Water level prediction device, water level prediction method, and program

Abstract

To provide a water level prediction device, a water level prediction method, and a program that enable highly accurate water level prediction that contributes to improvement of pump control performance.SOLUTION: A water level prediction device of an embodiment comprises a water level prediction unit that, on the basis of at least a water level of a pump well in a rainwater drainage system, a water level in a flow channel before rainwater flows into the pump well, driving operation history information including an operation amount for a certain past time of a drainage pump installed in the pump well, and planned driving operation information including the operation amount for a certain future time of the drainage pump, predicts a water level of the pump well at a time after the predetermined time in the future.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] An embodiment of the present invention relates to a water level prediction device, a water level prediction method, and a program. [Background technology]

[0002] In order to avoid flood damage caused by the increasing number of torrential rains and localized heavy rains (so-called "guerilla downpours") in recent years, stormwater countermeasures are required. In response to this situation, the Ministry of Land, Infrastructure, Transport and Tourism is promoting flood prevention projects that use not only hard countermeasures (hardware-related measures) such as the construction of flood prevention facilities such as stormwater mains, but also soft countermeasures (software-related measures) such as the use of observation data such as rainfall information. Against this background, a soft countermeasure has been developed to date: dynamic control technology for drainage pumps (also known as "dynamic control technology for stormwater pumps") based on predictions of the amount of stormwater inflow into pump wells that store stormwater within pumping stations, which are stormwater drainage facilities.

[0003] The stormwater pump dynamic control technology predicts the amount of stormwater flowing into the pump well within the pumping station, varies the start and stop water levels of the drainage pump according to the predicted amount of stormwater inflow, and efficiently starts and stops the drainage pump according to the changing start and stop water levels.This technology is expected to reduce the risk of flooding and the number of times the pump needs to be started and stopped. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 4399122 [Patent Document 2] Patent Publication No. 2021-95711 Summary of the Invention [Problem to be solved by the invention]

[0005] However, it is difficult to directly measure the amount of rainwater inflow into a pump well using a flow meter, etc. While it is possible to create a model to estimate the amount of rainwater inflow from the measured water level of the pump well and the pump discharge rate, the estimated value is likely to contain errors, which results in a decrease in the accuracy of the prediction of the amount of rainwater inflow and further decreases the control performance of the pump.

[0006] The problem that the invention aims to solve is to provide a water level prediction device, a water level prediction method, and a program that enable highly accurate water level predictions that contribute to improving the control performance of pumps. [Means for solving the problem]

[0007] The water level prediction device of the embodiment includes: The stormwater drainage facility includes a water level prediction unit that predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of a pump well in the stormwater drainage facility, the water level of the flow path before rainwater flows into the pump well, information on an operation history including a past operation amount of a drainage pump installed at the pump well, and information on a planned operation of the drainage pump including an operation amount for a certain time in the future, and the water level prediction unit creates a model for making the prediction using at least the water level of the pump well, the water level of the flow path before rainwater flows into the pump well, and the information on the operation history, and The water level of the flow path before rainwater flows into the pump well, information on the operating operation history, and information on the planned operating operation are input into the model, and a predicted value of the water level of the pump well at a certain time in the future is output from the model. In the prediction, the operating variable used in the previous prediction is input into the model as the current operating variable. If the predicted water level value obtained from the model deviates from a certain range of thresholds, the difference between the predicted value and the threshold is converted into the inflow amount of rainwater flowing into the pump well, the operating variable used in the previous prediction is changed according to the inflow amount, and the changed operating variable is input into the model as the current operating variable. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a diagram showing an example of the configuration of a stormwater drainage system including a water level prediction device according to a first embodiment. [Figure 2A] FIG. 10 is a diagram showing an example (part 1) of information on a planned driving operation input from the outside. [Figure 2B] 2B is a diagram showing an example of information on planned driving operations that is a copy of the information on planned driving operations shown in FIG. 2A with some changes made. FIG. [Figure 3A] FIG. 10 is a diagram showing an example (part 2) of information on a planned driving operation input from the outside. [Figure 3B] 3B is a diagram showing an example of information on planned driving operations that is a copy of the information on planned driving operations shown in FIG. 3A with some changes made. FIG. [Figure 4] FIG. 2 is a diagram showing an example of a model created and used by the water level prediction unit 22. [Figure 5] FIG. 10 is a diagram showing an example of predicting the water level of the pump well 13 after a certain time period based on information on the planned operation for that time period after the prediction execution time and information on the operation history before the prediction execution time. [Figure 6] FIG. 3 is a diagram showing an example of the operation of the water level prediction device 20 according to the first embodiment. [Figure 7]FIG. 10 is a diagram showing an example of the configuration of a storm water drainage system including a water level prediction device according to a second embodiment. [Figure 8] FIG. 10 is a diagram showing an example of predicting the water level of the pump well 13 at a certain time after the prediction execution time by making small changes to each operation amount included in the operation history. [Figure 9] FIG. 10 is a diagram showing an example of the operation of the water level prediction device 20 according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments will be described with reference to the drawings.

