Management method, management program and management system

Abstract

To reduce the dependence of a manager on experience and knowledge and to reduce the load of the manager.SOLUTION: A method for managing a sewage treatment plant 1 into which water to be treated flows from a sewage pipeline 2 of a split-type sewer system, the method including a step of obtaining a predicted value of an inflow amount of the water to be treated by inputting a forecast value of a precipitation amount in at least a predetermined region into a learned model constructed using, as training data, a set of an actual value of a precipitation amount in the predetermined region where the sewage pipeline 2 is laid and an actual value of an inflow amount of the water to be treated in the sewage treatment plant 1 when the actual value of the precipitation amount is recorded.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a management method, a management program, and a management system for a sewage treatment plant into which water to be treated flows in from a sewage pipe of a separate sewer system. [Background technology]

[0002] A separate sewer system is a type of sewer system that has separate stormwater and sewage pipes. Separate sewers are designed so that rainwater that falls in the area being treated is discharged into rivers through the stormwater pipes, and the design concept does not assume that rainwater will flow into the sewage pipes. However, in reality, there is a known phenomenon where rainwater does seep into the sewage pipes, and the amount of water to be treated that flows into the sewage treatment plant via the sewage pipes increases during rainfall.

[0003] There is a demand for sewage treatment plants to standardize their operating conditions as much as possible, because large fluctuations in operating conditions can make it difficult to stably operate the biological treatment adopted by many sewage treatment plants, and can also lead to the need to provide excess capacity to prepare for rare high loads.

[0004] Conventionally, methods for equalizing the operating conditions of sewage treatment plants as much as possible, regardless of fluctuations in the inflow of water to be treated due to precipitation, have relied heavily on the experience and knowledge of plant managers. This is because predicting the behavior of precipitation, which is a natural phenomenon, is difficult, and the manner in which precipitation affects the inflow behavior of water to be treated (such as the degree of impact and the timing of the impact) varies depending on the location of the sewage treatment plant, making it difficult to develop a general-purpose control program. As prior art that takes this situation into consideration, Japanese Patent Laid-Open Publication No. 2013-185320 (Patent Document 1) discloses a pump operation plan support device that enables the selection of operation plans that reflect the operator's intentions and enables the creation of an optimal pump operation plan. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2013-185320 Summary of the Invention [Problem to be solved by the invention]

[0006] Although the invention described in Patent Document 1 reduces the burden on sewage treatment plant managers, it is the same as the technology prior to this invention in that it is based on the manager's knowledge through experience. Management methods that rely on the manager's knowledge through experience have issues such as a heavy burden on the manager in monitoring and controlling the sewage treatment plant and the difficulty of transferring the skills.

[0007] Therefore, it is desirable to realize a management method, a management program, and a management device that can reduce the administrator's dependency on experiential knowledge and reduce the administrator's burden. [Means for solving the problem]

[0008] The management method of the present invention is a management method for a sewage treatment plant into which treated water flows in from a sewage pipe of a separate sewer system, and is characterized by including a step of inputting at least a forecast value of precipitation in the specified area into a trained model constructed using as training data a pair of an actual value of precipitation in the specified area in which the sewage pipe is laid and an actual value of the inflow of treated water into the sewage treatment plant at the time the actual value of precipitation was recorded, thereby obtaining a predicted value of the inflow of treated water.

[0009] The management program of the present invention is a management program for managing a sewage treatment plant into which treated water flows in from a sewage pipe of a separate sewer system, and is characterized in that when executed by a computer, it can realize the function of inputting at least the forecast value of precipitation in the specified area into a trained model constructed using as training data a set of the actual value of precipitation in the specified area in which the sewage pipe is laid and the actual value of the inflow of treated water into the sewage treatment plant at the time the actual value of precipitation was recorded, and obtaining a predicted value of the inflow.

[0010] The management system of the present invention is a management system that manages a sewage treatment plant into which treated water flows in from a sewage pipe of a separate sewer system, and is equipped with a pumping device that sends the treated water to a water treatment facility and a control device that controls the pumping device, and is characterized in that the control device is capable of realizing the following functions: inputting at least the forecast value of precipitation in the specified area into a trained model constructed using as training data a set of actual precipitation values ​​in the specified area in which the sewage pipe is laid and actual values ​​of the inflow of treated water at the sewage treatment plant at the time the actual precipitation values ​​were recorded, thereby obtaining a predicted value of the inflow volume; and controlling the pumping device based on the predicted value of the inflow volume.

[0011] These configurations allow for highly accurate prediction of the inflow of untreated water, making it easier to forecast the operating conditions of the sewage treatment plant. Furthermore, because the precipitation forecasts on which the predictions are based are often provided by public institutions, the data is highly reliable and relatively easy to obtain. Therefore, the inflow of untreated water can be predicted easily and with high accuracy. Since the above benefits are achieved without relying too much on the experience and knowledge of the manager, the burden on the manager can be reduced.

[0012] Preferred embodiments of the present invention will be described below, but the scope of the present invention is not limited to the preferred embodiments described below.

[0013] In one aspect of the management method of the present invention, it is preferable that the specified area is a portion of the entire area in which the sewage pipeline is laid, selected based on the likelihood of rainwater infiltrating into the sewage pipeline.

[0014] One of the causes of fluctuations in the inflow of untreated water is the infiltration of rainwater into sewage pipes. According to the above configuration, by selecting an area for which the forecast value of precipitation is taken into consideration based on the likelihood of rainwater infiltration, it is possible to reduce the number of data points of the forecast value to be taken into consideration without impairing the prediction accuracy of the inflow of untreated water.

