Inflow volume prediction system

Abstract

We provide an inflow volume prediction system that can accurately predict sewage inflow volume even when information on external inflow pipes is unknown. [Solution] The system includes a facility data collection unit 113, a weather data collection unit 115, a data storage unit 110, a learning processing unit 150 that learns a model to predict the amount of sewage flowing into the pumping station 10 based on facility data and weather data and creates a trained model, a trained model storage unit 123, a prediction unit 160 that predicts the amount of sewage flowing based on the trained model, facility data and weather data, and a control unit 170. The amount of sewage flowing is calculated by summing the fluctuation amount of water held in the inflow channel 13 and the amount of water pumped by the pump. The fluctuation amount of sewage stored in the external inflow pipe 11 is estimated and used as one of the input values ​​for predicting the amount of sewage flowing.

Description

【Technical Field】 【0001】 The present invention relates to an inflow prediction system for predicting the inflow of sewage flowing into a pump station, a pump building in a sewage treatment plant, a pipeline facility having inflow measurement means, or a sewage treatment plant without a pump building, etc. 【Background Art】 【0002】 Conventionally, predicting the inflow of sewage flowing into facilities such as a pump station, a pump building in a sewage treatment plant, a pipeline facility having inflow measurement means, or a sewage treatment plant without a pump building (hereinafter also referred to as "pump station, etc.") is important for stabilizing the treated water quality by an appropriate operation method of the facility, appropriately arranging personnel for maintenance management during rainy days, preventing overflows inside and outside the facility, etc. For this reason, conventionally, a skilled technician in charge of facility management predicts fluctuations in the inflow of sewage based on the characteristics of the facility, meteorological conditions, etc., and temporarily increases the flow rate or treatment volume before the occurrence of a rainfall event to preliminarily reduce the sewage retention volume in the sewer pipeline, pump station, or sewage treatment plant, or stores a part of the inflowing water increased during rainy days in the pipeline and performs flow-down and treatment after the end of the rainfall event. 【0003】 On the other hand, in recent years, as a method for appropriately predicting the inflow of sewage without relying on a skilled technician, a method for predicting the inflow of sewage using a machine learning technique has been studied. For example, Patent Document 1 discloses an inflow prediction system having a facility data collection unit that collects facility data of a pump station, etc., a meteorological data collection unit that collects meteorological data, a data storage unit that stores facility data and meteorological data, a learning processing unit that creates a learned model for predicting the inflow of sewage based on the facility data and meteorological data, a prediction unit that predicts the inflow of sewage based on the learned model, facility data, and meteorological data, and a control unit that receives the predicted value of the inflow of sewage and controls the operating state of a pump station, etc. 【Prior Art Documents】 【Patent Documents】 【0004】 [Patent Document 1] Japanese Patent Publication No. 2023-169099 [Overview of the Initiative] [Problems that the invention aims to solve] 【0005】 In Patent Document 1, the current sewage inflow to a pumping station is used as one of the explanatory variables to predict the amount of sewage that will flow into the pumping station at a future point in time. This sewage inflow is calculated by summing the fluctuations per unit time of the amount of water held in the tanks that make up the pumping station, the pumping rate, and the fluctuations per unit time of the amount of sewage stored in the off-site inflow pipe connected to the upstream side of the pumping station. The fluctuations per unit time of the amount of sewage stored in the off-site inflow pipe are calculated by referring to the water level in the tank to which the off-site inflow pipe is connected, as well as the gradient, pipe diameter, and structure of the off-site inflow pipe. 【0006】 However, in many existing sewage treatment plants, the gradient, diameter, and structure of the inflow pipes outside the facility are unknown. In such cases, it is impossible to calculate the amount of sewage stored in the inflow pipes outside the facility, so it was necessary to provide an alternative calculation method. 【0007】 Furthermore, while Reference 1 assumes that the operation of pumps and gates is controlled based on predicted sewage inflow rates to manage the operation of a sewage treatment plant, operators of sewage treatment plants who had not previously recorded sewage inflow rates found it difficult to understand the meaning of predicted sewage inflow rates (the value and method of using these rates), making it difficult to make appropriate operational management decisions based on these rates. Therefore, it was necessary to provide data that is easier to use for operational management, such as predicted values ​​of the water level in the tanks that fluctuate in response to changes in sewage inflow rates, and information on the degree of risk related to predicted sewage inflow rates. 【0008】 The present invention has been made in view of the above-mentioned points, and its purpose is to provide an inflow rate prediction system that can accurately predict the amount of sewage inflow, not only when the information of the external inflow pipe connected to the upstream side of the sewage treatment plant is known, but also when it is unknown or uncertain. Another objective of the present invention is to provide an inflow prediction system that can obtain data from the predicted sewage inflow that is easier to use for operational management. [Means for solving the problem] 【0009】 The present invention relates to an inflow rate prediction system for predicting the amount of sewage inflow at a pumping station or a sewage treatment plant, such as a pump building or pipeline facility having an inflow rate measuring means, or a sewage treatment plant, comprising: a facility data collection unit for collecting facility data of the pumping station or sewage treatment plant, such as a pump building or pipeline facility having an inflow rate measuring means, or a sewage treatment plant; a weather data collection unit for collecting weather data; a data storage unit for storing the facility data and weather data; a learning processing unit for learning a model to predict the amount of sewage inflow to a pumping station or a sewage treatment plant, such as a pump building or pipeline facility having an inflow rate measuring means, or a sewage treatment plant, based on the facility data and weather data received from the data storage unit, and creating a trained model; a trained model storage unit for storing the trained model created by the learning processing unit; a prediction unit for predicting the amount of sewage inflow at a future location based on the trained model, the facility data, and the weather data; and a unit that receives a predicted value of the amount of sewage inflow from the prediction unit and predicts the amount of sewage inflow at a pumping station or a sewage treatment plant. The inflow rate prediction system comprises a control unit that controls the operating state of a pump building within the facility or a pipeline facility or sewage treatment plant equipped with an inflow rate measuring means, and is characterized in that, when the target for predicting the sewage inflow rate is a pump building within the pumping station or sewage treatment plant, the sewage inflow rate is calculated by summing the amount of fluctuation per unit time of the amount of water held in one or more tanks of the pump building within the pumping station or sewage treatment plant and the amount of water pumped per unit time; on the other hand, when the target for predicting the sewage inflow rate is a pipeline facility or sewage treatment plant equipped with an inflow rate measuring means, the sewage inflow rate is calculated as the amount of fluctuation per unit time of the amount of water held in one or more tanks of the pipeline facility or sewage treatment plant; and further estimates the amount of fluctuation per unit time of the amount of sewage stored in an off-site inflow pipe connected to the upstream side of the pump building within the pumping station or sewage treatment plant or a pipeline facility or sewage treatment plant equipped with an inflow rate measuring means, and uses the estimated amount of fluctuation as one of the input values ​​used to predict the sewage inflow rate. In addition to sewage treatment plants that have pump buildings, there are also sewage treatment plants that do not have pump buildings, and this invention also includes such sewage treatment plants. Specific control actions performed by the control unit to determine the operating status of a pumping station include, for example, adjusting the gate opening, the number of pumps in operation, and the rotational speed of the pumps. Examples of pipeline facilities equipped with means for measuring flow rate include manholes that serve as water tanks and are fitted with water level gauges and flow meters. In this context, tanks whose water levels fluctuate in response to changes in sewage inflow include, in the case of pumping stations or pump buildings within sewage treatment plants, tanks that make up inflow channels, sedimentation basins, and pump wells, and in the case of sewage treatment plants without pump buildings, tanks that make up primary sedimentation tanks, reaction tanks, and final sedimentation tanks. According to the present invention, the amount of fluctuation per unit time in the amount of sewage stored in the inflow pipe outside the facility is estimated separately from the amount of sewage flowing into the pumping station, etc., without being added to the amount of sewage flowing into the pumping station, etc., and this fluctuation amount is used as one of the input values ​​for predicting the amount of sewage flow. Therefore, even if the information of the inflow pipe outside the facility is unknown or uncertain, the amount of sewage flow can be predicted with high accuracy by using this estimated value. Even when information about the external inflow pipes is available, estimating the rate of change in the amount of sewage stored in the external inflow pipes per unit time, separately from the amount of sewage flowing into pumping stations, etc., and treating this change as a separate input value, makes it easier to process within the system compared to treating the combined amount of sewage flowing in as a single input value. Furthermore, according to the present invention, even if a flow meter is not installed to measure the amount of sewage flowing into a pumping station or the like, the amount of sewage flowing in can be predicted with high accuracy. Furthermore, the meteorological data used in the prediction unit for predicting sewage inflow is preferably mesh-unit meteorological data containing weather values, longitude / latitude, and time information, and is limited to meteorological data of mesh portions that overlap with the pump building or pipeline facility with inflow measurement means or the treatment area of ​​the sewage treatment plant within the pumping station or sewage treatment plant. With this configuration, only the meteorological data of mesh portions that overlap with the treatment area of ​​the pumping station, etc., i.e., meteorological data that has a high influence on the sewage inflow into the pumping station, etc., is used as input values ​​for predicting sewage inflow, thereby improving the accuracy of sewage inflow prediction. 