[0010] [First embodiment] First, the first embodiment will be described.

[0011] (System configuration) FIG. 1 shows an example of the configuration of a storm water drainage system including a water level prediction device according to the first embodiment.

[0012] The stormwater drainage system of this embodiment includes n water level gauges 11-1, 11-2, . . . , 11-n, a water level gauge 12-1, a plurality of drainage pumps 14, a pump control device 15, a water level prediction device 20, and the like.

[0013] In a storm water drainage facility (in a pumping station), for example, storm water flowing through a main line 11 is configured to flow into a pump well 13 through an inlet culvert 12.

[0014] The trunk line 11 is provided with n water level gauges 11-1, 11-2, ..., 11-n that measure the water level in each part of the trunk line 11. In addition, a water level gauge 12-1 that measures the water level in the pump well 13 is provided at the boundary between the inflow conduit 12 and the pump well 13 or on the pump well 13 side. The water level measurement values ​​measured by the water level gauges 11-1, 11-2, ..., 11-n and 12-1 are sent to the water level prediction unit 22, which will be described later.

[0015] The pump well 13 is provided with a plurality of drainage pumps 14 for draining water stored in the pump well 13 to the outside. Although only one drainage pump is shown in FIG. 1, in reality, there are a plurality of drainage pumps. In addition, in this example, an example is shown in which the plurality of drainage pumps 14 are variable-speed drainage pumps (variable-speed pumps), but this is not limiting, and all or some of the plurality of drainage pumps 14 may be replaced with fixed-speed drainage pumps (fixed-speed pumps). Each drainage pump 14 is controlled by a pump control device 15, which will be described later.

[0016] A flow meter 14-1 is provided in the discharge pipe of each drainage pump 14 to measure the discharge rate of the drainage pump 14 from the flow rate of water discharged from the drainage pump 14. The measurement value of the discharge rate of each drainage pump 14 measured by the flow meter 14-1 is sent to the water level prediction unit 22, which will be described later.

[0017] The water level prediction device 20 is a device that predicts the future water level of the pump well 13 by performing arithmetic processing using various data, and outputs a predicted value of the water level of the pump well 13 (water level prediction value) as the result of the arithmetic processing. The water level prediction device 20 is configured as, for example, a computer. The various functions provided in the water level prediction device 20 are constructed as programs to be realized by the computer. The water level prediction value output from the water level prediction device 20 is sent to the pump control device 15.

[0018] The pump control device 15 controls the rotation speed or discharge rate of each drainage pump 14 based on the target water level value determined using the water level prediction value sent from the water level prediction device 20 and the water level of the pump well 13 measured by the water level gauge 12-1 or the like. The control by the pump control device 15 may use a stormwater pump dynamic control technique, and for example, the start and stop water levels of the drainage pump may be changed according to the water level prediction value. The operation amount of each drainage pump 14 operated by the pump control device 15 may be configured to be transmitted to the water level prediction device 20 in chronological order.

[0019] (Configuration of water level prediction device 20) The water level prediction device 20 includes an operation replicating unit 21, a water level predicting unit 22, and a storage unit 24.

[0020] The water level prediction device 20 has a function of predicting the water level of the pump well 13 at a certain time in the future based on at least the water level of the pump well 13 (e.g., the water level measured by the water level meter 12-1), the water level of the flow path before rainwater flows into the pump well 13 (e.g., the water level of the main line 11 measured by the water level meter 11-1, 11-2, ..., 11-n), information on the operation history including the past operation amount of each drainage pump 14 (e.g., the discharge rate of each drainage pump 14 measured by the flow meter 14-1, or the discharge rate of each drainage pump 14 calculated from the operation history (e.g., rotation speed) of each drainage pump 14), and information on planned operation including the operation amount of each drainage pump 14 for a certain time period in the future (e.g., information on planned operation input from an external higher-level system, etc. (hereinafter, sometimes referred to as "planned operation data")). This function is realized by the water level prediction unit 22. The operation amounts included in the information on the scheduled operation used for the prediction are the operation amounts of the drainage pumps scheduled for operation after the prediction execution time.