[0015] In one aspect of the management method of the present invention, it is preferable that the specified area is a portion of the entire area in which the sewage pipeline is laid, selected based on the distance from the sewage treatment plant.

[0016] Depending on the urban structure, there may be many areas near sewage treatment plants where rainwater is likely to infiltrate into sewage pipelines. In this case, the above-described configuration allows for the selection of areas for which precipitation forecast values ​​are considered based on the distance from the sewage treatment plant, thereby reducing the number of data points for forecast values ​​to be considered without compromising the prediction accuracy of the amount of inflow of treated water. For example, in coastal cities with rivers, urban areas are formed in the plains surrounding the river, and sewage treatment plants may be located near the mouths of rivers, where it is convenient to discharge treated water after purification. In such cases, densely populated areas are likely to be near sewage treatment plants. Because events that cause rainwater to infiltrate into sewage pipelines (such as aging pipes) are likely to occur in densely populated areas, densely populated areas are areas where rainwater is likely to infiltrate into sewage pipelines.

[0017] In one aspect of the management method of the present invention, it is preferable that the teaching data includes, as actual values ​​of precipitation in the specified area, a first accumulated value which is the accumulated value of precipitation over a specified first period, and a second accumulated value which is the accumulated value of precipitation over a specified second period which is longer than the first period.

[0018] This configuration allows for a trained model with particularly high prediction accuracy to be obtained.

[0019] Preferably, the management method according to the present invention further comprises the step of determining a pumping rate at the sewage treatment plant based on the predicted value of the inflow rate.

[0020] This configuration makes it easier to equalize the operating conditions of the sewage treatment plant.

[0021] Further features and advantages of the present invention will become more apparent from the following description of exemplary and non-limiting embodiments, which is given with reference to the drawings. [Brief explanation of the drawings]

[0022] [Figure 1] 1 is a partial cross-sectional view of a sewage treatment plant according to an embodiment. [Figure 2] FIG. 10 is a diagram showing an example of how precipitation data is provided. [Figure 3] 10 is a flowchart showing a method for determining a recommended pumping amount in the management method according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0023] An embodiment of a management method, a management program, and a management system for a sewage treatment plant according to the present invention will be described with reference to the drawings. In the following, an example in which the present invention is applied to the management of a sewage treatment plant 1 (FIG. 1) into which untreated water flows in from a sewage pipe 2 of a separate sewerage system will be described.

[0024] [Configuration of sewage treatment plant] First, we will explain the configuration of a sewage treatment plant 1 that is the object of management in this embodiment. The sewage treatment plant 1 is a facility that receives water to be treated from a sewage pipe 2 of a separate sewer system laid in the area to be treated (such as a local government), purifies it, and then releases it into a river (Fig. 1).

[0025] A separate sewer system has separate stormwater pipes (not shown) and sewage pipes 2. Rainwater that falls in the area to be treated is discharged into the river through the stormwater pipes. Meanwhile, sewage (water to be treated) generated in homes, businesses, etc. flows into the sewage treatment plant 1 through the sewage pipes 2, where it is purified before being discharged into the river.

[0026] As described above, the separate sewer system is designed to treat rainwater and sewage in separate systems, and the design concept does not anticipate rainwater flowing into the sewage pipeline 2. However, in reality, rainwater can seep into the sewage pipeline 2 due to events such as a pipe that should be connected to a stormwater pipeline being mistakenly connected to the sewage pipeline 2, or the sewage pipeline 2 or the piping connected to it becoming worn out. Therefore, the amount of water to be treated flowing into the sewage treatment plant 1 via the sewage pipeline 2 tends to increase during precipitation.

[0027] Within the target area, areas prone to rainwater infiltration into sewage pipeline 2 can be estimated empirically. For example, the number of incidents of misconnected stormwater pipelines and aging sewage pipelines is positively correlated with the total length of sewage pipelines, which in turn is positively correlated with the number of buildings. Therefore, areas with many buildings are prone to rainwater infiltration into sewage pipeline 2. In addition, because buildings tend to be located in areas with high population density, areas with high population density are also prone to rainwater infiltration into sewage pipeline 2. Furthermore, because the number of incidents of aging sewage pipelines is positively correlated with the age of the sewage pipeline, areas with many relatively old buildings are also prone to rainwater infiltration into sewage pipeline 2. Therefore, areas with many old buildings are typical areas prone to rainwater infiltration into sewage pipeline 2; for example, areas around train stations tend to meet this condition.

[0028] In coastal cities with rivers, urban areas are formed in the plains surrounding the river, and sewage treatment plants 1 are sometimes located near the mouths of rivers, where it is convenient to discharge treated water after purification. In such cases, densely populated areas are located relatively close to sewage treatment plant 1, which are areas prone to rainwater infiltration into sewage pipelines 2. In addition, sewage treatment plants 1 may be located near densely populated areas in order to limit the total length of sewage pipelines 2. Therefore, in these cases, it is possible to estimate areas prone to rainwater infiltration into sewage pipelines 2 based on the distance from sewage treatment plant 1.

[0029] FIG. 1 shows the components of a sewage treatment plant 1 that receives water to be treated from a wastewater pipeline 2. The sewage treatment plant 1 includes a reservoir 3 having a receiving well 31 into which the water to be treated flows from the wastewater pipeline 2, a grit chamber 32 that removes sediment from the water to be treated, and a pump well 33 that stores the water to be treated after the sediment has been removed. The water treatment facility 4 is further downstream. The water treatment facility 4 may be a known biological treatment device. The water treatment facility 4 may be located on the same site as the reservoir 3, or on a different site. In the latter case, the site where the reservoir 3 and its associated equipment are located may be referred to as a pumping station.

