3A, etc.

[0069] In addition to the cleaning timing determination process, the program 110 also performs future planning processes such as pipeline flow rate planning, which are normally performed by the water operation planning system 4.

[0070] Overall optimization is achieved by formulating and solving a planning problem that simultaneously determines the cleaning timing and the planned flow rate. This makes it possible to calculate more effective future planned values. The planning problem is formulated as follows:

[0071] As decision variables, in addition to the flow rate of each pipeline and pump operation information (ON / OFF information), which are the decision variables of the original water management planning problem, the following decision variables related to the timing of cleaning the filter basin are used. Filter basin 1: x1(1), x1(2), ………………., x1(Tb1) (values ​​are 0 or 1, 1 means cleaning) Filter basin 2: x2(1), x2(2), ………………., x2(Tb2) (values ​​are 0 or 1, 1 means cleaning) Filtration basin 3:

[0072] Here, x1(t) (t=1, 2, ..., Tb1) is a variable that holds information on whether or not cleaning of filter basin 1 will be performed after time t, and takes the value of 0 or 1. 0 means no cleaning, and 1 means cleaning will be performed. Tb1 and Tb2 are the reference cleaning times mentioned above. These decision variables are used in order to linearize the planning problem (to achieve a high-speed solution). Note that the decision variables are subject to constraints shown in the equations in Figure 16B.

[0073] A new objective function is constructed by adding the additional objective function shown in Figure 16C to the objective function of the original water management planning problem (for example, a function of daily electricity costs). Note that although equation (7) is an example of filter basin 1, the objective function is constructed by adding penalty functions for eight filter basins. By introducing the term in equation (7), the cleaning timing for each filter basin can be determined as close as possible to the standard cleaning timing.

[0074] As a constraint, the constraint equation shown in FIG. 16D is added to the constraints of the original water management planning problem (such as upper and lower limit constraints on the water level of the distributing reservoir). The lower limit value of the operation of the distributing reservoir (clean water reservoir), which is the supply source of the flushing water, is calculated using equation (9) to form the constraint equation. By utilizing equation (9), the lower limit of the water level immediately before flushing can be replaced from the original value L(t) to a new value Lnew_low. A value slightly larger than the original lower limit of the water level operation L is set as the value of Lnew_low. In the first embodiment, a constraint was set to ensure the flushing interval between the filtration basins, but this constraint replaces that. There is no problem if the water level is maintained even if the flushing periods are consecutive, and this constraint realizes this. Equation (10) is a constraint equation that prevents the flushing periods of each filtration basin from overlapping.

[0075] The program 110 solves the operation planning problem consisting of the above decision variables, objective function, and constraint conditions (finds the decision variables that minimize the objective function under the constraint conditions) and determines the cleaning timing for each filter basin.

[0076] 17 is a flowchart showing the processing flow of programs (programs 31, 32, and 110) for performing cleaning scheduling (determining the cleaning timing) according to the second embodiment. The control unit 15 (CPU) executes the programs to execute this processing flow at predetermined intervals (e.g., every 30 minutes or every hour). The control unit 15 starts processing from step 1600 and sequentially executes the processing of steps 1601 to 1604 described below, and then proceeds to step 1695, where this processing flow is temporarily terminated.

[0077] Step 1601: The control unit 15 reads out from the database 12 information on the head loss prediction model to be used.

[0078] Step 1602: The control unit 15 executes the processing of the program 31 to calculate the predetermined head loss arrival time.

[0079] Step 1603: The control unit 15 executes the processing of the program 32 to calculate the reference cleaning time for each filtration basin.

[0080] Step 1604: The control unit 15 executes the processing of the program 110 to determine a water operation plan (i.e., the planned values ​​for water intake, water conveyance, and water delivery of the water intake and delivery system) including the cleaning timing of each filter basin and the pipeline flow rate to be performed by the water operation planning system 4.

[0081] <Effects> As explained above, the filter basin cleaning scheduling device 1 according to the second embodiment of the present invention can more accurately determine the cleaning timing for each filter basin, as in the first embodiment. This filter basin cleaning scheduling device 1 utilizes a highly accurate head loss prediction model to determine the effective cleaning timing for each filter basin, and can determine a water operation plan including the cleaning timing for each filter basin and the pipeline flow rate to be executed by the water operation planning system 4. As a result, this filter basin cleaning scheduling device 1 can contribute to cost reduction.

[0082] <<Modifications>> The present invention is not limited to the above-described embodiments, and various modifications can be adopted within the scope of the present invention. Furthermore, the above-described embodiments can be combined with each other without departing from the scope of the present invention.

[0083] The present invention can also have the following configuration.

[0084] [1] A filter basin cleaning scheduling device that is applied to a water purification facility having a plurality of filter basins and includes an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It was configured as follows: Filter basin cleaning scheduling device. [2] In the filter basin cleaning scheduling device according to [1], The information processing device includes: an evaluation value is calculated by summing, for each time, the squares of the differences between the head loss predicted value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time, and determining the rising speed equivalent amount so that the evaluation value is minimized; It was configured as follows: Filter basin cleaning scheduling device.