[0021] The water level prediction unit 22 receives inputs of the water level measured by the water level gauge 12-1, the water level of the main line 11 measured by the water level gauges 11-1, 11-2, ..., 11-n, the discharge volume of each drainage pump 14 measured by the flow meter 14-1 (the past operation volume of each drainage pump 14), and planned operation data from an external higher-level system, etc. In addition, the water level prediction unit 22 may also receive inputs of design information for sewer pipes and pumping stations, meteorological data such as rainfall, rainfall intensity, and temperature, etc.

[0022] When there is another drainage pump 14 that has the same rated output as a certain drainage pump 14, the operation duplication unit 21 has a function of generating (duplicating) information on another planned operation by swapping the operation amounts included in the information on the planned operation of both drainage pumps 14. The information on the other planned operation that has been duplicated and the information on the planned operation input from outside are both sent to the water level prediction unit 22.

[0023] However, it is not always necessary to install the operation duplication unit 21. If the operation duplication unit 21 is not provided, the planned operation information input to the water level prediction device 20 is sent directly to the water level prediction unit 22. If the operation duplication unit 21 is provided, the water level prediction unit 22 uses both the information of another duplicated planned operation and the information of a planned operation input from outside when predicting the water level.

[0024] The water level prediction unit 22 also has a function of creating a model (described later) for predicting the water level using at least the water level of the pump well 13, the water level of the flow path before rainwater flows into the pump well 13, and information on the operation history. Furthermore, the water level prediction unit 22 has a function of inputting at least the water level of the pump well 13, the water level of the flow path before rainwater flows into the pump well 13, information on the operation history, and information on the planned operation into the model, and outputting a predicted water level value of the pump well 13 at a certain time in the future from the model.

[0025] In addition, when predicting water levels using the above model, the water level prediction unit 22 inputs the operating variable used in the previous prediction into the model as the operating variable for the current prediction, and if the water level prediction value obtained from the model deviates from a certain range of threshold values ​​(such as an upper limit value), converts the difference between the predicted value and the threshold value into the inflow amount of rainwater flowing into the pump well 13, changes the operating variable used in the previous prediction according to the inflow amount, and inputs the changed operating variable into the model as the operating variable for the current prediction.

[0026] The storage unit 24 stores various information input into the water level prediction device 20 and various information used by the operation replicating unit 21 and the water level predicting unit 22 for processing.

[0027] For example, the memory unit 24 may store time series data (e.g., meteorological data such as rainfall, rainfall intensity, and temperature, water levels at various points on the main line, the operation amount of each drainage pump, and past time series data on the water level of the pump well 13) obtained sequentially in a chronological order in the form of a list, together with time series data on the operation amount of each drainage pump scheduled after the prediction execution time.

[0028] (Details of the processing by the driving operation replicating unit 21) The operation duplication unit 21 inputs information about scheduled operation operations from the predicted execution time to the predicted target time (for example, up to three minutes from the predicted execution time), and outputs the information about scheduled operation operations from the predicted execution time to the predicted target time to the water level prediction unit 22 after increasing the amount of information about scheduled operation operations by duplication. The amount of increase in the information about scheduled operation operations (the number of duplications) at this time varies depending on the situation at the time, as shown below. Two examples are given here.

[0029] 1. When there is only one set of drainage pumps with the same rated output Fig. 2A shows an example (part 1) of information on planned operation that has been input from the outside. Fig. 2B shows an example of information on planned operation that has been duplicated by changing part of the information on planned operation shown in Fig. 2A.

[0030] For example, consider the case where there are three drainage pumps 14 (a drainage pump with a manipulated variable X, a drainage pump with a manipulated variable Y, and a drainage pump with a manipulated variable Z) as shown in FIG. 2A.

[0031] Here, it is assumed that there is another drainage pump 14 (drainage pump with operation amount Y) that has the same rated output as a certain drainage pump 14 (drainage pump with operation amount X) that is the object of operation.

[0032] In this case, as shown in Figure 2B, information on a different scheduled operating operation (information on a duplicated scheduled operating operation) is generated by swapping the values ​​of the operating quantities of the drainage pump with operating quantity X and the drainage pump with operating quantity Y from the predicted execution time up to, for example, three minutes ahead.