[0030] The sewage treatment plant 1 is equipped with, as ancillary equipment for the storage tank 3, a gate device 5 installed between the receiving well 31 and the settling tank 32, a lifting pump 6 (an example of a pumping device) that sends the treated water from the pump well 33 to the water treatment facility 4, water level gauges 7 (71, 72) that measure the water levels in the receiving well 31 and the pump well 33, and a control device 8 that controls the gate device 5 and the lifting pump 6.

[0031] Gate device 5 is a sluice gate installed between receiving well 31 and grit basin 32, and its opening degree can be controlled. The amount of untreated water that flows from receiving well 31 to grit basin 32 is positively correlated with the opening degree of gate device 5. When the opening degree of gate device 5 is relatively small and the inflow amount of untreated water is relatively large, the amount of untreated water that flows into receiving well 31 is likely to be greater than the amount that flows out of receiving well 31, and the water level in receiving well 31 is likely to rise. Furthermore, particularly in situations where the water level in receiving well 31 rises, sewage pipeline 2 may play a role in storing untreated water.

[0032] A plurality of pumping pumps 6 (five in this embodiment (not shown)) are provided, and each can be individually controlled to operate or stop. Furthermore, the output of one of the five pumping pumps 6 can be controlled by inverter control. Therefore, the amount of water pumped can be controlled by controlling the number of pumping pumps 6 to operate and the output of the inverter-controllable pumping pumps 6. The capacity of the multiple pumping pumps 6 may be the same or different. When the capacity of the multiple pumping pumps 6 is different, the desired amount of water pumped can be achieved by changing the combination of the pumping pumps 6 to operate according to the required amount of water pumped.

[0033] The water level gauges 7 are instruments that measure the water levels in the receiving well 31 and the pump well 33, with a first water level gauge 71 being provided in the receiving well 31 and a second water level gauge 72 being provided in the pump well 33. Each water level gauge 7 may be a known water level gauge such as a float type, ultrasonic type, or differential pressure type.

[0034] The control device 8 is a known computer device, and is electrically connected to the gate device 5, the water pump 6, and the water level gauges 7 (71, 72). The control device 8 is also connected to a network such as the Internet, and can obtain meteorological information data provided by public organizations such as the Japan Meteorological Agency (Japan) and the National Weather Service (USA) via the network.

[0035] The control device 8 may be a single computer device or a system composed of multiple computers. The control device 8 may be a system including, for example, a first computer installed near the gate device 5, the water pump 6, and the water level gauge 7 to control these devices, and a second computer connected to the first computer via a network to perform relatively heavy computational processing. In this case, the first computer may be a device such as a microcontroller or a programmable logic controller, and the second computer may be a device such as a server computer.

[0036] The control device 8 may also have devices that are common components of a computer, such as an output device such as a liquid crystal display, input devices such as a keyboard and a mouse, and a storage device such as a hard disk drive.

[0037] [Management system configuration] The gate device 5, water pump 6, water level gauge 7, and control device 8 described above constitute a management system 10 according to this embodiment. The management system 10 can generally achieve the functions of predicting the inflow of water to be treated at the sewage treatment plant 1, determining a recommended pumping rate for the water to be treated based on the predicted inflow rate, and determining a recommended opening degree for the gate device 5. A management program according to this embodiment is installed in the control device 8, and the above functions are achieved by the control device 8 executing the management program.

[0038] [Management method] Next, a description will be given of an embodiment of a method for managing the sewage treatment plant 1. The following description also describes the functions of the management program and the management system 10 according to this embodiment.

[0039] (1) Prediction of inflow of treated water First, a prediction is made of the inflow of water to be treated at the sewage treatment plant 1. In this embodiment, the prediction is made using a trained model in which the amount of precipitation is an explanatory variable and the inflow of water to be treated is a target variable.

[0040] The trained model used in this embodiment uses precipitation as an explanatory variable and the inflow of water to be treated as a target variable. The trained model is constructed using pairs of actual precipitation values ​​and actual inflow values ​​of water to be treated as training data. The trained model may be constructed by the management system 10 according to this embodiment, or by another computer, etc. The precipitation and inflow amounts may be used as variables themselves, or features obtained by applying any feature extraction process may be used as variables.

[0041] The actual precipitation value is the actual precipitation value in the area where the sewage pipeline 2 is installed, and is, for example, precipitation data provided by the Japan Meteorological Agency or the like. Here, the area for which the actual precipitation value is considered may be the entire area where the sewage pipeline 2 is installed (hereinafter referred to as the "treatment area"), but preferably, an area of ​​the treatment area where rainwater is likely to infiltrate the sewage pipeline 2 is selected. This is because selecting the area for which actual precipitation is considered based on the likelihood of rainwater infiltration can reduce the number of data points used when predicting the inflow of water to be treated. Reducing the number of data points is economically advantageous because it contributes to reducing the amount of calculation processing in the control device 8 and the amount of precipitation data purchased from the Japan Meteorological Agency or the like. Note that the area where the sewage pipeline 2 is installed refers to the treatment area covered by the sewage treatment plant 1 (the area where sewage generated in that area flows into the sewage treatment plant 1 via the sewage pipeline 2).

