Injection rate calculation system, water treatment system, and injection rate calculation method

Abstract

To improve the accuracy of the injection ratio of a chemical to be injected to treat raw water flowing into a water purification plant.SOLUTION: An injection ratio calculation system (2) comprises an acquiring unit (21) acquiring upstream data measured at least on one of a position of injecting a chemical and an upstream position relative to the injection position and an evaluation index evaluating the state of mixed water injected with a chemical, an injection ratio estimation unit (22) of estimating the injection ratio of the chemical from the upstream data using an estimation model, an injection ratio correction unit (41) of correcting the estimated injection ratio according to the evaluation index, and an updating unit (51) of updating the estimation model using teacher data using the upstream data as explanatory variables and using the corrected injection ratio as an object variable.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a technique for calculating the injection rate of chemicals used in water purification facilities.

Background Art

[0002] In a water treatment system, various chemicals such as chlorine and flocculants are added to produce water with a quality that meets predetermined standards. Therefore, it is important to control the amount of the above chemicals to be added.

[0003] The model generation device described in Patent Document 1 generates a regression model for estimating the injection rate of a flocculant to be injected into influent water flowing into a water purification plant. The model generation device clusters a set of data sets including water quality data including the turbidity before injection, which is the turbidity of the influent water before injecting the flocculant, and the injection rate of the flocculant injected into the influent water, to generate a plurality of clusters. Next, the model generation device generates the regression model for each of the plurality of generated clusters using the water quality data as an explanatory variable and the injection rate as an objective variable for learning data. [[ID=1十八]]

[0004] The flocculant injection control device described in Patent Document 2 performs feedback control using the aggregation state of flocs in the mixed water, which is the treated water into which the flocculant has been injected, as a control amount and the injection rate of the flocculant with respect to the treated water as a manipulated variable. Further, the flocculant injection control device determines the target value of the control amount in the feedback control based on the turbidity of a sedimentation tank that precipitates flocs in the mixed water and its target value.

[0005] In the flocculant injection control system described in Patent Document 3, the raw water turbidity charge amount estimation means estimates the charge amount on the surface of turbidity particles contained in the raw water, and the charge state measurement means after flocculation measures the charge state after injecting the flocculant. The flocculant injection rate setting device calculates a basic flocculant injection rate according to the charge amount on the surface of the turbidity particles contained in the raw water, corrects the basic flocculant injection rate according to the charge state after injecting the flocculant, and determines the actual injection rate.

Prior Art Documents

[0006] [Patent Document 1] Japanese Patent Publication No. 2021-098191 [Patent Document 2] Japanese Patent Publication No. 2019-193916 [Patent Document 3] Japanese Patent Publication No. 2002-205076 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] In the model generation device described in Patent Document 1, if the training data no longer conforms to the current conditions of the influent water, the estimated value of the coagulant injection rate may not be the optimal value. In the coagulant injection control device described in Patent Document 2, there is a time lag between measuring the turbidity of the sedimentation tank and changing the coagulant injection rate. Therefore, if the water quality of the treated water changes suddenly, the turbidity of the sedimentation tank will temporarily worsen. By combining feedforward (FF) control and feedback (FB) control, as in Patent Document 3, the above problems of Patent Documents 1 and 2 are solved to some extent, but there is still room for improvement.

[0008] One aspect of the present invention aims to realize an injection rate estimation system that can improve the accuracy of the injection rate of chemicals injected to treat raw water flowing into a water purification facility. [Means for solving the problem]

[0009] To solve the above problems, an injection rate calculation system according to one aspect of the present invention is an injection rate calculation system for calculating the injection rate of a chemical agent to be injected to treat raw water flowing into a water purification facility, comprising: an acquisition unit that acquires upstream data measured at least one of the injection position of the chemical agent and a position upstream of the injection position, and an evaluation index for evaluating the state of the mixed water after the chemical agent has been injected; an estimation unit that estimates the injection rate of the chemical agent from the upstream data using an estimation model; a correction unit that corrects the injection rate estimated by the estimation unit based on the evaluation index; and an update unit that updates the estimation model using training data in which the upstream data is an explanatory variable and the injection rate corrected by the correction unit is an objective variable.