[0033] In this way, two pieces of planned operation information (the information on the planned operation input from the outside + the information on the duplicated planned operation information) are generated based on the information on one input planned operation information. Note that if there is no drainage pump 14 with the same rated output, duplication is not performed.

[0034] 1. When there are two sets of drainage pumps with the same rated output Fig. 3A shows an example (part 2) of information on planned operation that has been input from the outside. Fig. 3B shows an example of information on planned operation that has been duplicated by changing part of the information on planned operation shown in Fig. 3A.

[0035] For example, consider a case where there are four drainage pumps 14 (a drainage pump with a manipulated variable W, a drainage pump with a manipulated variable X, a drainage pump with a manipulated variable Y, and a drainage pump with a manipulated variable Z) as shown in FIG. 3A.

[0036] Here, it is assumed that there is another drainage pump 14 (drainage pump with operation amount X) that has the same rated output as a certain drainage pump 14 (drainage pump with operation amount W) that is the object of operation, and there is also another drainage pump 14 (drainage pump with operation amount Z) that has the same rated output as a certain drainage pump 14 (drainage pump with operation amount Y) that is the object of operation.

[0037] In this case, three copies are made. First, as shown in Fig. 3B, information on another scheduled operation (information on a duplicated scheduled operation) is generated by swapping the values ​​of the operation amounts of the drainage pump with operation amount Y and the drainage pump with operation amount Z for, for example, three minutes from the predicted execution time.

[0038] The second, although not shown in the figure, is information on a different scheduled operation (information on a duplicated scheduled operation) in which the values ​​of the operation quantities of the drainage pump with operation quantity W and the drainage pump with operation quantity X are swapped, for example, up to three minutes from the predicted execution time.

[0039] The third, although not shown in the figure, is to generate information on another scheduled operation (information on a duplicated scheduled operation) by swapping the values ​​of the operation quantities for the drainage pump with operation quantity W and the drainage pump with operation quantity X up to, for example, three minutes from the predicted execution time, and by swapping the values ​​of the operation quantities for the drainage pump with operation quantity Y and the drainage pump with operation quantity Z up to, for example, three minutes from the predicted execution time.

[0040] In this way, based on the input information of one scheduled operation, information of four scheduled operation (information of the scheduled operation input from outside + information of three duplicated scheduled operation) is generated. Note that if there is no drainage pump 14 with the same rated output, duplication is not performed.

[0041] (Details of the processing of the water level prediction unit 22) The water level prediction unit 22 performs a process of creating a model for predicting the water level of the pump well 13 and a process of outputting a predicted water level value using the created model.

[0042] FIG. 4 shows an example of a model created and used by the water level prediction unit 22.

[0043] The model 22A shown in FIG. 4 has a fully connected layer with multiple intermediate layers between the input layer and the output layer, similar to a general deep learning model.

[0044] When creating the model 22A, the water level prediction unit 22 inputs into the input layer information on the operation history of the pumping station, including the pump discharge rate of each drainage pump and the rotation speed of each drainage pump, water levels at each point on the main line, meteorological information data such as rainfall, the water level of the pump well 13, and other time-synchronized time series data with time stamps, as well as design information on sewer pipes and pumping stations, and allows the model to learn.

[0045] When making a prediction, the water level prediction unit 22 inputs into the input layer information on the operation history of the pumping station, including the pump discharge rate of each drainage pump, the rotation speed of each drainage pump, and other operating variables, as described above, water levels at each point on the main line, meteorological information data such as rainfall, the water level of the pump well 13, and other time-synchronized time series data with time stamps, as well as design information for the sewer pipes and pumping stations.In addition, the water level prediction unit 22 inputs information on the planned operation for a certain period from the prediction execution time to the prediction target time (for example, up to 3 minutes from the prediction execution time) generated by the operation duplication unit 21, predicts the water level of the pump well 13 after that certain period, and outputs the water level prediction value.

[0046] When multiple scheduled operation information is input, water level prediction is performed for all of the multiple scheduled operation information, and a single predicted value is generated based on the results. At this time, a simple average or a weighted sum may be used.

[0047] FIG. 5 shows an example of predicting the water level of the pump well 13 after a certain time from information on the planned operation for a certain time after the prediction execution time and information on the operation history before the prediction execution time.

[0048] As shown in Figure 5, past time series data on meteorological data such as rainfall intensity and temperature, main water level, operation amounts X, Y, Z of each drainage pump, and water level of pump well 13 are managed using a memory unit 24, etc., along with time series data on the operation amounts of each drainage pump scheduled after the prediction execution time.