[0042] As an example, the area for which the precipitation record is considered can be selected based on the distance from the sewage treatment plant 1. For example, it is advisable to consider the precipitation record of measurement points whose distance from the sewage treatment plant 1 is equal to or less than a predetermined threshold. As mentioned above, if there is a densely populated area in an area relatively close to the sewage treatment plant 1, this area is likely to experience rainwater infiltration into the sewage pipes 2. Therefore, the amount of rainfall in this area has a greater impact on the inflow of water to be treated than the amount of rainfall in other areas. Therefore, it is appropriate to construct a trained model taking into account the precipitation record of the area.

[0043] An example of how precipitation data is provided will be described with reference to FIG. 2. FIG. 2 shows a map of the area around the sewage treatment plant 1, with precipitation data provided by the Japan Meteorological Agency for the grid intersections overlaid on the map. The grid interval is 1 km. P1 and P2 in FIG. 2 indicate points (grid intersections) for which precipitation data is provided that are in the central city of the municipality where the sewage treatment plant 1 is located. For example, these 36 points (P1 and P2 in FIG. 2) can be used as the area for considering actual precipitation data. P1 is the 9th point out of the 36 points in order of proximity to the sewage treatment plant 1, and is indicated by a square. P2 is the remaining 27 points, and is indicated by a circle.

[0044] Furthermore, of the 36 locations, only 9 locations (P1 in FIG. 2) that are particularly close to the sewage treatment plant 1 may be set as areas for considering the actual precipitation amount. A comparison between considering the actual precipitation amount for 36 locations and considering the actual precipitation amount for 9 locations will be explained in the Examples section with specific examples.

[0045] The actual precipitation value may be a numerical value for any time interval, such as an hourly precipitation value provided by the Japan Meteorological Agency, etc. Alternatively, instead of or in addition to the precipitation data provided by the Japan Meteorological Agency, etc., a value obtained by performing a predetermined calculation on the precipitation data may be used as the actual precipitation value. For example, a 6-hour precipitation value obtained by integrating the hourly precipitation values ​​provided by the Japan Meteorological Agency, etc., every 6 hours may be used as the actual precipitation value.

[0046] The actual precipitation values ​​included in the training data preferably include multiple integrated values ​​with different aggregation periods, i.e., a first integrated value that is the integrated value of precipitation over a predetermined first period, and a second integrated value that is the integrated value of precipitation over a predetermined second period that is longer than the first period. For example, it is preferable that the training data include precipitation data (an example of a first integrated value) that is the integrated value of precipitation over one hour (an example of a first period), and a 6-hour precipitation (an example of a second integrated value) that is the integrated value of precipitation over six hours (an example of a second period). An increase in the inflow of water to be treated is seen not only when there is heavy rainfall in a short period of time (such as a sudden downpour), but also when a relatively small amount of precipitation per unit time continues for a long period of time. Therefore, by constructing a trained model using training data that includes both the accumulated value of precipitation over a relatively short period (first accumulated value) and the accumulated value of precipitation over a relatively long period (second accumulated value), a trained model can be obtained that is capable of making predictions that take into account various possible modes that may be factors that cause fluctuations in the inflow of treated water, thereby improving prediction accuracy.

[0047] The actual value of the inflow amount of water to be treated can be determined based on the operation history of the sewage treatment plant 1. The following formula (1) is an example of a calculation formula for determining the inflow amount of water to be treated.

number

[0048] Equation (1) shows the balance of the water to be treated in the last hour of time t. Q in (t) is the inflow of treated water in the last hour at time t, and Q out (t) is the pumping volume for the last hour of time t. Q out (t) is determined from the actual value of the pumping amount of the water pumping pump 6.

[0049] V(t-1) and V(t) are the amounts of water to be treated stored in the sewage pipeline 2 one hour before time t and at time t, respectively (hereinafter referred to as "storage amount"). In other words, the term (V(t-1) - V(t)) represents the increase or decrease in the amount of stored water to be treated in the hour immediately preceding time t. The amount of stored water to be treated at time t is positively correlated with the water level in the receiving well 31 at time t, and this correlation is understood by the manager of the sewage treatment plant 1. Therefore, the term (V(t-1) - V(t)) is determined based on the actual water level measured by the water level gauge 71.

[0050] As described above, the right side of equation (1) is determined based on the actual value of the pumping amount of the water pumping pump 6 and the actual value of the water level measured by the water level meter 71. Therefore, based on these actual values, the actual value Q of the inflow of the water to be treated in (t) can be identified.

[0051] Note that the pair of actual precipitation values ​​and actual values ​​of the inflow rate of treated water does not require simultaneity. This is because the inflow rate of treated water increases or decreases slightly after changes in rainfall. In other words, the actual value of the inflow rate of treated water at a sewage treatment plant when an actual value of precipitation is recorded refers to the actual value of the inflow rate of treated water at a moment or time period that can be affected by the actual precipitation at a certain moment or time period.

[0052] The algorithm used to generate a trained model from training data is not particularly limited. Examples include, but are not limited to, support vector machines (regression, classification), decision trees, random forests, gradient boosting, light GBM, logistic regression, neural networks (simple perceptron, multilayer perceptron), Gaussian process regression, Bayesian networks, k-nearest neighbors, lasso regression, multiple regression analysis, ridge regression, elastic net, and partial least squares regression. Note that the validity of the generated trained model may be verified by using part of the training data and the rest as test data.