【0010】 In addition to the above features, the present invention is characterized in that, if the gradient, diameter, and structure of the external inlet pipe are known, the amount of fluctuation per unit time of the amount of sewage stored in the external inlet pipe is estimated from the fluctuation of the water level in the tank to which the external inlet pipe is connected, as well as from the gradient, diameter, and structure of the external inlet pipe. According to the present invention, if the gradient, diameter, and structure of the external inlet pipe are known, the amount of fluctuation per unit time of the amount of sewage stored in the external inlet pipe is estimated by referring to these factors and the fluctuation of the water level in the tank to which the external inlet pipe is connected. Instead of adding this fluctuation amount to the amount of sewage flowing into the pumping station, the accuracy of the sewage flow rate prediction is improved by using it separately as one of the explanatory variables used to predict the amount of sewage flowing into the pumping station, etc. 【0011】 Furthermore, in addition to the above features, the present invention is characterized in that, if the gradient, diameter, and structure of the external inlet pipe are not known, the amount of fluctuation per unit time of the amount of sewage stored in the external inlet pipe is estimated and calculated by estimating the correspondence between the water level in the water tank to which the external inlet pipe is connected and the amount of sewage stored in the external inlet pipe, based on the fluctuation of the water level in the water tank to which the external inlet pipe is connected when the gate installed at the pumping station, etc. is closed, and the amount of sewage flowing into the pumping station, etc. According to the present invention, even when the gradient, diameter, and structure of the external inlet pipe are not known, the relationship between the water level in the tank to which the external inlet pipe is connected and the amount of sewage stored in the external inlet pipe is estimated. Instead of adding the estimated fluctuation amount of sewage stored in the external inlet pipe to the amount of sewage flowing into the pumping station, the present invention improves the accuracy of sewage flow rate prediction by using it separately as one of the explanatory variables used to predict the amount of sewage flowing into the pumping station, etc. 【0012】 In addition to the above features, the present invention is characterized in that, if the gradient, diameter, and structure of the external inflow pipe are not known, multiple candidate values ​​for the amount of sewage stored in the external inflow pipe are calculated using multiple candidate values ​​for the gradient, diameter, and structure of the external inflow pipe, and the candidate value that provides the best prediction accuracy when each of these candidate values ​​is used as one of the input values ​​for predicting the amount of sewage inflow is selected to estimate and calculate the amount of sewage stored in the external inflow pipe. According to the present invention, even when the gradient, diameter, and structure of the inflow pipe outside the facility are not known, multiple candidate values ​​for the amount of sewage stored in the inflow pipe outside the facility are calculated using multiple candidate values ​​for the gradient, diameter, and structure of the inflow pipe outside the facility. An estimated value of the amount of sewage stored in the inflow pipe outside the facility is calculated by adopting the candidate value that yields the best prediction accuracy when these are used as explanatory variables for prediction. Instead of adding the estimated fluctuation amount of sewage stored in the inflow pipe outside the facility to the amount of sewage flowing into the pumping station, etc., it is used separately as one of the explanatory variables used to predict the amount of sewage flowing into the pumping station, etc., thereby improving the accuracy of sewage flow prediction. 【0013】 Furthermore, in addition to the above features, the present invention is characterized in that at least one of the following is installed in the treatment area such as the pumping station: an upstream pumping station that collects and flows in sewage from an upstream treatment area located upstream of the treatment area, or a pipeline facility located upstream of the treatment area that has means for measuring the inflow rate of the upstream treatment area. The prediction value of the pumping volume of the upstream pumping station, or the prediction value of the sewage inflow at the pipeline facility having means for measuring the inflow rate of the upstream treatment area, is set to an arbitrary value, and the prediction of the sewage inflow by the prediction unit at the pumping station or the downstream station is set to the prediction value of the pumping volume of the upstream pumping station, or the prediction value of the sewage inflow at the pipeline facility having means for measuring the inflow rate of the upstream treatment area. According to the present invention, for example, it becomes possible to simulate the increase or decrease in sewage inflow at a downstream pumping station, etc., assuming that rain were to fall in the treatment area of ​​an upstream pumping station, even though it is not currently raining there. 【0014】 Furthermore, in addition to the above features, the present invention is characterized in that, when the target for predicting the amount of sewage inflow is a pumping station or a pump building within a sewage treatment plant, the amount of pump water pumped per unit time at a future location is specified to an arbitrary value or to the current value, and the water level of one or more tanks in the pumping station or pump building, the amount of pump water pumped per unit time, and the predicted value of the amount of sewage inflow are used to predict the water level of one or more tanks in the pumping station or pump building at the future location. According to the present invention, it is possible to perform simulations, for example, to predict how the water level in a tank will fluctuate at a future location, based on predicted values ​​of sewage inflow, when the pump pumping volume at a future location is set to a predetermined constant value. In other words, since the water level in the tank can be predicted simultaneously with the prediction of sewage inflow, it becomes possible to plan and manage the pump pumping volume by referring to this. 【0015】 Furthermore, in addition to the above features, the present invention is characterized in that, when the target for predicting the amount of sewage inflow is a pump station or a pump building within a sewage treatment plant, the water level of one or more tanks in the pump station or pump building at a future location is specified to an arbitrary value or to the current value, and the water level and structure of one or more tanks in the pump station or pump building, along with the predicted amount of sewage inflow, is used to predict the amount of water pumped per unit time at the future location such that the water level of one or more tanks in the pump station or pump building is maintained at the specified arbitrary value or the specified current value. According to the present invention, it is possible to perform simulations such as predicting how the amount of water pumped from a tank at a future location will fluctuate, based on predicted values ​​of sewage inflow, for example, when the water level in a tank at a future location such as a pumping station is set to a predetermined constant value. 【0016】 In addition to the above features, the present invention is also characterized by predicting the amount of sewage inflow for each unit time within a specified period based on facility data and weather data over an arbitrary long period, using the facility data and weather data prior to each unit time, and outputting predicted values ​​equivalent to those obtained when predicting the amount of sewage inflow for each unit time within that period. According to the present invention, it is possible not only to predict the current and future sewage inflow from facility data and meteorological data for the minimum period necessary for predicting sewage inflow, but also to output past predicted values ​​of sewage inflow for each unit time within that period as a single database by using longer-term facility data and meteorological data as input data. This makes it possible to monitor the model performance over a long period and analyze trends in predicted values. In other words, instead of real-time data, we can obtain "predicted values ​​that would have been obtained if a prediction had been made at a certain point in the past." By varying this "certain point in time" every unit of time, we can output predicted values ​​using data from 0:00 to 10:00, predicted values ​​using data from 1:00 to 11:00, predicted values ​​using data from 2:00 to 12:00, and so on, for every unit of time within the period, and determine the accuracy of the model. 【0017】 In addition to the above features, the present invention designates any type of variable among the facility data and weather data used for predicting the sewage inflow volume, various calculated values for prediction calculated using the facility data and weather data, and the predicted sewage inflow volume, and designates any reference value and logical operator, thereby setting any type of alarm output condition and outputting an alarm for the prediction result. According to the present invention, when a plurality of reference values and logical operators are designated for any type of variable, for example, when various data, various calculated values, or predicted values of sewage inflow volume exceed the reference value, a plurality of types of alarms are output. Therefore, not only the predicted value of the sewage inflow volume, but also the predicted value of the pumping volume of the upstream pump station and a plurality of reference values such as weather data can be combined to output a plurality of alarms having probabilistically different characteristics. As a result, a dangerous state or the like can be visualized in an easily understandable form and used for the operation determination of the sewage treatment plant. 【Advantages of the Invention】 【0018】 According to the present invention, the sewage inflow volume can be accurately predicted not only when the information of the external inflow pipe of the facility connected to the upstream side of the sewage treatment plant is clear, but also when it is unclear or uncertain. In addition, according to the present invention, by providing predicted values and alarms of the water level, pump pumping volume, etc. together with the predicted value of the sewage inflow volume, it is possible to obtain an inflow volume prediction system that makes it easier to perform the operation management of the sewage treatment plant. 