[0049] The example in Figure 5 shows how the water level of the pump well 13 at a certain time after the predicted execution time (4 minutes ahead) is predicted based on information on the planned operation of each drainage pump for a certain time after the predicted execution time (operation amounts X, Y, Z of each drainage pump from the predicted execution time up to 3 minutes ahead) and information on the history before the predicted execution time (1 minute to 30 minutes ahead).

[0050] (Example of operation) Next, an example of the operation of the water level prediction device 20 according to the first embodiment will be described with reference to FIG.

[0051] First, various data including time-series data are input from the outside into the water level prediction device 20 (step S11). At this time, data other than the planned operation data (information on the planned operation) is sent to the water level prediction unit 22 without going through the operation duplication unit 21.

[0052] Here, if there are drainage pumps 14 with the same rated output in the pumping station, the operation duplication unit 21 generates multiple pieces of planned operation information (planned operation information input from outside + one or more pieces of duplicated planned operation information) based on the information on one planned operation input to the water level prediction device 20 (planned operation data), and sends this to the water level prediction unit 22 (step S12). Note that if there are no drainage pumps 14 with the same rated output, the processing of step S12 is not performed, and the information on the planned operation input to the water level prediction device 20 is sent directly to the water level prediction unit 22.

[0053] Next, in the water level prediction unit 22, a model 22A such as that described in Fig. 4 is created (step S13). When creating the model 22A, information on the operation history of the pumping station, including the pump discharge rate of each drainage pump, the rotation speed of each drainage pump, and other operation amounts, water levels at each point on the main line, meteorological information data such as rainfall, the water level of the pump well 13, and other time-synchronized time-series data with time stamps are input to the input layer, and learning is performed.

[0054] Finally, the water level prediction unit 22 performs water level prediction using the model 22A (step S14). When the prediction is performed, data equivalent to the data input in step S13 is input to the input layer, and information on the planned operation for a certain period from the prediction execution time to the prediction target time generated by the operation duplication unit 21 is input, the water level of the pump well 13 after that certain period is predicted, and a water level prediction value is output.

[0055] Furthermore, when predicting the water level, the following procedure may be adopted as a method for determining information on the planned operation.

[0056] (1) Based on the results of the previous prediction, the operation amount of the planned operation that was optimized by each optimization method at the time of the previous prediction is input into the model as the operation amount for this time.

[0057] (2) Predict the water level. If the predicted water level does not deviate from a certain threshold (such as an upper limit), the prediction is terminated.

[0058] (3) If the predicted water level value deviates from a certain range of threshold values ​​(such as an upper limit value), the difference between the predicted value and the threshold value is converted into the amount of rainwater flowing into the pump well 13, and various optimization methods are used to change the operating variables used in the previous prediction so that they can accommodate this amount of rainwater inflow.The changed operating variables are then input into the model as the current operating variables, and the process returns to step (2) above.

[0059] According to the first embodiment, the water level is predicted using a model that predicts the water level itself, which is an actual measured value, rather than predicting the amount of rainwater inflow, which is an estimated value. This reduces errors and improves the robustness of the predicted value, thereby improving the performance of pump control. In addition, since predictions can be made using more information on planned operation, more stable predictions become possible.

[0060] As mentioned above, it is difficult to directly measure the amount of rainwater inflow, and in many cases the water level of the pump well has been recorded as data. Although this water level is proportional to the amount of rainwater inflow, it also changes due to control of pump operation, etc., so in the past, the amount of rainwater inflow was estimated by using a model to eliminate the influence of control of pump operation, etc. from the change in water level. When this estimation was performed, the change in water level was used, so the estimation results were unstable and contained many errors.

[0061] In contrast, the water level prediction device 20 of this embodiment directly predicts the water level of the pump well 13 by adding the influence of future control such as pump operation to the explanatory variables. The pump control device 15 then controls the pump using the predicted water level. When it is necessary to control the pump using the predicted inflow, the predicted change in water level can be converted into the amount of rainwater inflow and used to control the pump. In this case, the change in water level is used, but there is no need to convert the change in water level for the data used when learning the model as in the conventional method, which improves accuracy compared to the conventional method.

[0062] For example, when creating a model to predict the water level five minutes from now, the water level changes depending on the control for the five minutes from the time the prediction is made, so it is necessary to add the control history for these five minutes to the explanatory variables. Conventionally, this is data from after the time the prediction is made, i.e., future data, and since the control amount cannot be determined before the water level prediction result is available, it has not generally been added to the explanatory variables. However, when considering actual prediction, the control content can be set arbitrarily, so there is no problem in adding it to the explanatory variables. In this embodiment, this future operating data (i.e., planned operating operation data) is input as an explanatory variable at the time of prediction.