[0053] When a forecasted precipitation value is input as an explanatory variable to the trained model, a predicted value of the inflow of water to be treated is output as a target variable. The forecasted precipitation value is the forecasted precipitation value for the area where the sewage pipe 2 is laid (an example of a predetermined area), and is, for example, precipitation forecast data provided by the Japan Meteorological Agency. Here, the area in which the forecasted precipitation value is considered is preferably the same as the area considered when constructing the trained model. Therefore, the area in which the forecasted precipitation value is considered may be the entire treatment area, but preferably, an area within the treatment area where rainwater is likely to infiltrate the sewage pipe 2 is selected. As an example, the actual precipitation amount at a measurement point whose distance from the sewage treatment plant 1 is equal to or less than a predetermined threshold is considered.

[0054] Precipitation forecast values ​​can also be numerical values ​​for any time interval. For example, the Japan Meteorological Agency provides several types of forecast values ​​with different spatial and temporal accuracy, including short-term precipitation forecast values ​​(1 km mesh, every 10 minutes, up to 6 hours later), 15-hour precipitation forecast values ​​(5 km mesh, every 60 minutes, up to 7 to 15 hours later), and mesoscale model prediction values ​​(0.05 degrees latitude and 0.0625 degrees longitude increments, every 12 hours, up to 79 hours later). Using the short-term precipitation forecast values ​​and 15-hour precipitation forecast values ​​as explanatory variables makes it easier to predict the inflow of treated water over a relatively long time span while improving the accuracy of the prediction, especially for the near future.

[0055] Furthermore, in addition to the forecasted precipitation amount, the actual precipitation amount may also be taken into consideration. For example, the inflow amount of water to be treated several hours after the time point at which the prediction is made (hereinafter referred to as the "reference time") depends on the amount of precipitation after the reference time as well as the amount of precipitation up to the reference time. Therefore, a combination of the actual and forecasted precipitation amounts for the time period surrounding the reference time may be used as an explanatory variable to predict the inflow amount of water to be treated.

[0056] The above method allows the inflow of treated water to be predicted for as far into the future as precipitation forecast values ​​are available. Therefore, it is possible to obtain a predicted inflow value for each hour of the future period for which precipitation forecast values ​​are available, and to predict the time-dependent change in inflow during that period. Note that the accuracy of the inflow prediction depends on the accuracy of the precipitation forecast value, with higher accuracy for the near future than for the distant future. In this embodiment, as an example, the inflow is predicted for up to 72 hours in advance.

[0057] (2) Determining the recommended pumping volume Second, a recommended value for the amount of water to be pumped is determined for the sewage treatment plant 1. The recommended value for the amount of water to be pumped is determined based on a predicted value for the inflow amount of water to be treated.

[0058] The basic concept of the sewage treatment plant 1 is to standardize the operating conditions of the water treatment facility 4 as much as possible. To achieve this, it is desirable to determine the operating conditions so as to minimize fluctuations in the amount of water pumped by the lift pump 6. However, it is also necessary to consider the balance of the water to be treated in the storage tank 3, i.e., the difference between the inflow amount and the amount of water pumped.

[0059] When the inflow of water to be treated is less than the pumping rate, the amount of stored water decreases as the operation of the sewage treatment plant 1 continues, and it may become necessary to reduce the pumping rate. In this case, the pumping rate may suddenly decrease, which may significantly change the operating conditions of the water treatment facility 4. Therefore, it is desirable to operate the plant so that the pumping rate gradually decreases as the stored water volume decreases.

[0060] When the inflow of water to be treated is greater than the pumping rate, the amount of water stored increases as the operation of the sewage treatment plant 1 continues, and it may become necessary to increase the pumping rate to prevent the storage tank 3 from overflowing. In this case, the pumping rate may suddenly increase, which may significantly change the operating conditions of the water treatment facility 4. Therefore, it is desirable to operate the plant so that the pumping rate gradually increases as the amount of stored water increases.

[0061] That is, the recommended pumping rate can be determined by adjusting the pumping rate according to the amount of stored water to be treated, based on the basic idea that the pumping rate should be balanced with the inflow rate of water to be treated. Figure 3 is a flowchart showing a method for determining the recommended pumping rate based on the above idea. The following describes a method for determining the recommended pumping rate Qr(t) at time t.

[0062] In the first step #10, the recommended pumping rate Qr(t) is determined according to the following equation (2).

number

[0063] Ave.Q in the first term on the right side of equation (2) in (t) is the average value of the inflow of water to be treated for each hour for the 12 hours before and after time t. The value of this term is determined using the trained model as described above.

[0064] Ave. V(t) in the second term on the right side of equation (2) is the average value of the storage volume of the water to be treated in the 24 hours immediately prior to time t. As mentioned above, the storage volume of the water to be treated has a positive correlation with the water level in the storage tank 3 and is determined based on the water level measured by the water level gauge 7. Vs in the second term on the right side of equation (2) is the reference value for the storage volume of the water to be treated. Therefore, the term (Ave. V(t) - Vs) represents the difference between the average storage volume of the water to be treated and the reference value. A positive value for this term indicates that the storage volume tends to be greater than the reference value, and a negative value for this term indicates that the storage volume tends to be less than the reference value. Note that A1 in the second term on the right side of equation (2) is a constant multiplied by the difference between the average storage volume of the water to be treated and the reference value.

[0065] Ri in the third term on the right-hand side of equation (2) is an index value that indicates the tendency of fluctuations in the storage volume of untreated water, and is hereinafter referred to as the "increase index." The increase index is determined as the sum of the maximum hourly increase in the inflow volume of untreated water in the 12 hours before and after time t and the maximum hourly increase in the storage volume of untreated water in the 24 hours immediately before time t, multiplied by 24. In other words, the increase index assumes that the behavior of the inflow of untreated water and the behavior of increases and decreases in the storage volume, which are taken into account when calculating the recommended pumping rate Qr(t), work to the maximum extent possible in the direction of increasing the storage volume.