[0010] An injection rate calculation system according to another aspect of the present invention is an injection rate calculation system for calculating the injection rate of a chemical agent to be injected to treat raw water flowing into a water purification facility, comprising: an acquisition unit that acquires upstream data measured at at least one of the injection position of the chemical agent and a position upstream of the injection position; and an optimization unit that optimizes the injection rate of the chemical agent using an estimation model constructed with training data in which the upstream data and the injection rate of the chemical agent are explanatory variables and an evaluation index for evaluating the state of the mixed water after the chemical agent has been injected is the objective variable, so that the evaluation index estimated from the upstream data and the injection rate of the chemical agent becomes a target value.

[0011] A water treatment system according to yet another aspect of the present invention includes an injection rate calculation system having the above configuration, and a control device that controls the injection of the chemical agent according to the injection rate calculated by the injection rate calculation system.

[0012] A method for calculating an injection rate according to yet another aspect of the present invention includes: an acquisition step of acquiring upstream data measured at at least one of the injection position of a chemical agent to be injected to treat raw water flowing into a water purification facility and a position upstream of the injection position, and an evaluation index for evaluating the state of the mixed water after the chemical agent has been injected; an estimation step of estimating the injection rate of the chemical agent from the upstream data using an estimation model; a correction step of correcting the injection rate estimated in the estimation step based on the evaluation index; and an update step of updating the estimation model using training data in which the upstream data is an explanatory variable and the injection rate corrected by the correction unit is an objective variable.

[0013] A method for calculating an injection rate according to yet another aspect of the present invention includes: an acquisition step of acquiring upstream data measured at at least one of the injection position of a chemical agent to be injected to treat raw water flowing into a water purification facility and a position upstream of the injection position; and an optimization step of optimizing the injection rate of the chemical agent using an estimation model in which the upstream data and the injection rate of the chemical agent are explanatory variables and an evaluation index for evaluating the state of the mixed water after the chemical agent has been injected is the dependent variable, so that the evaluation index estimated from the upstream data and the injection rate of the chemical agent becomes a target value. [Effects of the Invention]

[0014] According to one aspect of the present invention, the accuracy of the injection rate of chemicals injected to treat raw water flowing into a water purification facility can be improved. [Brief explanation of the drawing]

[0015] [Figure 1] This figure shows an overview of a water treatment system according to one embodiment of the present invention. [Figure 2] This block diagram shows an example of the main components of the injection rate calculation system in the water treatment system described above. [Figure 3] This flowchart shows an example of the flow of the injection rate calculation process in the water treatment system with the above configuration. [Figure 4]A graph showing the temporal changes of the raw water turbidity, which is an example of an explanatory variable in the injection rate calculation system, and the injection rate of the flocculant, which is the objective variable, in one embodiment of the water treatment system. [Figure 5] A graph showing the temporal changes of the raw water turbidity, which is an example of an explanatory variable, and the injection rate of the flocculant, which is the objective variable, in a modified example of the water treatment system. [Figure 6] A graph showing the temporal changes of the injection rate of powdered activated carbon, which is an example of an explanatory variable, and the injection rate of the flocculant, which is the objective variable, in another embodiment of the water treatment system. [Figure 7] A diagram showing an overview of a water treatment system according to another embodiment of the present invention. [Figure 8] A block diagram showing an example of the main configuration of an injection rate calculation system in the water treatment system. [Figure 9] A flowchart showing an example of the flow of the injection rate calculation process in the previous stage of the water treatment system having the above configuration. [Figure 10] A flowchart showing an example of the flow of the injection rate calculation process in the middle stage of the water treatment system having the above configuration. [Figure 11] A flowchart showing an example of the flow of the injection rate calculation process in the latter stage of the water treatment system having the above configuration. [Figure 12] A block diagram showing an example of the main configuration of an injection rate calculation system in a water treatment system according to yet another embodiment of the present invention. [Figure 13] A flowchart showing an example of the flow of the injection rate calculation process in the water treatment system having the above configuration.

Embodiments for Carrying Out the Invention

[0016] Hereinafter, embodiments of the present invention will be described in detail. For convenience of explanation, members having the same functions as those shown in each embodiment are denoted by the same reference numerals, and the description thereof will be omitted as appropriate. <000010!> 〔Embodiment 1〕 One embodiment of the present invention will be described with reference to Figures 1 to 3.

[0018] (Water treatment system) Figure 1 is a diagram illustrating the overview of the water treatment system according to this embodiment. The water treatment system 1 of this embodiment is a system that treats raw water flowing into a water purification facility from, for example, a river, into clean water. Specifically, the water treatment system 1 of this embodiment realizes a coagulation and sedimentation process in which suspended solids and other substances contained in the raw water are coagulated by a coagulant and separated from the raw water by gravity sedimentation.