[0063] As a result, the explanatory variables used in the model are all measured data, not estimated data, and the prediction target is also water level, making it possible to make highly accurate predictions, unlike conventional models.As mentioned above, if the amount of rainwater inflow is required for pump control, the water level can be converted into the amount of rainwater inflow based on the structure of the pump well.

[0064] (Second embodiment) Next, a second embodiment will be described, focusing on the differences from the first embodiment.

[0065] (System configuration) FIG. 7 shows an example of the configuration of a storm water drainage system including a water level prediction device according to the second embodiment.

[0066] In the second embodiment, the water level prediction device 20 does not include the water level prediction unit 22, but instead includes a data padding unit 23.

[0067] The data padding unit 23 has the function of generating new operation quantities obtained by making small changes to the operation quantities included in the aforementioned operating operation history (for example, the operation quantities one minute before the predicted execution time), and new water level information obtained by applying a function of the small changes to the water level of the pump well 13.

[0068] In this case, the water level predicting unit 22 uses the new manipulated variable and the new water level information in predicting the water level.

[0069] The data padding unit 23 also has a function of setting the value of the minute change to be equal to or less than a certain threshold value.

[0070] The data padding unit 23 also has a function of externally obtaining information (padded data) indicating the amount, rate, or proportion of the minute change, or the proportion or rate at which data is padded by the minute change (padded proportion or padding rate). Here, the padded proportion or padding rate indicates what proportion of the amount of original data to be padded is newly generated, or what percentage is newly generated.

[0071] (Details of the processing of the data padding unit 23) In the case of the aforementioned model 22A, there are many explanatory variables and a large amount of data is required for learning. However, there may be cases where model 22A is desired to be used even when there is little data, or where there is a large amount of data but the variation in the amount of operation is small, preventing learning from progressing.

[0072] Therefore, in the second embodiment, the above-mentioned model 22A is used, and by making small changes to each operation amount included in the operating operation history (for example, the operation amount one minute before the predicted execution time), the water level of the pump well 13 is made small changes for a certain period from the predicted execution time to the predicted target time (for example, up to three minutes from the predicted execution time), and the water level of the pump well 13 at a time after that certain period (four minutes from now) is predicted.

[0073] FIG. 8 shows an example of a case where the water level of the pump well 13 at a certain time after the prediction execution time is predicted by making small changes to each operation amount included in the operation history.

[0074] In the example of Figure 8, if there are small changes of Δx, Δy, and Δz in the operation volume x, operation volume y, and operation volume z, respectively, in the operation history from one minute ago, for example, the extent to which the water level will change is calculated using a mathematical function model. For example, the function model showing the change in water level when there is a change of Δx in the operation volume x is expressed by the function f(Δx). Also, the function model showing the change in water level when there is a change of Δy in the operation volume y is expressed by the function g(Δy). Also, the function model showing the change in water level when there is a change of Δz in the operation volume z is expressed by the function h(Δz).

[0075] In other words, by adding minute changes Δx, Δy, Δz to the operation volume x, operation volume y, and operation volume z one minute before the prediction execution time, the water level of the pump well 13 for a certain period of time from the prediction execution time up to three minutes into the future is changed by f(Δx)+g(Δy)+h(Δz), and the water level of the pump well 13 at a time after the certain period of time (four minutes into the future) is predicted from the changed operation volumes X, Y, Z of each drainage pump up to three minutes into the future from the prediction execution time.

[0076] As described above, by adding various variations of minute changes to various types of time series data to augment the data, the augmented data can be used for learning the model 22A of the water level prediction unit 22. However, if the amount of augmentation is increased too much, the characteristics of the above-mentioned function model will be learned, which may result in a decrease in the accuracy of the model 22A. Therefore, it is desirable to confirm the ratio of the data augmentation to the actual data volume before setting it.

[0077] (Example of operation) Next, an example of the operation of the water level prediction device 20 according to the second embodiment will be described with reference to FIG.

[0078] First, various data including time-series data are input from the outside into the water level prediction device 20 (step S11). At this time, data such as inflated data (information indicating the amount, rate, or proportion of minute changes), the water level of the pump well 13, and the discharge rate of each drainage pump 14 are input to the data padding unit 23. Meanwhile, data such as design information, meteorological data such as rainfall, planned operation data (planned operation information), and water levels at each point on the main line are input to the water level prediction unit 22.