[0066] B in the third term on the right side of equation (2) is a coefficient determined by case classification based on the storage volume V(t-1) one hour before time t, and is selected from three candidates (B1, B2, and B3 in order of decreasing size) using the following procedure.

[0067] In the first branch #11, the storage volume V(t-1) one hour before time t is compared with the first threshold T1. The first threshold T1 indicates the upper limit of the storage volume that is permissible under normal circumstances (when there are no exceptional factors that would increase the storage volume, such as heavy rain). If the storage volume V(t-1) one hour before time t is smaller than the first threshold T1, a low level coefficient B1 is selected as the coefficient B (#13). If the storage volume V(t-1) one hour before time t is equal to or greater than the first threshold T1, the process proceeds to the second branch #12.

[0068] In the second branch #12, the storage volume V(t-1) one hour before time t is compared with a second threshold T2. The second threshold T2 indicates an upper limit that is not allowed to be exceeded even when exceptional factors such as heavy rain are taken into consideration. If the storage volume V(t-1) one hour before time t is smaller than the second threshold T2, a medium-level coefficient B2 is selected as coefficient B (#14). If the storage volume V(t-1) one hour before time t is equal to or greater than the second threshold T2, a high-level coefficient B3 is selected as coefficient B (#15).

[0069] That is, the larger the storage volume V(t-1) one hour before time t, the larger the coefficient B, and the larger the recommended pumping volume Qr(t) determined according to equation (2) becomes. This is intended to reduce the storage volume of water to be treated by increasing the recommended pumping volume Qr(t) when the storage volume of water to be treated exceeds either threshold T1 or T2.

[0070] The coefficient B determined in the above procedure is applied to determine the recommended pumping rate Qr(t) according to equation (2) (#16).

[0071] In the second step #20, the recommended pumping rate Qr(t) is corrected so that it does not exceed the physical tolerance. In the first step #10, a recommended value may be determined that exceeds the physical constraints caused by the capacity of the sewage pipeline 2 and the storage tank 3, so this is corrected.

[0072] In the third branch #21, the estimated storage volume Ve(t) at time t is compared with a third threshold T3. The estimated storage volume Ve(t) is the sum of the storage volume V(t-1) one hour before time t and the predicted value of the inflow volume of water to be treated in the hour immediately before time t, minus the pumping volume in the hour immediately before time t. The third threshold T3 indicates the lower limit of the allowable storage volume of water to be treated. If the estimated storage volume Ve(t) at time t is smaller than the third threshold T3, the process proceeds to the fourth branch #22, and if the estimated storage volume Ve(t) at time t is equal to or greater than the third threshold T3, the process proceeds to the fifth branch #23.

[0073] In the fourth branch #22, the estimated storage volume Ve(t) at time t is compared with 0. If the estimated storage volume Ve(t) at time t is less than 0, the recommended pumping volume Qr(t) is replaced with a value determined according to the following equation (3) (#24). If the estimated storage volume Ve(t) at time t is 0 or greater, the recommended pumping volume Qr(t) is replaced with a value determined according to the following equation (4) (#25). In either case, the process then proceeds to step #27.

number

[0074] When the estimated storage volume Ve(t) is below 0, the storage volume of the water to be treated is 0 or very close to 0, so the storage volume one hour before time t plus the inflow volume is the maximum amount of water to be treated present on the primary side of the pumping pump 6. Therefore, it is not possible to achieve a pumping volume that exceeds this amount. Equation (3) reflects this situation, and the recommended pumping volume Qr(t) is calculated by multiplying the storage volume V(t-1) one hour before time t by the predicted inflow volume Q for the hour immediately preceding time t. in In this situation, it is not possible to increase the amount of water stored to be treated, so the highest priority is given to leveling the operating conditions of the water treatment facility 4.

[0075] If the estimated storage volume Ve(t) is equal to or greater than 0, the recommended pumping volume Qr(t) is determined with the aim of increasing the storage volume of the water to be treated up to the third threshold T3. That is, the recommended pumping volume Qr(t) is set to the value obtained by subtracting the third threshold T3 from the maximum pumping volume considered in equation (3) (equation (4)), thereby suppressing the pumping volume and increasing the storage volume.

[0076] In the fifth branch #23, the estimated storage volume Ve(t) at time t is compared with a fourth threshold T4. The fourth threshold T4 indicates the upper limit of the storage volume of the water to be treated. If the estimated storage volume Ve(t) at time t is greater than the fourth threshold T4, the recommended pumping volume Qr(t) is replaced with a value determined according to the following equation (5) (#26), and then processing proceeds to #27. If the estimated storage volume Ve(t) at time t is equal to or less than the fourth threshold T4, the recommended pumping volume Qr(t) is not changed and processing proceeds to #27.

number

[0077] If the estimated storage volume Ve(t) is greater than the fourth threshold T4, the recommended pumping volume Qr(t) is determined with the aim of reducing the storage volume of untreated water to the fourth threshold T4 in mind. In other words, the entire difference between the estimated storage volume Ve(t) and the fourth threshold T4 is used for pumping, and any untreated water that has flowed in is also used for pumping, thereby reducing the storage volume.