[0085] [3] In the filter basin cleaning scheduling device according to [1], The information processing device includes: using a fixed weighting coefficient value less than 1, calculate an evaluation value by summing up for each time the values ​​obtained by multiplying the square of the difference between the head loss prediction value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time by a coefficient calculated by raising the weighting coefficient value by a power value according to the time, the coefficient becoming smaller the earlier the time, and determine the rising speed equivalent amount so that the evaluation value is minimized. It was configured as follows: Filter basin cleaning scheduling device.

[0086] [4] In the filter basin cleaning scheduling device according to [3], The measurement values ​​include turbidity measurement values ​​of each of the filtration basins measured at a predetermined period in each of the filtration basins, The information processing device includes: The weighting coefficient value is changed based on an absolute value of a turbidity change amount for a predetermined period in the past based on the turbidity measurement value. Composed of Filter basin cleaning scheduling device.

[0087] [5] In the filter basin cleaning scheduling device according to [4], The information processing device includes: The weighting coefficient value is changed so that the smaller the absolute value of the amount of change in turbidity over the predetermined period in the past, the larger the weighting coefficient value. It was configured as follows: Filter basin cleaning scheduling device.

[0088] [6] In the filter basin cleaning scheduling device according to [1], The measurement values ​​include turbidity measurement values ​​of each of the filtration basins measured at a predetermined period in each of the filtration basins, The information processing device includes: a process of acquiring a relational expression representing the relationship between the turbidity measurement value and the head loss rise speed equivalent amount predicted in the past by learning the relationship; A process of calculating an amount equivalent to the rising speed of the head loss using the relational expression; To execute It was configured as follows: Filter basin cleaning scheduling device.

[0089] [7] [1] to [6], the filter basin cleaning scheduling device according to any one of [1] to [6], the information processing device is configured to receive a planned value of the filtration amount from an external device, The information processing device includes: The planned value of the filtration amount is used as a future predicted value of the filtration amount. It was configured as follows: Filter basin cleaning scheduling device.

[0090] [8] [1] to [6], the filter basin cleaning scheduling device according to any one of [1] to [6], The information processing device includes: a filtration amount based on the filtration amount measurement value is used as a future predicted value of the filtration amount; It was configured as follows: Filter basin cleaning scheduling device.

[0091] [9] In the filter basin cleaning scheduling device according to any one of [1] to [8], The information processing device includes: A standard cleaning time, which is a desirable cleaning time for each of the filtration basins, is calculated based on information including the arrival time calculated for each of the filtration basins, a penalty function is set whose value increases as the cleaning time for each of the filtration basins deviates from the standard cleaning time, and the cleaning time for each of the filtration basins is determined so that the sum of the penalty function values ​​for each of the filtration basins is minimized under predetermined constraint conditions. It was configured as follows: Filter basin cleaning scheduling device.

[0092]

[10] [1] to [9], the filter basin cleaning scheduling device according to any one of [1] to [9], The information processing device includes: A first objective function is provided for achieving optimal operation of a water intake and conveyance system applied to the water purification treatment facility; A standard cleaning time, which is a desirable cleaning time for each of the filtration basins, is calculated based on information including the arrival time calculated for each of the filtration basins, and a penalty function is set whose value increases as the cleaning time for each of the filtration basins deviates from the standard cleaning time. A second objective function is established by adding the first objective function to the sum of the penalty function values, and the cleaning timing of the filtration basin and the planned values ​​of the water intake, conveyance and water conveyance of the water intake and conveyance system are calculated so as to minimize the second objective function under predetermined constraints including upper and lower limit constraints of the distribution reservoir in the water intake and conveyance system. It was configured as follows: Filter basin cleaning scheduling device.

[0093]

[11] It is applied to water purification plants with multiple filtration basins, A filter cleaning scheduling system including an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined cycle in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It was configured as follows: Filtration basin cleaning scheduling system.

[0094]

[12] A filter basin cleaning scheduling method that is applied to a water purification facility having a plurality of filter basins and uses an information processing device to which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period are input, comprising: By the information processing device, The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; Filter basin cleaning scheduling method. [Explanation of symbols]

[0095] 1...Filter basin cleaning scheduling device, 2...Monitoring and control device, 3...Water supply facility, 4...Water operation planning system, 11...Cleaning plan development program, 12...Database, 13...Storage device, 15...Control unit, 16...Data input device, 17...Display device, 30, 31, 32, 33, 110...Program

Claims

1. A filter basin cleaning scheduling device that is applied to a water purification facility having a plurality of filter basins and includes an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It is configured as follows: The information processing device includes: an evaluation value is calculated by summing, for each time, the squares of the differences between the head loss predicted value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time, and determining the rising speed equivalent amount so that the evaluation value is minimized; It was configured as follows: Filter basin cleaning scheduling device.