【Brief Description of the Drawings】 【0019】 [Figure 1] It is a schematic configuration diagram showing an example of the pump station 10. [Figure 2] It is a system configuration diagram showing an example of the inflow volume prediction system 100. [Figure 3] It is a diagram showing an outline of the processing procedure of the sewage inflow volume prediction method. [Figure 4] It is a diagram showing the calculation procedure of weather data. [Figure 5]This is a schematic diagram showing an example of a sewage treatment plant 200 with a pump building. [Figure 6] This is a schematic diagram showing an example of a sewage treatment plant 300 that does not have a pump building. [Figure 7] This is a schematic plan view showing the case where the upstream pumping station 500 of the upstream treatment area 501 is connected to the upstream side of the treatment area 401 of the sewage treatment plant 400. [Figure 8] This diagram shows a schematic representation of the processing procedure for predicting the amount of sewage inflow at both the sewage treatment plant 400 and the upstream pumping station 500. [Figure 9] This figure shows an example of a simulation, assuming that it rains in the upstream treatment area 501. [Figure 10] This figure shows an example of a simulation where the pump water output at a future location is specified as an arbitrary value. [Figure 11] This figure shows an example of a simulation where the water level in the tank of the pump station 10 at a future location is set to an arbitrary value. [Modes for carrying out the invention] 【0020】 Embodiments of the present invention will be described in detail below with reference to the drawings. Figure 1 is a schematic diagram showing an example of a pump station 10 controlled by an inflow prediction system 100 according to one embodiment of the present invention, and Figure 2 is a system configuration diagram showing an example of the inflow prediction system 100. 【0021】 Pumping station 10 is installed to collect wastewater discharged from homes and factories in sewer pipes and send it to a sewage treatment plant, or to collect rainwater in sewer pipes and discharge it into rivers, etc. In the case of a combined sewer system, it is installed to send the combined wastewater and rainwater to a sewage treatment plant. 【0022】 As shown in Figure 1, for example, the pump station 10 is configured to include an internal inlet pipe 12 that allows sewage to flow into the pump station 10 from an external inlet pipe 11, an inlet channel 13 consisting of a tank into which sewage from the internal inlet pipe 12 flows, a grit chamber 15 consisting of a tank into which sewage from the inlet channel 13 is introduced and solid matter such as garbage and sand settles, a pump well 17 consisting of a tank into which sewage from which solid matter has been removed in the grit chamber 15 is introduced, pumped up by a pump P and flowed further downstream into a pipe (sewer pipe) 17A, and an inflow rate prediction system 100 that predicts the amount of sewage flowing into the pump station 10 and controls the operating state of the pump station 10. 【0023】 Upstream of the external inlet pipe 11, a network of sewer pipelines that extend throughout the watershed (treatment area) where sewage is collected by the pumping station 10 is connected, and the sewage from the treatment area is collected in the external inlet pipe 11. As mentioned above, the collected sewage may consist only of wastewater, only of rainwater, or a mixture of wastewater and rainwater. 【0024】 Since the inlet pipe 12 within the facility is piping within the pump station 10, its gradient, diameter, and structure are known, and it is controlled as equipment on the pump station 10 side. If the gradient, diameter, and structure of the inlet pipe 12 within the facility are unknown, it can be assumed to be part of the inlet pipe 11 outside the facility, and the following control can be performed. Furthermore, it is possible that the inlet pipe 12 within the facility is not installed, in which case the inlet pipe 11 outside the facility will be directly connected to the inlet channel 13. 【0025】 Water level gauges 19, 21, and 23 are installed in the inflow channel 13, the sedimentation basin 15, and the pump well 17, respectively, to detect changes in the water level of each tank and transmit the detected values ​​to the inflow prediction system 100. Note that water level gauge 21 may not be installed in the sedimentation basin 15; in such cases, the water level of the sedimentation basin 15 may be treated as the same as the water level of the pump well 17. Furthermore, only one water level gauge may be installed, and the water levels of the inflow channel 13, sedimentation basin 15, and pump well 17 may be treated as the same. 【0026】 A gate 25 is installed between the inflow channel 13 and the sedimentation basin 15. The opening degree of this gate 25 is controlled by a control signal from the inflow rate prediction system 100. 【0027】 Multiple pumps P are installed in parallel (only one is shown in Figure 1). The operating state (on / off and rotation speed) of each pump P is controlled by a control signal from the inflow prediction system 100. Pumps P pump up the sewage from the pump well 17 and transfer it to the next stage via piping 17A. 【0028】 As shown in Figure 2, the inflow rate prediction system 100 is configured to include a data collection unit 110 having a facility data collection unit 113 and a weather data collection unit 115, a storage unit 120 for storing various data, a learning processing unit 150 for creating a trained model, a prediction unit 160 for predicting the amount of sewage inflow into the pumping station 10, and a control unit 170 for controlling the operating state of the pumping station 10 based on the predicted value of the sewage inflow rate. 【0029】 The facility data PD collected by the facility data collection unit 113 includes data obtained from the operating status of the facility, the pump station 10, such as water level data from the water level gauges 19, 21, and 23, and pump pumping volume data obtained from the operating status (speed and number of operating pumps) of the pumps P controlled by the control unit 170 described below. The pump pumping volume may also be determined by installing a flow meter separately on the pumps P. The facility data PD also includes various data related to the pump station 10, such as the cross-sectional area, shape and dimensions of each water tank 13, 15, and 17, and other various data, as well as various data such as the gradient, pipe diameter, and structure of the inlet pipe 12 inside the facility, and if known, various data such as the gradient, pipe diameter, and structure of the inlet pipe 11 outside the facility. 【0030】 The meteorological data collected by the meteorological data collection unit 115 is data entered at predetermined time intervals from meteorological data distributed over the network. The meteorological data to be entered is actual measured data (including short-term forecast data; the same applies hereinafter) such as precipitation at the same time in the area including the treatment area of ​​the pumping station 10, as shown in the meteorological data 185 in Figure 4(b) below. Specifically, analyzed rainfall and short-term precipitation forecast data (1km mesh shape data) provided by the Japan Meteorological Agency are used. As described above, the actual measured data is meteorological data in mesh units, and is data that links meteorological values, longitude / latitude, and time information to mesh units, that is, area units (1km mesh shape units) into which the region is divided into multiple small parts. The meteorological values ​​are one or more data including precipitation from among precipitation, weather, temperature, wind speed, wind direction, and atmospheric pressure. Precipitation is the meteorological value that has the greatest influence on the prediction of sewage inflow, but other meteorological values ​​also influence the prediction of sewage inflow. Therefore, by adding weather data other than precipitation to the input values ​​according to the topography and weather characteristics of the treatment area, it becomes possible to further improve the accuracy of sewage inflow prediction. 【0031】 The memory unit 120 includes a data storage unit 121, a trained model storage unit 123, and a geographic information storage unit 125. 【0032】 The data storage unit 121 stores and stores facility data and weather data collected by the facility data collection unit 113 and the weather data collection unit 115 over time. The trained model storage unit 123 stores the trained model created by the learning processing unit 150 described below. This trained model is updated as the inflow forecasting system 100 is put into actual operation and new various data are updated. The geographic information storage unit 125 stores geographic information about the processing area input from a geographic information system (GIS) distributed over the network. 【0033】 The learning processing unit 150 is a means for learning a model to predict the amount of sewage flowing into the pumping station 10 based on the facility data and weather data received from the data storage unit 121, and creating a trained model. The learning processing unit 150 is a means for creating a model using machine learning techniques, and in this example, an RNN (Recurrent Neural Network) is used as the machine learning technique (more specifically, an advanced form of RNN, LSTM (Long Short-Term Memory), is used). RNN is a deep learning technique suitable for predicting time-series data whose values ​​change over time, and is suitable for predictions using time-series data, such as facility data and weather data at predetermined time intervals, as in this inflow prediction system 100. 【0034】 In other words, in an RNN, the hidden layer has a path to transmit the output from the hidden layer at a given time to the hidden layer at the next time, so that the hidden layer at time t accepts input from the hidden layer at the previous time t-1, in addition to input from the input layer at the same time t. This makes it possible to create a neural network that takes into account the influence of time intervals. 【0035】 In this embodiment, the time step, one of the hyperparameters of the RNN (LSTM), is the difference between the peak time of sewage inflow and the peak time of precipitation in the treatment area (approximately 10 hours in this example), and / or the fluctuation period of sewage inflow at the sewage treatment plant (approximately 24 hours in this example). Sewage inflow at sewage treatment plants has a periodicity unique to each individual sewage treatment plant (particularly noticeable at sewage treatment plants in urban areas). Therefore, by using the fluctuation period, learning and prediction values ​​can be calculated while considering the periodicity of sewage inflow fluctuations, thereby reducing computational costs or the number of learning trials, and improving the accuracy of sewage inflow prediction. In this example, using 24 hours as the time step resulted in higher accuracy. 【0036】 The trained model generated by the learning processing unit 150 is stored in the trained model storage unit 123 and is updated at regular intervals. 【0037】 The prediction unit 160 predicts the amount of sewage inflow at future locations based on the trained model stored in the trained model storage unit 123 and the facility data, weather data, and geographic information stored in the data storage unit 121. The accuracy of the predicted value compared to the measured value of the inflow into the pumping station 10 was a correlation coefficient of 0.966 and a root mean square error (RMSE) of 424. 【0038】 The control unit 170 receives a predicted value of the sewage inflow from the prediction unit 160 and controls the opening degree of the gate 25 in the pumping station 10, the number of operating pumps P, and the rotation speed of the operating pumps P, that is, the operating state of the pumping station 10. 