[0078] In process #27, a correction is made taking into account the capacity of the pumping pump 6. The recommended pumping rate Qr(t) determined in the steps up to this point is compared with the maximum pumping rate Qmax and minimum pumping rate Qmin based on the capacity of the pumping pump 6. The maximum pumping rate Qmax is, for example, the pumping rate when all five pumping pumps 6 are operating at maximum output. The minimum pumping rate Qmin is, for example, the pumping rate when only one pumping pump 6 is operating at minimum output. If the recommended pumping rate Qr(t) is greater than or equal to the minimum Qmin and less than or equal to the maximum Qmax, the recommended pumping rate Qr(t) is not changed. If the recommended pumping rate Qr(t) exceeds the maximum Qmax, the recommended pumping rate Qr(t) is replaced with the maximum Qmax. If the recommended pumping rate Qr(t) is less than the minimum Qmin, the recommended pumping rate Qr(t) is replaced with the minimum Qmin.

[0079] The recommended pumping rate Qr(t) determined by the above procedure is presented to the manager of the sewage treatment plant 1, for example, via an output device of the control device 8. The manager, who receives the recommended pumping rate Qr(t), sets the pumping rate by, for example, changing the setting value of the pumping pump 6. Note that the control device 8 may itself have a function to control the pumping pump 6 according to the recommended pumping rate Qr(t) without manual setting operations by the manager, for example. Furthermore, the manager of the sewage treatment plant 1 may be presented with recommended examples of combinations of pumping pumps 6 to be started in conjunction with the recommended pumping rate Qr(t).

[0080] (3) Determining the recommended gate opening value Third, a recommended value for the opening degree of the gate device 5 is determined. The recommended value for the opening degree of the gate device 5 is determined based on the recommended value for the pumping rate and the water level of the storage tank 3 (measured value of the water level gauge 7). As described above, this embodiment takes the approach of minimizing fluctuations in the amount of water pumped by the pumping pump 6. Since the relationship between the output of the pumping pump 6 and the amount of water pumped is influenced by the water level (water pressure) on the primary side of the pumping pump 6, suppressing fluctuations in the water level of the storage tank 3 is advantageous in suppressing fluctuations in the amount of water pumped. Therefore, the opening degree of the gate device 5 is determined so as to suppress fluctuations in the water level of the storage tank 3.

[0081] In this embodiment, a trained model is used to determine a recommended opening value for the gate device 5, with the inflow rate and pumping rate of the water to be treated as explanatory variables and the opening value of the gate device 5 that can keep fluctuations in the water level of the pump well 33 (measurement value of the second water level gauge 72) within a predetermined range based on the inflow rate and pumping rate as the objective variable. A predicted value for the inflow rate of the water to be treated and a recommended value for the pumping rate are input into the trained model, and a recommended opening value for the gate device 5 is obtained.

[0082] The trained model used here is constructed using as training data a set of the actual values ​​of the inflow rate of the water to be treated, the actual values ​​of the pumping rate, the actual values ​​of the measurements of the second water level gauge 72, and the actual value of the opening degree of the gate device 5. The trained model may be constructed by the management system 10 according to this embodiment, or by another computer, etc. Each actual value constituting the training data can be identified based on the operating history of the sewage treatment plant 1.

[0083] The algorithm used to generate a trained model from training data is not particularly limited. Examples include, but are not limited to, support vector machines (regression, classification), decision trees, random forests, gradient boosting, light GBM, logistic regression, neural networks (simple perceptron, multilayer perceptron), Gaussian process regression, Bayesian networks, k-nearest neighbors, lasso regression, multiple regression analysis, ridge regression, elastic net, and partial least squares regression. Note that the validity of the generated trained model may be verified by using part of the training data and the rest as test data.

[0084] The recommended value of the opening degree of the gate device 5 determined using the trained model is presented to the manager or the like of the sewage treatment plant 1 via an output device or the like of the control device 8. The manager or the like who receives the recommended value of the opening degree of the gate device 5 sets the opening degree of the gate device 5. Note that the control device 8 may also have a function to control the gate device 5 itself according to the recommended value without manual setting operations by the manager or the like.

[0085] Other Embodiments Finally, other embodiments of the management method, management program, and management system according to the present invention will be described. Note that the configurations disclosed in the following embodiments can be applied in combination with the configurations disclosed in other embodiments, as long as no contradiction occurs.

[0086] In the above embodiment, an example of a management method including a process for determining a recommended value for the pumping rate and a recommended value for the opening degree of the gate device 5 has been described. However, in the present invention, it is optional whether or not to determine these recommended values. Also, only the recommended value for the pumping rate may be determined. Furthermore, the method for determining each recommended value in the above embodiment is merely an example.

[0087] Regarding other configurations, it should be understood that the embodiments disclosed in this specification are illustrative in all respects and that the scope of the present invention is not limited thereby. Those skilled in the art will easily understand that appropriate modifications are possible without departing from the spirit of the present invention. Therefore, other embodiments modified without departing from the spirit of the present invention are naturally included in the scope of the present invention. [Example]

[0088] The present invention will be further described below with reference to examples, but the present invention is not limited to these examples.

[0089] (1) Trained model for predicting the inflow of treated water Example 1 For sewage treatment plant 1 of a certain municipality in Japan, the actual (hourly) inflow volume of treated water was identified based on the nine-year operation history from 2012 to 2021. In addition, precipitation data (hourly) showing the actual precipitation volume around sewage treatment plant 1 was obtained from the Japan Meteorological Agency.