[0019] As shown in Figure 1, the water treatment system 1 includes an intake well 101, a rapid mixing tank 102, a flocculation tank 103, and a sedimentation tank 104 as equipment for realizing the above-mentioned coagulation and sedimentation process.

[0020] The intake well 101 is a tank that suppresses surface disturbances caused by raw water supplied through the water conduit and ensures the hydraulic stability of the raw water supplied to the facilities beyond the intake well 101. Chlorine and pH (hydrogen ion concentration index) adjusters may be injected into the raw water. Here, the chlorine is an agent used to disinfect the raw water. The chlorine is an agent such as sodium hypochlorite, and is injected between the intake well 101 and the rapid mixing tank 102 by an injection device (not shown). The pH adjuster is an agent used to adjust the pH of the raw water. The pH adjuster is an agent such as sulfuric acid or sodium hydroxide (caustic soda), and is injected between the intake well 101 and the rapid mixing tank 102 by an injection device (not shown). In addition, powdered activated carbon may be introduced between the intake point upstream of the intake well 101 and the rapid mixing tank 102. Powdered activated carbon is used for the removal of odor substances, trihalomethanes and trihalomethane precursors, etc.

[0021] A water quality meter 111 is installed in the intake well 101. The water quality meter 111 measures the quality of the raw water that enters the intake well 101. This water quality is at least turbidity, and other examples include color, water temperature, conductivity, pH (hydrogen ion concentration index), alkalinity (acid consumption), ultraviolet absorbance, etc. The water quality meter 111 may be installed in the rapid mixing tank 102, or in the waterway upstream of the rapid mixing tank 102. The water quality data measured by the water quality meter 111 is input into the injection rate calculation system 2.

[0022] A flow meter 112 is installed in the water conduit between the intake well 101 and the rapid mixing tank 102. The flow meter 112 measures the flow rate of raw water sent from the intake well 101 to the rapid mixing tank 102. The flow rate data measured by the flow meter 112 is input to the coagulant injection device 113.

[0023] The rapid mixing tank 102 is a mixing tank for rapidly mixing the raw water sent from the intake well 101 by injecting a coagulant. Here, the coagulant is a chemical agent used to coagulate substances contained in the raw water. The coagulant is a chemical agent such as polyaluminum chloride (PAC) or aluminum sulfate (aluminum sulfate), and is injected into the rapid mixing tank 102 by the coagulant injection device 113.

[0024] Furthermore, the rapid mixing tank 102 is equipped with a rapid agitator 114 for performing the rapid agitation described above. The rapid agitator 114 is, for example, a flash mixer. The rapid agitator 114 may operate at a constant agitation speed, or its agitation speed may be adjustable by motor control. In the rapid mixing tank 102, minute flocs are formed by the injection of the coagulant and the agitation by the rapid agitator 114. The treated water containing these minute flocs is sent to the subsequent floc formation tank 103, where further aggregation of the flocs is promoted in the equipment from the floc formation tank 103 onward.

[0025] The flocculation tank 103 is a tank for slow agitation in order to form larger flocs in the treated water. The flocculation tank 103 is equipped with a slow agitator 115. The agitation of the treated water by the slow agitator 115 promotes further aggregation of flocs. The slow agitator 115 is, for example, a flocculator. After a predetermined time of agitation, the treated water from the flocculation tank 103 is sent to the subsequent sedimentation tank 104. The flocculation tank 103 may be a bypass-flow type without an agitator.

[0026] Furthermore, a camera 116 is provided in the floc formation pond 103. The camera 116 photographs the flocs formed in the floc formation pond 103. From the static images of the flocs, the diameter, area, volume, etc., of the flocs can be estimated, and the formation state of the flocs can be confirmed. Alternatively, the movement speed of the flocs can be estimated from the dynamic images of the flocs, and the formation state of the flocs can be confirmed. The data from the images is input into the injection rate calculation system 2 as an evaluation index for evaluating the aggregation state.

[0027] Furthermore, it is desirable to install camera 116 in an appropriate position depending on the stage of floc formation that you wish to check. If you want to check the floc formation stage as early as possible, installing the camera further upstream within the camera's field of view will shorten the time delay between estimating the flocculation rate and determining the quality of flocculation, thereby minimizing the adverse effects on the water quality of the settled water due to poor flocculation in a short period of time. Conversely, if you want to check the floc formation stage when floc formation has progressed sufficiently, you should install the camera further downstream within the camera's field of view.