【0039】 Figure 3 shows a schematic diagram of a processing procedure in which a trained model of the inflow rate prediction system 100 is constructed at the pumping station 10, the amount of sewage flowing into the pumping station 10 is predicted using this trained model, and the operating state of the pumping station 10 is controlled based on the prediction. In this processing procedure, a trained model that predicts the amount of sewage flowing in is first constructed using measured values ​​such as facility data, and then, during the actual operation of the inflow rate prediction system 100, the operating state of the pumping station 10 is controlled while reconstructing the trained model using new measured values. Note that the order of processing is not limited to this processing procedure and can be changed in various ways. 【0040】 As shown in the figure, the inflow prediction system 100 has a facility data collection unit 113 that collects facility data PD, which includes water level data for the inflow channel 13, sedimentation basin 15, and pump well 17 of the pumping station 10, as well as pump pumping volume data for pump P. 【0041】 Meanwhile, the weather data collection unit 115 collects weather data 185 (Step 2). 【0042】 Next, the prediction unit 160 uses the collected facility data PD to calculate the past and present inflow rates of sewage into the pumping station 10 (step 3). Specifically, the sewage inflow rate is calculated by summing the fluctuations per unit time of the water volume held in each tank, i.e., the inflow channel 13, the sedimentation basin 15, and the pump well 17, which fluctuate in accordance with the fluctuations in the sewage inflow rate of the pumping station 10, the pump pumping rate per unit time of the pump P, and the fluctuations per unit time of the amount of water stored in the facility inflow pipe 12 within the pumping station 10. At this time, the fluctuations per unit time of the amount of sewage stored in the facility inflow pipe 11, which is an off-site pipeline (hereinafter also referred to as "sewage storage fluctuations") are not added to this sewage inflow rate, but are estimated (calculated) separately. 【0043】 The rate of change per unit time of the water volume held in each tank 13, 15, and 17 can be easily and accurately calculated by multiplying the cross-sectional area of ​​each tank 13, 15, and 17 by the rate of change in water level per unit time. Furthermore, the rate of change in the amount of sewage stored in the inflow pipe 12 within the facility can also be calculated by referring to the water level in the inflow channel 13, etc. 【0044】 Next, the amount of sewage storage fluctuation in the external inflow pipe 11 is estimated as follows. [When the gradient, diameter, and structure of the external inlet pipe 11 are known] If the gradient, diameter, and structure of the external inlet pipe 11 are known, the fluctuation in the amount of sewage stored in the external inlet pipe 11 is estimated (calculated) by assuming that the hydraulic gradient between the water level of the inlet channel 13, which is a water tank to which the external inlet pipe 11 is connected via the internal inlet pipe 12, and the water level of the sewage stored in the external inlet pipe 11 is constant, and by referring to the fluctuation in the water level of the inlet channel 13 and the gradient, diameter, and structure of the external inlet pipe 11. 【0045】 When performing the prediction calculation of sewage storage fluctuations, it is necessary to pre-determine a one-to-one correspondence between water level and water volume for the inflow channel 13, which is an in-house tank, or, if considered as an input variable, for the in-house inflow pipe 12 and the out-of-house inflow pipe 11. For the out-of-house inflow pipe 11, the function f representing this correspondence is determined using the pipe's gradient, diameter, and structure, if these are known. If L1 is the water level at time t and L2 is the water level at time t+1, these can be obtained as measured or predicted values, so f(L2)-f(L1) becomes the fluctuation amount of storage volume per unit time. 【0046】 [When the gradient, diameter, and structure of the external inlet pipe 11 are not clear (Part 1)] If the gradient, diameter, and structure of the external inlet pipe 11 are not known, the amount of sewage storage fluctuation in the external inlet pipe 11 is estimated by estimating the correspondence between the water level of the inlet channel 13 and the amount of sewage stored in the external inlet pipe 11, based on the water level fluctuation of the inlet channel 13, which is a water tank to which the external inlet pipe 11 is connected, when the gate 25 installed at the pumping station 10 is closed, and the amount of sewage flowing into the pumping station 10. 【0047】 In this example, gate 25 is located between the inflow channel 13 and the sedimentation basin 15. When gate 25 is closed, incoming sewage will flow only into the external inflow pipe 11, the internal inflow pipe 12, and the inflow channel 13. For example, using data up to the time immediately before closing gate 25, a predicted value of the sewage inflow one hour ahead can be calculated (let's call this predicted value W). For convenience, let f be a function that represents the correspondence between the water level in the inflow channel 13 and the total amount of water present in the inflow channel 13, the internal inflow pipe 12, and the external inflow pipe 11. If L1 is the water level immediately before closing and L2 is the water level one hour later, then W ≈ f(L2) - f(L1). From this, the behavior of f can be estimated to some extent, and based on this, the correspondence between the water level in the inflow channel 13 and the amount of sewage stored in the external inflow pipe 11 can be estimated. 【0048】 [When the gradient, diameter, and structure of the external inlet pipe 11 are not clear (Part 2)] As another method, if the gradient, diameter, and structure of the external inlet pipe 11 are not known, the amount of sewage storage fluctuations within the external inlet pipe 11 can also be estimated and calculated by using multiple candidate values ​​for the gradient, diameter, and structure of the external inlet pipe 11 to calculate multiple candidate values ​​for the amount of sewage storage within the external inlet pipe, and then selecting the candidate value that yields the best prediction accuracy when each of these candidate values ​​is used as one of the input values ​​(explanatory variables) for predicting the amount of sewage inflow. 【0049】 If the above method for estimating sewage storage fluctuations, which uses the water level fluctuations of the inlet channel 13 when gate 25 is closed, proves unsuccessful, this method using multiple candidate values ​​is effective. Specifically, by varying the gradient, pipe diameter, and structure of the external inlet pipe 11 (within reasonable limits), functions f1, f2, f3, ... representing the correspondence between water level and water volume for the external inlet pipe 11 are mechanically generated. Using each of these functions f1, f2, f3, ... the amount of fluctuation per unit time in the sewage storage volume of the external inlet pipe 11 is then calculated, and each of these is individually added as an input variable to repeat the prediction process. Finally, the function fn with the best prediction accuracy is adopted as the calculation process. 【0050】 As described above, by calculating the amount of sewage inflow from the fluctuations in water level, it is possible to calculate the amount of sewage inflow with high accuracy even if a flow meter for measuring the amount of sewage inflow to the pumping station 10 is not installed. 【0051】 As in this embodiment, estimating the amount of sewage stored in the external inflow pipe 11 per unit time separately from the amount of sewage flowing into the pumping station 10, and treating this fluctuation as a separate input value (explanatory variable) from the sewage inflow, makes the system easier to process internally compared to treating the two as a single sewage inflow. This will be explained below. 【0052】 For example, given the water level L and the amount of sewage stored in the external inflow pipe 11 S, the function f such that S = f(L) is complex, and the inverse function f -1It is not possible to determine this. Therefore, when predicting future water levels, if the sewage inflow is defined as "change in sewage storage volume + change in tank water volume + pump pumping volume", 1. Calculate the total volume of water in the external inlet pipe 11 and the water tanks 13, 15, and 17 (and also the internal inlet pipe 12, and so on) within the pumping station 10 based on the current water level (using function f). 2.Calculate the total volume of water in the external inlet pipe 11 and tanks 13, 15, and 17 after 2.1 hours (total volume + predicted inflow - pump output). 3. Calculate the water level after 1 hour (using the inverse function f -1 (Since this is unknown, we have no choice but to estimate it using methods such as the bisection method.) This process is necessary. 【0053】 On the other hand, if the sewage inflow is defined as "fluctuation in tank water volume + pump pumping volume", 1. Calculate the water volume in tanks 13, 15, and 17 within pump station 10 based on the current water level. 2.Calculate the water volume in tanks 13, 15, and 17 after 2.1 hours (water volume in tank + predicted inflow - pump output). 3. Calculate the water level in the tank after 1 hour (it is easy to determine the water level from the volume of water in the tank). 4. Calculate the amount of sewage stored in the external inflow pipe 11 using the water level inside the tank (using function f). This is the process. 【0054】 By defining "fluctuation in tank water volume + pump output" as the sewage inflow and treating "fluctuation in sewage storage volume" as another input (explanatory variable), it becomes easier to perform machine learning and various calculations, and the system can handle these values ​​more easily, compared to treating them together as a single sewage inflow. 【0055】 Next, the prediction unit 160 uses the previously stored geographic information to calculate meteorological data for the treatment area into which sewage flows into the pumping station 10 (step 4). The procedure for calculating meteorological data will be explained using Figure 4. 【0056】 First, the geographic information 181 shown in Figure 4(a) and the meteorological data 185 shown in Figure 4(b) are read out. The geographic information 181 stores the treatment area 183 where the sewer pipes flowing into the pumping station 10 are laid. The treatment area 183 is a mesh-like treatment area 183 whose size (area) matches the mesh size (area) of the meteorological data 185, but the present invention is not limited to this. Then, as shown in Figure 4(c), the geographic information 181 and the meteorological data 185 are superimposed, and as shown in Figure 4(d), the meteorological data 185A for only the portion of the treatment area 183 is extracted. For processing this data extraction, for example, the clipping function in a geographic information system (GIS) that allows for integrated analysis of data by visualizing the geographic information (map data) 181 on a computer and combining it with other information such as meteorological data 185 is used. 【0057】 In this way, out of the mesh-unit weather data 185, the weather data 185A for the mesh portion that overlaps with the treatment area 183 of the pumping station 10 at the same time, i.e., the weather data 185A that has an effect on the amount of sewage flowing into the pumping station 10, is used as the input value for predicting the amount of sewage flowing in. Furthermore, by using a geographic information system (GIS), it is possible to easily and accurately extract only the weather data 185A that overlaps with the treatment area 183 from the weather data 185. From these points, weather data such as precipitation can be accurately calculated for each mesh and each time, and the accurate peak time of precipitation, which is the sum of the precipitation for each mesh, can be calculated, thereby improving the accuracy of sewage flow prediction. 