[0079] Next, the data padding unit 23 pads the data using padding data (information indicating the amount, rate, or proportion of the minute change) (step S12').

[0080] Next, in the water level prediction unit 22, a model 22A such as that described in Fig. 4 is created (step S13'). When creating the model 22A, information on the operation history of the pumping station, including the pump discharge rate of each drainage pump, the rotation speed of each drainage pump, and other operation variables, water levels at each point on the main line, meteorological information data such as rainfall, the water level of the pump well 13, and other time-synchronized time-series data with time stamps are input to the input layer, and further data required for learning is read out from the data augmentation unit 23 and input, and learning is performed.

[0081] Finally, the water level prediction unit 22 performs water level prediction using the model 22A (step S14'). When the prediction is performed, data equivalent to the data input in step S13' is input to the input layer, and data necessary for water level prediction is read and input from the data padding unit 23. The water level of the pump well 13 after a certain time is predicted, and a water level prediction value is output.

[0082] According to the second embodiment, even in a situation where there is little data to create the model 22A, or where there is a large amount of data but the variation in the operation amount is small and learning does not progress, the model 22A can be efficiently trained, thereby improving the accuracy of water level prediction and ultimately improving pump control performance.

[0083] As described above in detail, according to the embodiment, it is possible to perform highly accurate water level prediction that contributes to improving the control performance of the pump.

[0084] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0085] 11...main line, 11-1, 11-2, ..., 11-n...water level gauge, 12...inlet culvert, 12-1...water level gauge, 13...pump well, 14...drainage pump, 15...pump control device, 20...water level prediction device, 22...water level prediction unit, 23...data padding unit, 24...memory unit.

Claims

1. A water level prediction unit is provided which predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of a pump well in a stormwater drainage facility, the water level of the flow path before stormwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed in the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future, The water level prediction unit creating a model for making the prediction using at least the water level of the pump well, the water level of the flow path before the rainwater flows into the pump well, and information on the operation history; inputting at least the water level of the pump well, the water level of the flow path before rainwater flows into the pump well, information on the operation history, and information on the planned operation into the model, and outputting a predicted value of the water level of the pump well at a time after the certain time in the future from the model; In the prediction, the manipulated variable used in the previous prediction is input to the model as the manipulated variable for this time, and if the predicted value of the water level obtained from the model deviates from a certain range of thresholds, the difference between the predicted value and the threshold is converted into the inflow amount of rainwater flowing into the pump well, the manipulated variable used in the previous prediction is changed according to the inflow amount, and the changed manipulated variable is input to the model as the manipulated variable for this time. Water level prediction device.

2. A water level prediction unit that predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of a pump well in a stormwater drainage facility, the water level of the flow path before stormwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed at the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future; an operation duplication unit that generates information on another scheduled operation of the drainage pump by exchanging the operation amounts included in the information on the scheduled operation of both the drainage pumps when there is another drainage pump having the same rated output as the drainage pump; Equipped with The water level prediction unit uses both information on the other scheduled operating operation and information on the scheduled operating operation in the prediction. Water level prediction device.

3. A water level prediction unit that predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of a pump well in a stormwater drainage facility, the water level of the flow path before stormwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed at the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future; a data augmentation unit that generates new operation variables obtained by applying minute changes to the operation variables included in the operation history, and new water level information obtained by applying a function of the minute changes to the water level of the pump well; Equipped with The water level prediction unit uses information on the new manipulated variable and the new water level in the prediction. Water level prediction device.

4. the data padding unit sets the value of the small change to be equal to or less than a certain threshold value. The water level prediction device according to claim 3.

5. the data padding unit externally obtains information indicating the amount, rate, or proportion of the minute change, or the proportion or rate at which the data is padded by the minute change; The water level prediction device according to claim 3 or 4.

6. and a water level prediction unit predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of the pump well in the storm water drainage facility, the water level of the flow path before the rainwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed in the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future, Predicting the water level of the pump well Creating a model for making the prediction using at least the water level of the pump well, the water level of the flow path before rainwater flows into the pump well, and information on the operation history; inputting at least the water level of the pump well, the water level of the flow path before rainwater flows into the pump well, information on the operation history, and information on the scheduled operation into the model, and outputting a predicted value of the water level of the pump well at a time after the certain time in the future from the model; In the prediction, the manipulated variable used in the previous prediction is input to the model as the manipulated variable for this time, and when the predicted value of the water level obtained from the model deviates from a certain range of threshold, the difference between the predicted value and the threshold is converted into the inflow amount of rainwater flowing into the pump well, the manipulated variable used in the previous prediction is changed according to the inflow amount, and the changed manipulated variable is input to the model as the manipulated variable for this time; Including, Water level prediction methods.