[0090] FIG. 2 is a map of the area around the sewage treatment plant 1, and precipitation data is provided by the Japan Meteorological Agency for the intersections of the grids overlaid on the map. The grid spacing is 1 km. In Example 1, the target locations for acquiring precipitation data were 36 locations (P1 and P2 in FIG. 2) within the area where the sewage pipes 2 of the sewage treatment plant 1 are laid, for which precipitation data is provided by the Japan Meteorological Agency. P1 is the 9th location out of the 36 locations closest to the sewage treatment plant 1, and is indicated by a square mark. P2 is the remaining 27 locations, and is indicated by a circle. The central city of the municipality where the sewage treatment plant 1 is located is located within the distribution range of the 36 locations selected here.

[0091] The rainfall data was used after extracting features. The features used were (a) to (d) below. (a) Feature representing time (time elapsed from a reference point) (b) Average hourly precipitation at P1 and P2 (36 locations in total) (multiple locations per location) (c) Average 6-hour precipitation at P1 and P2 (36 locations in total) (multiple points for each location) (d) Average 6-hour precipitation at P1 (9 locations) (multiple points for each location)

[0092] As a result, training data consisting of approximately 2.8 million pairs of actual values ​​of precipitation and actual values ​​of inflow of water to be treated was obtained.

[0093] Using Light GBM in a Python environment, we generated and verified a trained model using the above training data. Light GBM was used as the algorithm to generate the trained model. The training data was divided into nine parts by year and cross-validation was performed. The coefficient of determination R of the obtained trained model was 2 was 0.659, and a trained model with sufficient accuracy for practical use was obtained.

[0094] Example 2 A trained model was generated and verified using training data in which precipitation data was acquired only from P1 in Figure 2. A trained model was generated using the same procedure as above, except that the rainfall-related features of the training data were changed to (a), (b'), and (d) below. (a) Feature representing time (time elapsed from a reference point) (b') Average hourly precipitation at P1 (9 locations) (multiple points for each location) (d) Average 6-hour precipitation at P1 (9 locations) (multiple points for each location)

[0095] The coefficient of determination R of the trained model obtained 2 The coefficient of determination (R) was 0.663, which means that a trained model with sufficient accuracy for practical use was obtained. 2 were equivalent.

[0096] (2) A trained model for determining the recommended gate opening value Example 3 For a sewage treatment plant of a certain municipality in Japan, we identified the actual values ​​of the inflow volume of treated water, the actual value of the pumped volume, the actual value of the water level in the pump well, and the actual value of the gate device opening, based on the operating history for one year up to 2023. This resulted in training data consisting of approximately 80,000 sets of the above actual values.

[0097] Using Light GBM in a Python environment, we generated and verified a trained model using the above training data. Light GBM was used as the algorithm used to generate the trained model. The obtained trained model was verified by cross-validation. The coefficient of determination R of the obtained trained model 2 was 0.669, and a trained model with sufficient accuracy for practical use was obtained. [Industrial Applicability]

[0098] INDUSTRIAL APPLICABILITY The present invention can be used to manage a sewage treatment plant into which water to be treated flows in from a sewage pipe of a separate sewer system. [Explanation of symbols]

[0099] 1: Sewage treatment plant 2: Sewage pipe 3: Reservoir 31: Landing well 32: Settling pond 33: Pump well 4: Water treatment facilities 5: Gate device 6: Water pump 7: Water level gauge 71:First water level gauge 72:Second water level gauge 8: Control device 10: Management System

Claims

1. A management method for a sewage treatment plant into which treated water flows in from a sewage pipe of a separate sewer system, comprising: A management method including a step of inputting at least the forecast value of precipitation in a specified area into a trained model constructed using as training data a pair of actual precipitation values ​​in the specified area where the sewage pipeline is laid and actual values ​​of the inflow of treated water at the sewage treatment plant at the time the actual precipitation values ​​were recorded, thereby obtaining a predicted value of the inflow of treated water.

2. The management method described in claim 1, wherein the specified area is a portion of the entire area in which the sewage pipeline is laid, selected based on the likelihood of rainwater infiltration into the sewage pipeline.

3. 2. The management method according to claim 1, wherein the predetermined area is a portion of the entire area in which the sewage pipeline is laid, selected based on the distance from the sewage treatment plant.

4. The management method described in claim 1, wherein the teaching data includes, as actual values ​​of precipitation in the specified area, a first accumulated value which is the accumulated value of precipitation over a specified first period, and a second accumulated value which is the accumulated value of precipitation over a specified second period which is longer than the first period.

5. The management method according to any one of claims 1 to 4, further comprising the step of determining a pumping rate at the sewage treatment plant based on the predicted value of the inflow rate.

6. A management program for managing a sewage treatment plant into which treated water flows in from a sewage pipe of a separate sewer system, When executed by a computer, A management program that can realize the function of inputting at least the forecast value of precipitation in a specified area into a trained model constructed using as training data a pair of actual precipitation values ​​in the specified area where the sewage pipeline is laid and actual values ​​of the inflow volume of treated water at the sewage treatment plant at the time the actual precipitation values ​​were recorded, and obtaining a predicted value of the inflow volume.

7. A management system for managing a sewage treatment plant into which water to be treated flows from a sewage pipe of a separate sewer system, a pumping device that sends the water to be treated to a water treatment facility; a control device for controlling the water pumping device; The control device a function of inputting a forecast value of at least the precipitation amount in the predetermined area into a trained model constructed using a set of an actual value of the precipitation amount in the predetermined area where the sewage pipeline is laid and an actual value of the inflow amount of water to be treated at the sewage treatment plant when the actual value of the precipitation amount was recorded as training data, and obtaining a predicted value of the inflow amount; A management system capable of realizing a function of controlling the pumping device based on the predicted value of the inflow amount.

Images