【0058】 Step 5: The learning processing unit 150 is trained to construct a trained model that predicts the amount of sewage flowing into the pumping station 10, using the facility data obtained in Steps 1 to 4, weather data in the treatment area, the amount of sewage flowing into the pumping station 10, and the amount of sewage storage fluctuations in the external inflow pipe 11. 【0059】 The prediction unit 160 uses the calculated current sewage inflow into the pumping station 10, the estimated (calculated) current sewage storage fluctuation in the external inflow pipe 11, and the calculated current weather data 185A in the treatment area 183 as explanatory variables in the constructed trained model, and predicts the amount of sewage that will flow into the pumping station 10 at a future point (for example, one hour later) as the dependent variable (step 6). 【0060】 Then, the system moves to step 7, where, based on the predicted values, the current opening of gate 25, the number of operating pumps P, and the rotation speed of each operating pump P are controlled (step 7). The control may involve, for example, temporarily increasing the number of operating pumps P or increasing the pump rotation speed in advance, based on a predicted increase in the amount of sewage flowing into pump station 10, thereby increasing the amount of sewage flowing downstream from pump station 10 and reducing the amount of sewage held in the sewage pipelines within the treatment area or at pump station 10. Although not shown in the diagram, the system may also control the opening and closing state of another gate installed at any point in the sewage pipelines within the treatment area. 【0061】 Next, after a predetermined time has elapsed (the interval at which the above operation flow is performed) ("Y" in step 8), the process returns to step 1, and, as described above, various measured values ​​are collected and calculated to reconstruct the trained model and control the gates and pumps based on the predicted sewage inflow, and the prediction accuracy of the trained model is improved through this repeated process. At this time, in step 5, the trained model may be reconstructed using not only measured values ​​but also predicted values, but it is also possible to reconstruct the trained model without using predicted values. The reconstruction of the trained model includes, for example, the estimated value of the sewage storage fluctuation amount in the external inflow pipe 11 that was estimated above, along with other elements, in order to improve the prediction accuracy (for example, so that the predicted future water level in the inflow channel 13 approaches the actual water level). 【0062】 The predetermined time in step 8 is the acquisition interval for the facility data PD and weather data 185 used to predict the sewage inflow volume. For example, an hourly interval, more preferably a 30-minute interval, is preferable for improving prediction accuracy. On the other hand, if the acquisition interval is shortened to, for example, 30 minutes or less, the computational cost or memory usage for learning or outputting predicted values ​​in machine learning will increase, which may cause problems in actual operation due to lags in the output values ​​of the predicted values. On the other hand, if the acquisition interval is extended to 3 hours or more, the prediction accuracy will decrease, and the significance of calculating predicted values ​​will be lost if the interval is too long. For these reasons, the acquisition interval is preferably 30 minutes to 3 hours, and even more preferably 30 minutes to 1 hour. 【0063】 As explained above, by using the above-described inflow rate prediction system 100, only the weather data 185A of the mesh portion that overlaps with the treatment area 183 of the pumping station 10, i.e., the weather data 185A that has a high influence on the amount of sewage flowing into the pumping station 10, is used as input values ​​for predicting the amount of sewage flowing in, thereby improving the accuracy of the sewage flow rate prediction. 【0064】 The pump station 10 described above consists of an inlet channel 13, a sedimentation basin 15, and a pump well 17, but the inlet channel 13 and sedimentation basin 15 may be omitted in some cases. Conversely, tanks other than those described above may also be installed (the same applies to the pump building 10-2 described below). In either case, the sewage inflow is calculated by summing the amount of water held in each tank, which fluctuates in accordance with the fluctuations in the sewage inflow, per unit time, and the amount of water pumped per unit time. 【0065】 Furthermore, although the above processing procedure describes the case where the geographic information system (GIS) and weather data 185 are input separately into the inflow forecasting system 100 and the weather data in the processing area is calculated by overlaying both sets of data, it is also possible to input geographic information that has been processed in advance by importing the weather data 185 into the geographic information system (GIS) and then overlaying it into the inflow forecasting system 100, and then calculate the weather data in the processing area. 【0066】 Figure 5 is a schematic diagram showing an example of a sewage treatment plant 200 with a pump building according to another embodiment controlled by the inflow rate prediction system 100. 【0067】 The sewage treatment plant with a pump building 200 is constructed by connecting the pump building 10-2, the primary sedimentation tank 201, the reaction tank 211, the final sedimentation tank 221, and the disinfection equipment 231 from the upstream side using pipes 17A, 201A, 211A, and 221A, respectively. The tanks may also be connected directly to each other without using pipes. 【0068】 Pump building 10-2 has the same configuration as pump station 10 shown in Figure 1, so the same reference numerals are used for the same or equivalent parts as in pump station 10 shown in Figure 1, and detailed explanations are omitted. Also, in Figure 5, the facility inlet pipe 12 shown in Figure 1 is omitted (note that the facility inlet pipe 12 shown in Figure 1 does not necessarily have to be installed). 【0069】 The difference between this pump building 10-2 and the pump station 10 is that, in addition to the gate 25 and pump P installed in the pump building 10-2, the equipment controlled by the flow rate prediction system 100 includes a gate 241 installed at the location where treated sewage from the sewage treatment plant 200 with this pump building is discharged into rivers, etc. 【0070】 In this inflow prediction system 100, as shown in Figure 3, facility data PD is collected, which includes water level data for the inflow channel 13, sedimentation basin 15, and pump well 17 of the pump building 10-2, as well as pump pumping volume data for pump P (Step 1), and simultaneously collects meteorological data 185 (Step 2). 【0071】 Next, the current inflow rate of sewage into the pump building 10-2 and the fluctuation rate of sewage storage in the external inflow pipe 11 are calculated and estimated using the same method as described above (Step 3). Then, using the same method as described above, meteorological data 185A for the treatment area 183 into which sewage flows into the pump building 10-2 are calculated (Step 4). 【0072】 Next, a trained model for predicting sewage inflow is constructed using facility data, weather data, sewage inflow, and sewage storage fluctuations in the external inflow pipe 11 obtained in steps 1 to 4 (step 5). Then, the sewage inflow is predicted using the constructed trained model (step 6), and based on the predicted value, the opening and closing state of gate 25 and pump P, as well as the opening and closing state of gate 241 are controlled (step 7). The acquisition of the above data, retraining of the trained model, and prediction are repeated at predetermined time intervals ("Y" in step 8). For example, based on a predicted value that the amount of sewage flowing into the pump building 10-2 will increase, the number of operating pumps P and their rotation speeds are increased in advance, and at the same time, the opening of gates 25 and 241 is increased, and the water levels of each tank (inflow channel 13, sedimentation basin 15, and pump well 17), primary sedimentation tank 201, reaction tank 211, final sedimentation tank 221, and disinfection equipment 231 in the pump building 10-2 are lowered in advance. This allows the sewage treatment plant 200 with a pump building to be operated and controlled in accordance with a predetermined predicted sewage inflow. 【0073】 The individual tanks constituting the above-mentioned sewage treatment plant with pump building 200 may be omitted in some cases. Conversely, tanks other than those mentioned above may also be installed. 【0074】 Figure 6 is a schematic diagram showing an example of a sewage treatment plant 300 without a pump building, according to yet another embodiment controlled by the inflow rate prediction system 100. 【0075】 This sewage treatment plant 300 is constructed by connecting the primary sedimentation tank 301, reaction tank 311, final sedimentation tank 321, and disinfection equipment 331 from the upstream side using pipes 301A, 311A, and 321A, respectively. The tanks may also be connected directly to each other without using pipes. 【0076】 Since this sewage treatment plant 300 does not have a pump building, water level gauges 303, 313, 323, and 333 are installed in the primary sedimentation tank 301, reaction tank 311, final sedimentation tank 321, and disinfection equipment 331, respectively. These water level gauges 303, 313, 323, and 333 detect changes in the water level of each tank and transmit the detected values ​​to the inflow rate prediction system 100, which is used to calculate the amount of sewage flowing into this sewage treatment plant 300. 【0077】 Furthermore, an external inlet pipe 351 is connected to the primary sedimentation tank 301 of this sewage treatment plant 300. Also, in Figure 6, the internal inlet pipe 12 shown in Figure 1 is omitted (note that the internal inlet pipe 12 shown in Figure 1 does not necessarily have to be installed). Upstream of the external inlet pipe 351, the sewage pipeline network that runs throughout the watershed (treatment area) that collects sewage to this sewage treatment plant 300 is connected, and the sewage from that treatment area is collected. As mentioned above, the collected sewage may consist only of wastewater, only of rainwater, or a mixture of wastewater and rainwater. 【0078】 A gate 371 is installed at the point where the inlet pipe 351 outside the facility of the primary sedimentation tank 301 is connected, and another gate 391 is installed at the location where treated wastewater from the sewage treatment plant 300 is discharged into rivers or other bodies of water. The opening of each gate 371 and 391 is controlled by the inflow rate prediction system 100. 【0079】 In this inflow prediction system 100, first, as shown in Figure 3, water level data from the primary sedimentation tank 301, reaction tank 311, final sedimentation tank 321, and disinfection equipment 331 of the sewage treatment plant 300 are collected as facility data PD (Step 1), and at the same time, meteorological data 185 is collected (Step 2). 