2. A filter basin cleaning scheduling device that is applied to a water purification facility having a plurality of filter basins and includes an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It is configured as follows: The information processing device includes: a fixed weighting coefficient value less than 1 is used to calculate an evaluation value by summing up for each time the values ​​obtained by multiplying the square of the difference between the head loss prediction value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time by a coefficient calculated by raising the weighting coefficient value by a power value according to the time, the coefficient becoming smaller the earlier the time is, and the rising speed equivalent amount is determined so that the evaluation value is minimized; It was configured as follows: Filter basin cleaning scheduling device.

3. The filter basin cleaning scheduling device according to claim 2, The measurement values ​​include turbidity measurement values ​​of each of the filtration basins measured at a predetermined period in each of the filtration basins, The information processing device includes: changing the weighting coefficient value based on the absolute value of the amount of change in turbidity over a predetermined period of time in the past based on the turbidity measurement value; It was configured as follows: Filter basin cleaning scheduling device.

4. The filter basin cleaning scheduling device according to claim 3, The information processing device includes: The weighting coefficient value is changed so that the smaller the absolute value of the amount of change in turbidity over the predetermined period in the past, the larger the weighting coefficient value. It was configured as follows: Filter basin cleaning scheduling device.

5. A filter basin cleaning scheduling device that is applied to a water purification facility having a plurality of filter basins and includes an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It is configured as follows: The measurement values ​​include turbidity measurement values ​​of each of the filtration basins measured at a predetermined period in each of the filtration basins, The information processing device includes: a process of acquiring a relational expression representing the relationship between the turbidity measurement value and the head loss rise speed equivalent amount predicted in the past by learning the relationship; A process of calculating an amount equivalent to the rising speed of the head loss using the relational expression; To execute It was configured as follows: Filter basin cleaning scheduling device.

6. The filter basin cleaning scheduling device according to claim 1, the information processing device is configured to receive a planned value of the filtration amount from an external device, The information processing device includes: The planned value of the filtration amount is used as a future predicted value of the filtration amount. It was configured as follows: Filter basin cleaning scheduling device.

7. The filter basin cleaning scheduling device according to claim 1, The information processing device includes: a filtration amount based on the filtration amount measurement value is used as a future predicted value of the filtration amount; It was configured as follows: Filter basin cleaning scheduling device.

8. The filter basin cleaning scheduling device according to claim 1, The information processing device includes: A standard cleaning time, which is a desirable cleaning time for each of the filtration basins, is calculated based on information including the arrival time calculated for each of the filtration basins, a penalty function is set whose value increases as the cleaning time for each of the filtration basins deviates from the standard cleaning time, and the cleaning time for each of the filtration basins is determined so that the sum of the penalty function values ​​for each of the filtration basins is minimized under predetermined constraint conditions. It was configured as follows: Filter basin cleaning scheduling device.

9. A filter basin cleaning scheduling device that is applied to a water purification facility having a plurality of filter basins and includes an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It is configured as follows: The information processing device includes: A first objective function is provided for realizing optimal operation of a water intake and conveyance system applied to the water purification treatment facility; A standard cleaning time, which is a desirable cleaning time for each of the filtration basins, is calculated based on information including the arrival time calculated for each of the filtration basins, and a penalty function is set whose value increases as the cleaning time for each of the filtration basins deviates from the standard cleaning time. A second objective function is established by adding the first objective function to the sum of the penalty function values, and the cleaning timing of the filtration basin and the planned values ​​of the water intake, conveyance and water conveyance of the water intake and conveyance system are calculated so as to minimize the second objective function under predetermined constraints including upper and lower limit constraints of the distribution reservoir in the water intake and conveyance system. It was configured as follows: Filter basin cleaning scheduling device.

10. It is applied to water purification plants with multiple filtration basins, A filter cleaning scheduling system including an information processing device into which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined cycle in each of the filter basins are input, The information processing device includes: The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; It is configured as follows: The information processing device includes: an evaluation value is calculated by summing, for each time, the squares of the differences between the head loss predicted value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time, and determining the rising speed equivalent amount so that the evaluation value is minimized; It was configured as follows: Filtration basin cleaning scheduling system.

11. A filter basin cleaning scheduling method that is applied to a water purification facility having a plurality of filter basins and uses an information processing device to which measurement values ​​including a filtration volume measurement value and a head loss measurement value of each of the filter basins measured at a predetermined period are input, comprising: By the information processing device, The head loss increase speed equivalent for each filter is predicted from the measured values, Based on the predicted future value of the head loss increase speed equivalent and the filtration amount, predicting future head loss for each of the filtration basins, calculating the time when the predicted head loss will reach a predetermined value, and determining the timing of cleaning each of the filtration basins based on the calculated arrival time; By the information processing device, an evaluation value is calculated by summing, for each time, the squares of the differences between the head loss predicted value at each time calculated from the rising speed equivalent amount and the time series data of the filtration volume measurement value and the head loss measurement value at the same time, and determining the rising speed equivalent amount so that the evaluation value is minimized; Filter basin cleaning scheduling method.

Images