7. The water level prediction unit predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of the pump well in the stormwater drainage facility, the water level of the flow path before stormwater flows into the pump well, information on the operating history including the past operating amount of the drainage pump installed at the pump well, and information on the planned operating operation including the operating amount of the drainage pump for a certain time in the future; an operation duplication unit, when there is another drainage pump having the same rated output as the drainage pump, generates information on another scheduled operation in which the operation amounts included in the information on the scheduled operation of both drainage pumps are swapped; Including, In the prediction, both information on the other planned driving operation and information on the planned driving operation are used. Water level prediction methods.

8. The water level prediction unit predicts the water level of the pump well at a time after a certain time in the future based on at least the water level of the pump well in the stormwater drainage facility, the water level of the flow path before stormwater flows into the pump well, information on the operating history including the past operating amount of the drainage pump installed at the pump well, and information on the planned operating operation including the operating amount of the drainage pump for a certain time in the future; generating, by a data augmentation unit, new operation amounts obtained by applying minute changes to the operation amounts included in the operation history, and new water level information obtained by applying a function of the minute changes to the water level of the pump well; Including, The prediction uses information on the new manipulated variable and the new water level. Water level prediction methods.

9. In the generation, the value of the small change is set to be equal to or less than a certain threshold. The water level prediction method according to claim 8.

10. In the generation, information indicating the amount, rate, or proportion of the minute change, or the proportion or rate at which the minute change will inflate the data, is obtained from an external source. The water level prediction method according to claim 8 or 9.

11. On the computer, A program for realizing a water level prediction function that predicts the water level of a pump well at a time after a certain time in the future based on at least the water level of a pump well in a stormwater drainage facility, the water level of a flow path before rainwater flows into the pump well, information on an operation history including past operation amounts of a drainage pump installed in the pump well, and information on planned operation operations including operation amounts of the drainage pump for a certain time in the future, The water level prediction function is creating a model for making the prediction using at least the water level of the pump well, the water level of the flow path before the rainwater flows into the pump well, and information on the operation history; inputting at least the water level of the pump well, the water level of the flow path before rainwater flows into the pump well, information on the operation history, and information on the planned operation into the model, and outputting a predicted value of the water level of the pump well at a time after the certain time in the future from the model; In the prediction, the manipulated variable used in the previous prediction is input to the model as the manipulated variable for this time, and if the predicted value of the water level obtained from the model deviates from a certain range of thresholds, the difference between the predicted value and the threshold is converted into the inflow amount of rainwater flowing into the pump well, the manipulated variable used in the previous prediction is changed according to the inflow amount, and the changed manipulated variable is input to the model as the manipulated variable for this time. program.

12. A computer comprising: a water level prediction function that predicts the water level of the pump well at a certain time in the future based on at least the water level of a pump well in the storm water drainage facility, the water level of the flow path before rainwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed in the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future; an operation duplication function that generates information on another scheduled operation of the drainage pump by exchanging the operation amounts included in the information on the scheduled operation of both the drainage pumps when there is another drainage pump having the same rated output as the drainage pump; A program for realizing the above, The water level prediction function uses both information about the other planned operating operation and information about the planned operating operation in the prediction. program.

13. A computer comprising: a water level prediction function that predicts the water level of the pump well at a certain time in the future based on at least the water level of a pump well in the storm water drainage facility, the water level of the flow path before rainwater flows into the pump well, information on the operation history including the past operation amount of the drainage pump installed in the pump well, and information on the planned operation including the operation amount of the drainage pump for a certain time in the future; a data augmentation function that generates new operation variables obtained by applying minute changes to the operation variables included in the operation history, and new water level information obtained by applying a function of the minute changes to the water level of the pump well; A program for realizing the above, The water level prediction function uses the new manipulated variable and the new water level information in the prediction. program.

14. The data padding function is set so that the value of the small change is equal to or less than a certain threshold. The program according to claim 13.

15. The data padding function externally obtains information indicating the amount, rate, or proportion of the minute change, or the proportion or rate at which the data is padded by the minute change.

15. The program according to claim 13 or 14.

Images