【0080】 Next, the current inflow rate of sewage into the sewage treatment plant 300 and the fluctuation in the amount of sewage stored in the external inflow pipe 351 are calculated and estimated using the facility data PD collected above (Step 3). Specifically, the sewage inflow rate is calculated by summing the fluctuation rate per unit time of the amount of water held in each tank, namely the primary sedimentation tank 301, reaction tank 311, final sedimentation tank 321, and disinfection equipment 331, which fluctuate in accordance with the fluctuation in the amount of sewage inflow into the sewage treatment plant 300, and, if necessary, the fluctuation rate per unit time of the amount of water held in the pipes 301A, 311A, and 321A connecting each of these tanks. 【0081】 The amount of water volume held in each tank 301, 311, 321, and 331 fluctuates per unit time. Specifically, this can be easily and accurately calculated by multiplying the cross-sectional area of ​​each tank 301, 311, 321, and 331 by the amount of water level fluctuation per unit time. Furthermore, the amount of sewage storage fluctuation in the external inflow pipe 351 can be estimated using the same method as described in the embodiment shown in Figure 1 above. 【0082】 Next, using the same method as described above, meteorological data 185A for the treatment area 183 into which sewage flows into the sewage treatment plant 300 is calculated (Step 4). Then, using the facility data, meteorological data, sewage inflow, and sewage storage fluctuations in the external inflow pipe 351 obtained in Steps 1 to 4, a trained model for predicting sewage inflow is constructed (Step 5). 【0083】 Then, the trained model is used to predict the amount of sewage flowing in (step 6), and based on the predicted value, the opening and closing states of gates 371 and 391 are controlled (step 7). The acquisition of the above data, retraining of the trained model, and prediction are repeated at predetermined time intervals ("Y" in step 8). The control may involve, for example, increasing the opening of gates 371 and 391 in advance and lowering the water levels in each tank within the sewage treatment plant 300 in advance, based on a predicted increase in the amount of sewage flowing into the sewage treatment plant 300. This allows the sewage treatment plant 300 to be operated and controlled in accordance with the predicted amount of sewage flowing in. 【0084】 The above-mentioned sewage treatment plant 300 consists of a primary sedimentation tank 301, a reaction tank 311, a final sedimentation tank 321, and a disinfection facility 331. However, any of these tanks may be omitted, or conversely, tanks other than those described above may be installed. 【0085】 Figure 7 is a schematic plan view showing the case where an upstream pumping station 500, which collects sewage from the upstream treatment area 501, is connected to one of the sewage pipelines in the treatment area 401 of the sewage treatment plant 400. 【0086】 The upstream pumping station 500 is a relay pumping station and has the same configuration as the pumping station 10 shown in Figure 1. The sewage treatment plant 400 has the same configuration as the sewage treatment plant 200 with a pump building shown in Figure 5 or the sewage treatment plant 300 shown in Figure 6. In the same figure, r1 shows an image of rainfall on the upstream treatment area 501, and r2 shows an image of rainfall on the treatment area 401. As the pumping station 500 is located upstream of the sewage treatment plant 400, the amount of sewage flowing into the sewage treatment plant 400 is affected by the amount of sewage pumped at the pumping station 500. 【0087】 Alternatively, the upstream pump station 500 may have a configuration similar to the sewage treatment plant 200 with a pump building shown in Figure 5 or the sewage treatment plant 300 shown in Figure 6, or the sewage treatment plant 400 may have a configuration similar to the pump station 10 shown in Figure 1. Furthermore, the upstream pump station 500 may be a pipeline facility with means for measuring inflow volume (for example, a facility with a water level meter and flow meter installed in a manhole that serves as a water tank). 【0088】 Figure 8 shows an outline of a processing procedure in which trained models of the inflow rate prediction system 100 are constructed at both the sewage treatment plant 400 and the upstream pumping station 500, these trained models are used to predict the inflow rate of sewage flowing into the sewage treatment plant 400 and the upstream pumping station 500, and the operating state of the sewage treatment plant 400 and the upstream pumping station 500 is controlled based on the prediction. Note that the order of operations is not limited to this processing procedure and can be changed in various ways. In this example, the operation control of the sewage treatment plant 400 and the upstream pumping station 500 is performed simultaneously by the same inflow rate prediction system 100, but separate inflow rate prediction systems 100 may be installed and controlled separately at the same time. The inflow rate prediction system 100 used in this example has the same configuration as the inflow rate prediction system 100 shown in Figure 2 above. 【0089】 In Figure 8, the inflow prediction system 100 first collects facility data PD and meteorological data 185 at the upstream pumping station 500 and the sewage treatment plant 400, respectively (steps 1,1A, 2,2A). 【0090】 Next, the sewage inflow volume (sewage inflow volume into the upstream pumping station 500 and sewage inflow volume into the sewage treatment plant 400) and the sewage storage fluctuation volume in the external inflow pipes of each facility are calculated and estimated from the above facility data PD (Step 3, 3A). 【0091】 Next, meteorological data 185A for treatment areas 401 and 501 is calculated from the respective meteorological data 185 (Step 4, 4A). Then, a trained model for predicting sewage inflow is constructed using the facility data, meteorological data, sewage inflow, and sewage storage fluctuations in the inflow pipe outside the facility obtained in Steps 1, 1A to 4, 4A (Step 5, 5A). It is preferable to further use the measured pumping volume of the pumps at the upstream pumping station 500 when constructing the trained model to be used for the sewage treatment plant 400. 【0092】 Next, at the upstream pumping station 500, the constructed trained model is used to predict the amount of sewage inflow to the upstream pumping station 500, the current amount of sewage storage fluctuation in the external inflow pipe estimated separately, and the current weather data within the treatment area 501 calculated above, as explanatory variables (step 6A). Furthermore, using the predicted amount of sewage inflow, a predicted value for the amount of water pumped by the pumps at the upstream pumping station 500 is calculated (step 6B). This step 6B, which calculates the predicted amount of water pumped at the upstream pumping station 500, is called the upstream water pumping prediction means. 【0093】 Meanwhile, at the sewage treatment plant 400, the trained model is constructed to predict the amount of sewage inflow (step 6). In doing so, the current amount of sewage inflow into the sewage treatment plant 400 calculated above, the separately estimated fluctuations in sewage storage in the external inflow pipe, the current weather data 185A within the treatment area 401 calculated above, and the predicted pumping volume of the upstream pumping station 500 obtained in step 6B (and also the current pumping volume of the upstream pumping station 500) are input into the trained model for the sewage treatment plant 400 as explanatory variables to predict the amount of sewage inflow into the sewage treatment plant 400 at future locations (step 6). 【0094】 With this configuration, even if there is an upstream pumping station 500 that collects sewage from the upstream treatment area 501 and flows it into the downstream treatment area 401, it becomes possible to predict the amount of sewage inflow to the downstream sewage treatment plant 400 with high accuracy and ease by calculating the predicted pumping volume of the upstream pumping station 500 and inputting it into a trained model for the sewage treatment plant 400. In other words, by inputting the predicted pumping volume of the upstream pumping station 500 into a trained model for the sewage treatment plant 400, data from the current pumping volume of the upstream pumping station 500 to the future pumping volume can be used as data for predicting the amount of sewage inflow to the sewage treatment plant 400, thus enabling more accurate predictions. 【0095】 Then, based on the predicted values ​​mentioned above, the gates and pumps of the upstream pumping station 500 and the sewage treatment plant 400 are controlled (steps 7, 7A), and the trained model is repeatedly retrained and predicted at predetermined time intervals ("Y" in steps 8, 8A). 【0096】 The embodiment shown in Figure 7 illustrates a case where an upstream pumping station 500, which collects sewage from an upstream treatment area 501, is connected to a sewage pipeline in the treatment area 401 of a sewage treatment plant 400. However, the present invention can also be similarly applied when there are three or more multi-stage treatment areas connected, such as when an upstream pumping station that collects sewage from an even further upstream treatment area is connected to a sewage pipeline in the treatment area 501 of the upstream pumping station 500, or when a pipeline facility with an inflow rate measuring means is connected. 【0097】 When predicting the inflow into the sewage treatment plant 400 using the present invention, if the predicted pumping volume of the upstream pumping station 500 was not used as an explanatory variable, the accuracy of the predicted inflow volume after 1 hour relative to the measured value was a correlation coefficient of 0.914 and a root mean square error (RMSE) of 687. Furthermore, the accuracy of the predicted inflow volume after 10 hours relative to the measured value was a correlation coefficient of 0.719 and a root mean square error (RMSE) of 1186. On the other hand, when the inflow volume of a pipeline facility equipped with an inflow volume measuring means and its predicted value were used as explanatory variables instead of the upstream pumping station 500, the accuracy of the predicted inflow volume after 1 hour relative to the measured value was a correlation coefficient of 0.934 and a root mean square error (RMSE) of 615. Furthermore, the accuracy of the predicted inflow volume after 10 hours relative to the measured value was a correlation coefficient of 0.798 and a root mean square error (RMSE) of 1051. In this way, by inputting the predicted values ​​of the pumping volume or inflow volume of the pipeline facility having a means for measuring the inflow volume of the upstream pumping station 500 into a trained model for the sewage treatment plant 400, the accuracy of the inflow volume prediction to the sewage treatment plant 400 can be improved. 【0098】 As described above, the present invention constructs a trained model that predicts sewage inflow using mesh-level meteorological data, and can calculate the degree of influence of meteorological data on sewage inflow in each mesh. If the meteorological data includes precipitation and the target of the prediction is a separate sewer system, it is also possible to identify areas with a high degree of influence as areas with a high probability of generating unidentified water. Furthermore, by using meteorological data for each mesh portion that overlaps with the area of ​​a pumping station or pump building of a sewage treatment plant, pipeline facilities with inflow measurement means, or the treatment area of ​​a sewage treatment plant, more accurate identification becomes possible. 