[0028] Alternatively, instead of the camera 116, a measuring instrument capable of measuring indicators that can evaluate the quality of the flocculation state, such as zeta potential and flow current values, may be provided. In this case, the water quality data measured by the measuring instrument is input into the injection rate calculation system 2 as an evaluation index for feedback correction of the injection rate.

[0029] The sedimentation tank 104 is a tank for settling and separating aggregated flocs contained in the influent water from the floc formation tank 103. As the influent water remains in the sedimentation tank 104 for a predetermined time (for example, 1 to 3 hours), the larger flocs formed in the floc formation tank 103 settle due to gravity. This separates the flocs from the treated water, and the supernatant liquid is sent to a filtration tank (not shown) as sedimentation water. Further downstream from the sedimentation tank 104, equipment may be provided to perform additional treatments on the treated water sent to the filtration tank, such as ozone treatment or biological activated carbon treatment. The flocs settled in the sedimentation tank 104 are then extracted as sludge and sent to a sludge treatment facility (not shown).

[0030] If the floc formation tank 103 is omitted and the sedimentation tank 104 is changed to a high-speed coagulation sedimentation tank, the floc formation state can be evaluated in the sludge blanket zone where flocs are grown in the high-speed coagulation sedimentation tank.

[0031] The above-mentioned filtration pond is a pond equipped with filtration equipment for filtering the settled water flowing in from the sedimentation pond 104. In the above-mentioned filtration pond, minute suspended solids remaining in the settled water are separated by filtration. After being disinfected with a chlorine agent, the filtered settled water is supplied to consumers as tap water.

[0032] (Injection rate calculation system) As shown in Figure 1, the water treatment system 1 includes an injection rate calculation system 2. The injection rate calculation system 2 calculates the injection rate of coagulant to be injected into the raw water according to the attributes of the raw water, etc. Here, the injection rate is the ratio of the coagulant injected into the raw water to the raw water. Conventionally, the injection rate has been determined using the results of jar tests, past knowledge, predetermined calculation formulas, etc.

[0033] The injection rate calculation system 2 outputs the calculated injection rate to the coagulant injection device 113. Based on the flow rate data from the flow meter 112 and the injection rate from the injection rate calculation system 2, the coagulant injection device 113 determines the amount of coagulant to inject and injects it into the rapid mixing tank 102. The injection rate calculation system 2 may be composed of one information processing device equipped with a set of hardware and software components, or it may be composed of multiple information processing devices.

[0034] Figure 2 is a block diagram showing an example of the main components of the injection rate calculation system 2. The injection rate calculation system 2 includes an estimation device 11, a correction device 12, a model update device 13, an input device 14, an output device 15, and a storage device 16.

[0035] The input device 14 receives user input and outputs an input signal based on that input to the estimation device 11, the correction device 12, or the model update device 13. The input device 14 also outputs input signals from various devices to the estimation device 11 or the correction device 12.

[0036] The output device 15 outputs the information generated by the correction device 12. The output method by the output device 15 is not particularly limited. The output device 15 may be, for example, a display device that displays the information as an image, or an audio output device that outputs the information as sound. Alternatively, the output device 15 may be a control signal output device that outputs the information as an electrical signal to various devices such as the coagulant injection device 113.

[0037] The storage device 16 holds the programs and data used in the injection rate calculation system 2. The storage device 16 includes an estimation model storage unit 31, a target value storage unit 32, and a training data storage unit 33. The data held (stored) by these storage units 31 to 33 will be described later.

[0038] (estimation device) The estimation device 11 uses an estimation model to estimate the injection rate of coagulant to be injected into the raw water. The estimation device 11 includes a control unit 20. The control unit 20 controls all parts of the estimation device 11. The control unit 20 is implemented, for example, by a processor and memory. In this example, the processor accesses storage (not shown), loads a program (not shown) stored in storage into memory, and executes a series of instructions contained in the program. This constitutes the various parts included in the control unit 20.

[0039] The control unit 20 includes an acquisition unit 21 and an injection rate estimation unit 22 (estimation unit). The estimation model storage unit 31 stores the estimation model.

[0040] The acquisition unit 21 acquires input data, which is data input from the input device 14. The input data is upstream data, which is data acquired at at least one of the injection position where the coagulant injection device 113 injects the coagulant and a position upstream of the injection position. The upstream data includes attribute data indicating the attributes of the raw water, the injection rate of chemicals other than the coagulant injected between the intake point upstream of the intake well 101 and the rapid mixing tank 102, and the amount of return water from the distribution reservoir that is mixed with the raw water at the intake well 101.