【0099】 Furthermore, in this invention, the trained model may be rebuilt at regular intervals, for example, every six months to five years, preferably every six months to two years. By using the trained model built at regular intervals to calculate the degree of influence of rainfall on sewage inflow for each mesh and the change in that influence over each period, it becomes possible to take measures such as issuing a warning or prioritizing the repair of pipelines in areas where the change in the influence is large, assuming that the deterioration of pipelines in that mesh has progressed during that period. In other words, the accuracy of the information can be improved by setting the period to six months or more, and timely repair responses can be made by setting the period to five years or less. 【0100】 Various simulation functions can be installed in the inflow prediction system 100. For example, the inflow prediction system 100 installed in a sewage treatment plant 400 having an upstream pumping station 500 may be equipped with simulation means to simulate the increase or decrease in sewage inflow at downstream pumping stations, etc., in cases such as assuming that it rains in the upstream treatment area 501 of the upstream pumping station 500 even though it is not currently raining, or assuming that the amount of rainfall increases or decreases even though it is currently raining in the upstream treatment area 501 of the upstream pumping station 500, or assuming that the rain stops even though it is currently raining in the upstream treatment area 501 of the upstream pumping station 500. 【0101】 Figure 9 is a diagram illustrating a processing procedure, showing an example of a simulation in which it is assumed that rain falls in the upstream treatment area 501 of the upstream pumping station 500 even though it is not currently raining, or in which it is assumed that the amount of rainfall increases or decreases in the upstream treatment area 501 of the upstream pumping station 500. 【0102】 In this case, the operator sets and inputs, for example, an arbitrary value for the predicted increase or decrease in the pumping volume of the upstream pumping station 500 (or, if a pipeline facility with an inflow measurement means is installed instead of the upstream pumping station 500, the predicted increase or decrease in the sewage inflow volume at that pipeline facility) from an input means (not shown) installed in the inflow volume prediction system 100 (step 101). This predicted increase or decrease may be automatically generated by the inflow volume prediction system 100. 【0103】 Next, the predicted value of the actual pump pumping volume at the operational upstream pump station 500, which was predicted in step 6B of Figure 8, is added to the arbitrary predicted value (increase / decrease value) entered above to obtain a new predicted value of the pump pumping volume at the upstream pump station 500 (step 102). 【0104】 The new predicted values, along with other factors, are input into the trained model constructed in step 5 of Figure 8 for the operational sewage treatment plant 400 to predict the sewage inflow into the sewage treatment plant 400 (step 103), and the results of the simulation are output (step 104). The output destination may be the display or printing means of the inflow prediction system 100, or various other devices. 【0105】 According to this simulation method, the prediction of the sewage inflow at the downstream sewage treatment plant 400 by the prediction unit 160 uses the predicted value of the pumping volume of the upstream pumping station 500 (or the predicted value of the sewage inflow volume at the pipeline facility having an inflow volume measuring means for the upstream treatment area 501), which is set to an arbitrary value. This makes it possible to simulate fluctuations in the increase or decrease of sewage inflow at the downstream pumping station, etc., in cases such as assuming that it rains in the treatment area of ​​the upstream pumping station even though it is not currently raining, or assuming that it is raining but the rain becomes heavier or lighter. This allows operators to consider countermeasures for various conditions that may occur in the future. 【0106】 Next, the present invention may also provide, for example, an inflow rate prediction system 100 installed in the pumping station 10 shown in Figure 1, with another simulation means, which predicts the water level of one or more tanks 13, 15, 17 of the pumping station 10 at the future point by specifying an arbitrary value for the amount of water pumped per unit time at the future point. 【0107】 Figure 10 is a diagram illustrating a processing procedure for an example simulation where the pump water volume per unit time at a future location is specified to an arbitrary value (it may also be the current value). In this figure, first, for example, the operator specifies the pump water volume per unit time at a future location to an arbitrary value using an input means (not shown) installed in the inflow prediction system 100 (step 201). This specified pump water volume may be automatically generated by the inflow prediction system 100. 【0108】 Next, using the current water level and structure data of one or more tanks 13, 15, 17 of the pumping station 10, the current pumping rate per unit time of the operating pumping station 10, the predicted value of the sewage inflow rate of the operating pumping station 10, and the pumping rate per unit time of the specified future location, the water level of one or more tanks 13, 15, 17 of the pumping station 10 at the specified future location is predicted (step 202). 【0109】 This simulation method allows for the prediction of how the water level in a tank will fluctuate at a future location, for example, if the pumping volume at a future location is set to a predetermined constant value, based on predicted values ​​of sewage inflow. By referring to this prediction, it becomes possible to plan and manage the pumping volume or the amount of effluent discharged from the tank following the pumping station 10. This simulation method can also be used for pump buildings within sewage treatment plants. 【0110】 Next, the present invention may also include, for example, in the inflow rate prediction system 100 installed in the pumping station 10 shown in Figure 1, a simulation means that predicts the amount of water pumped per unit time at a future location by specifying the water levels of one or more water tanks 13, 15, 17 of the pumping station 10 at a future location to arbitrary values. 【0111】 Figure 11 is a diagram illustrating a processing procedure for an example of a simulation in which the water levels of one or more tanks 13, 15, 17 of the pump station 10 at a future location are specified to any value (the current value may also be used). In this figure, first, for example, an operator specifies the water levels of one or more tanks 13, 15, 17 of the pump station 10 at a future location to any value (the current value may also be used) using an input means (not shown) installed in the inflow prediction system 100 (step 301). Note that the specified water levels may be automatically generated by the inflow prediction system 100. 【0112】 Next, using the current water level and structure data of one or more tanks 13, 15, 17 of the pumping station 10, the predicted sewage inflow rate of the operating pumping station 10, and the water levels of the tanks 13, 15, 17 at the specified future location, the pumping rate per unit time at the future location is predicted so as to maintain the water levels of one or more tanks 13, 15, 17 of the pumping station 10 at the specified arbitrary value (step 302). 【0113】 This simulation method allows for simulations such as predicting how the pumping volume of the pump station 10 will fluctuate at a future location, based on predicted values ​​of sewage inflow, for example, when the water levels of the tanks 13, 15, and 17 of the pump station 10 at a future location are set to a predetermined constant value. For example, by setting the predicted pumping volume as a recommended value, it becomes possible to plan and manage the pumping volume or the amount of discharged water from the subsequent tanks of the pump station 10 by referring to this value. This simulation method can also be used for pump buildings within sewage treatment plants. 【0114】 Next, the present invention may also include a means for outputting long-term past prediction values ​​that can use past data stored in the data storage unit 121 of the inflow prediction system 100 to verify trends in long-term past prediction values ​​and model performance. 【0115】 The long-term historical forecast output means uses facility data and weather data stored in the data storage unit 121 over an arbitrary long period of time, and based on this long-term historical data, predicts the sewage inflow for all unit time periods included within that period, using the facility data and weather data prior to each unit time period. It then outputs a forecast value equivalent to that obtained if the sewage inflow prediction was performed for all unit time periods included within that period. The output may be made to a display means (not shown) of the inflow prediction system 100, to a printing means (not shown), or to any other device. 【0116】 The inflow prediction system 100 predicts sewage inflow using facility data and meteorological data for the minimum period necessary for the prediction. However, by using the long-term historical prediction output means, longer-term facility data and meteorological data can be used as input data, allowing historical prediction values ​​of sewage inflow for each unit time within that period to be output as a single database. This enables monitoring of the performance of the trained model over long periods and trend analysis of prediction values. 【0117】 In other words, instead of real-time data, we can obtain "predicted values ​​that would have been obtained if a prediction had been made at a certain point in the past." By varying this "certain point in time," we can output predicted values ​​using data from 0:00 to 10:00, predicted values ​​using data from 1:00 to 11:00, predicted values ​​using data from 2:00 to 12:00, and so on, for every unit of time within that period. This data can be used, for example, to determine the accuracy of a trained model. 【0118】 Next, the present invention may also provide an alarm means that outputs an alarm by combining conditions of various variables, such as facility data and weather data stored in the data storage unit 121 of the inflow prediction system 100, and various data calculated (estimated) using these data. 【0119】 The alarm system allows the user to specify any type of variable from among facility data and meteorological data used to predict sewage inflow, various calculated values ​​for the prediction calculated (estimated) using said facility data and meteorological data (such as the predicted pumping volume of the upstream pumping station 500 and the calculated values ​​of meteorological data in the treatment area), and the predicted sewage inflow. By specifying arbitrary reference values ​​and logical operators for these specified variables, the system sets arbitrary types of alarm output conditions and outputs an alarm based on the prediction results. 【0120】 This alarm system allows for the specification of reference values ​​and logical operators for any type of variable. For example, when one or more variables exceed the reference value, an alarm is output. By combining reference values ​​for multiple variables, such as the predicted sewage inflow rate, the predicted pumping rate of the upstream pumping station 500, and calculated meteorological data in the treatment area, it is possible to output multiple alarms with probabilistically different characteristics. This allows for the visualization of dangerous conditions in an easily understandable way, which can be used to make operational decisions at the sewage treatment plant. 【0121】 For example, "The predicted value of the sewage inflow is 200 (m³) 3If the pumping volume of the upstream treatment area 501 is 100 (m³ / min) or more, or if the predicted value of the pumping volume of the upstream treatment area 501 3 hours prior is 100 (m³ / min) or more 3 Alarm 1 is set with the condition "the predicted average rainfall in the area 3 hours ago is 5 (mm / km) or more", and Alarm 1 is set with this condition, and "the predicted average rainfall in the area 3 hours ago is 5 (mm / km) or more", and 2 Alarm 2 is set under the condition "220 (m³) or more" 3 You can set alarm 3 based on conditions such as "more than / min". 【0122】 As described above, by using the various simulation and alarm means of the present invention, it is possible to obtain and utilize data (such as simulation data and alarm data) that is easier to use in operation from predicted sewage inflow and other various data. 【0123】 Although embodiments of the present invention have been described above, the present invention is not limited to the above embodiments, and various modifications are possible within the scope of the claims, specification, and drawings. Furthermore, any configuration not directly described in the specification and drawings is within the scope of the technical idea of ​​the present invention as long as it achieves the function and effect of the present invention. Also, the embodiments described above and shown in the figures can be combined as long as there is no contradiction in their purpose and configuration. Moreover, even a part of the descriptions above and shown in the figures can constitute an independent embodiment, and the embodiments of the present invention are not limited to a single embodiment combining the above descriptions and figures. [Explanation of Symbols] 【0124】 10...Pumping station, 11...External inlet pipe, 12...Internal inlet pipe, 13...Inlet channel (water tank), 15...Sedimentation basin (water tank), 17...Pump well (water tank), 19,21,23...Water level gauge, 25...Gate, P...Pump, 100...Inflow prediction system, 110...Data collection unit, 113...Facility data collection unit, 115...Meteorological data collection unit, 120...Storage unit, 121...Data storage unit, 123...Trained model storage unit, 125...Geographic information storage unit, 150...Learning processing unit, 160...Prediction unit, 170...Control unit, PD...Facility data, GIS...Geographic information system, 181...Geographic information, 183...Processing area, 185, 185A...Meteorological data, 200...Sewage treatment plant with pump building, 10-2...Pump building, 201...Primary sedimentation tank (water tank), 211...Reaction tank (water tank), 221...Final sedimentation tank (water tank), 231...Disinfection equipment (water tank), 241...Gate, 300...Sewage treatment plant without pump building, 301...Primary sedimentation tank (water tank), 311...Reaction tank (water tank), 321...Final sedimentation tank (water tank), 331...Disinfection equipment (water tank), 303, 313, 323, 333...Water level gauge, 351...Outside inlet pipe, 371, 391...Gate, 400...Sewage treatment plant, 401...Treatment area, 500...Upstream pumping station, 501...Upstream treatment area.

Claims

[Claim 1] In a pumping station or a pump building within a sewage treatment plant, or in a pipeline facility having means for measuring inflow, or in an inflow rate prediction system for predicting sewage inflow at a sewage treatment plant, A facility data collection unit that collects facility data of the pump building or pipeline facilities with inflow rate measuring means within the pumping station or sewage treatment plant or the sewage treatment plant, The weather data collection unit collects weather data, A data storage unit that stores the aforementioned facility data and weather data, A learning processing unit that learns a model to predict the amount of sewage flowing into a pump building or pipeline facility with an inflow measurement means or sewage treatment plant, or a pump building or sewage treatment plant, based on the facility data and weather data received from the data storage unit, and creates a trained model. A trained model storage unit that stores the trained model created by the aforementioned learning processing unit, A prediction unit that predicts the amount of sewage inflow at a future location based on the trained model, the facility data, and the weather data, A control unit receives a predicted value of the sewage inflow from the prediction unit and controls the operating state of the pump building or pipeline facility having an inflow measurement means or the sewage treatment plant within the pumping station or sewage treatment plant, It has, If the target for predicting the sewage inflow is a pumping station or a pump building within a sewage treatment plant, the sewage inflow is calculated by summing the amount of water held in one or more tanks within the pumping station or pump building within the sewage treatment plant per unit time, and the amount of water pumped per unit time. On the other hand, if the target for predicting the sewage inflow is a pipeline facility or sewage treatment plant equipped with the inflow measurement means, the sewage inflow is calculated as the amount of fluctuation per unit time of the amount of water held in one or more tanks owned by the pipeline facility or sewage treatment plant. Furthermore, the inflow rate prediction system is characterized by estimating the amount of fluctuation per unit time of the amount of sewage stored in a pump building or pipeline facility with an inflow rate measuring means within the pumping station or sewage treatment plant, or in an external inflow pipe connected to the upstream side of the sewage treatment plant, and using the estimated amount of fluctuation as one of the input values ​​used to predict the amount of sewage inflow. [Claim 2] The amount of fluctuation per unit time of the amount of sewage stored in the inflow pipe outside the facility is, If the gradient, diameter, and structure of the inflow pipe outside the facility are known, The inflow rate prediction system according to claim 1, characterized in that it estimates the fluctuations in the water level of the tank to which the external inflow pipe is connected, and the gradient, pipe diameter, and structure of the external inflow pipe. [Claim 3] The amount of fluctuation per unit time of the amount of sewage stored in the inflow pipe outside the facility is, If the gradient, diameter, and structure of the inflow pipe outside the facility are not clear, The inflow rate prediction system according to claim 1, characterized in that it estimates and calculates the correspondence between the water level in the tank to which the inflow pipe outside the facility is connected and the amount of sewage stored in the inflow pipe outside the facility, based on the fluctuation of the water level in the tank to which the inflow pipe outside the facility is connected when the pump building or pipeline facility having an inflow rate measuring means or a gate installed in the pumping station or sewage treatment plant is closed, and the amount of sewage flowing into the pump building or pipeline facility having an inflow rate measuring means or a sewage treatment plant. [Claim 4] The amount of fluctuation per unit time of the amount of sewage stored in the inflow pipe outside the facility is, If the gradient, diameter, and structure of the inflow pipe outside the facility are not clear, Using multiple candidate values ​​for the gradient, diameter, and structure of the external inflow pipe, multiple candidate values ​​for the amount of sewage stored in the external inflow pipe are calculated. The inflow rate prediction system according to claim 1, characterized in that it estimates and calculates the amount of sewage stored in the inflow pipe outside the facility by adopting the candidate value that provides the best prediction accuracy when each of these candidate values ​​is used as one of the input values ​​for predicting the amount of sewage inflow. [Claim 5] At least one of the following is installed in the pumping station or the pump building within the sewage treatment plant or in a pipeline facility having means for measuring inflow, or in the treatment area of ​​the sewage treatment plant: an upstream pumping station that collects and flows in sewage from an upstream treatment area located upstream of the treatment area, or a pipeline facility having means for measuring inflow from an upstream treatment area located upstream of the treatment area. After specifying an arbitrary value for the predicted volume of water pumped at the upstream pumping station, or the predicted volume of sewage inflow at the pipeline facility equipped with an inflow measurement means for the upstream treatment area, The inflow rate prediction system according to claim 1, characterized in that the prediction unit at the pumping station or pump building within the sewage treatment plant or pipeline facility having an inflow rate measuring means or sewage treatment plant located downstream uses a predicted value of the pumping volume of the upstream pumping station, or a predicted value of the sewage inflow volume of the pipeline facility having an inflow rate measuring means in the upstream treatment area, which is set to an arbitrary value, for predicting the sewage inflow rate. [Claim 6] If the target for predicting the amount of sewage inflow is the pumping station or the pump building within the sewage treatment plant, By specifying the pump water output per unit time at a future location to any value, or to the current value, Using the water level and structure of one or more tanks in the aforementioned pumping station or pump building, the pumping rate per unit time, and the predicted value of the sewage inflow, The inflow prediction system according to claim 1, characterized in that it predicts the water level at the future point in one or more water tanks of the pumping station or pumping building. [Claim 7] If the target for predicting the amount of sewage inflow is the pumping station or the pump building within the sewage treatment plant, By specifying the water level of one or more tanks in the pump station or pump building at a future location to an arbitrary value, or to the current value, Using the water level and structure of one or more tanks in the pumping station or pumping building, and the predicted value of the sewage inflow, The inflow rate prediction system according to claim 1, characterized in that it predicts the pump pumping rate per unit time at a future location such that the water level of one or more tanks in the pumping station or pumping building is maintained at the specified arbitrary value or the specified current value. [Claim 8] The inflow rate prediction system according to claim 1, characterized in that, based on facility data and weather data over an arbitrary long period, it predicts the sewage inflow rate for each unit time included within that period using the facility data and weather data prior to each unit time, and outputs a predicted value equivalent to that which would be obtained if the sewage inflow rate prediction was performed for each unit time included within that period. [Claim 9] The inflow rate prediction system according to claim 1, characterized in that it sets any type of alarm output condition and outputs an alarm for the prediction result by specifying any type of variable from among the facility data and weather data used to predict the sewage inflow rate, various calculated values ​​for prediction calculated using said facility data and weather data, and the predicted sewage inflow rate, and by specifying any reference value and logical operator.

Images