[0041] Typical examples of the attribute data mentioned above include raw water quality, including raw water turbidity, which is the turbidity of the raw water. Other examples of attribute data mentioned above include the flow rate of the raw water, whether or not there is discharge from a dam located upstream of the water treatment facility, and the mixing ratio of water from multiple raw water sources when there are multiple sources. Examples of chemicals other than the coagulants mentioned above include chlorine agents, acid agents such as sulfuric acid, alkaline agents such as sodium hydroxide, and powdered activated carbon.

[0042] The acquisition unit 21 stores the acquired input data in the teacher data storage unit 33. By storing the input data in the teacher data storage unit 33 at predetermined intervals, the teacher data storage unit 33 stores the collection of input data.

[0043] The injection rate estimation unit 22 estimates the injection rate of the coagulant to be injected into the raw water using the estimation model read from the estimation model storage unit 31. The injection rate estimation unit 22 obtains the coagulant injection rate by inputting the upstream data into the estimation model. The injection rate estimation unit 22 then outputs the obtained injection rate to the correction device 12.

[0044] (Correction device) The correction device 12 corrects the coagulant injection rate estimated by the estimation device 11 based on an evaluation index for the coagulation state in the floc formation pond 103 after the coagulant has been injected. The correction device 12 includes a control unit 40. The control unit 40 controls all parts of the correction device 12. Note that the control unit 40 has the same hardware configuration as the control unit 20, so its explanation is omitted.

[0045] The control unit 40 includes an injection rate correction unit 41 (correction unit). The target value storage unit 32 of the storage device 16 stores a target value for the evaluation index of the aggregation state in the floc formation pond 103. The above target value may be stored in the target value storage unit 32 via the input device 14.

[0046] The injection rate correction unit 41 acquires the evaluation index from the input device 14 and corrects the injection rate according to the difference between the acquired evaluation index and the target value of the evaluation index stored in the target value storage unit 32. The injection rate correction unit 41 then outputs the corrected injection rate as a calculation result to the output device 15. If the output device 15 is the control signal output device described above, the coagulant injection device 113 can inject the coagulant into the raw water based on the injection rate.

[0047] Furthermore, the injection rate correction unit 41 stores the corrected injection rate in the training data storage unit 33. As a result, the training data storage unit 33 stores a set of training data in which the upstream data is used as the explanatory variable and the corrected injection rate is used as the dependent variable.

[0048] (Model update device) The model update device 13 is an estimation model for estimating the injection rate of coagulant to be injected into raw water, and updates the estimation model stored in the estimation model storage unit 31. The model update device 13 is equipped with a control unit 50. The control unit 50 controls all parts of the model update device 13. Note that the control unit 50 has the same hardware configuration as the control unit 40, so its explanation is omitted.

[0049] The control unit 50 includes an update unit 51. The update unit 51 performs machine learning using the training data set read from the training data storage unit 33 and constructs an estimation model as a learning model. In the training data set, the upstream data is the explanatory variable, and the corrected injection rate is the target variable. When the upstream data is input to the estimation model, the corrected injection rate corresponding to the upstream data is output from the estimation model.

[0050] The update unit 51 generates the above-mentioned estimation model as a regression model obtained, for example, by a random forest (RF), gradient boosting (GBR), or neural network (NN). The update unit 51 updates the estimation model stored in the estimation model storage unit 31 using the constructed estimation model.

[0051] According to the above configuration, the coagulant injection rate estimated from upstream data using the estimation model is corrected based on an evaluation index that assesses the state of the mixed water after the coagulant is injected. Then, the estimation model is updated using training data in which the corrected coagulant injection rate is the dependent variable and the upstream data is the independent variable. By using the updated estimation model, the coagulant injection rate can be estimated with high accuracy from the upstream data. As a result, the accuracy of the coagulant injection rate can be improved.

[0052] (Injection rate calculation process) Figure 3 is a flowchart showing an example of the flow of the injection rate calculation process in the water treatment system 1 with the above configuration. First, the acquisition unit 21 acquires upstream data obtained at least one of the coagulant injection position and a position upstream of the injection position, and stores it in the training data storage unit 33 as an explanatory variable (step S11, the description of "step" will be omitted hereafter). Next, the injection rate estimation unit 22 estimates the coagulant injection rate from the upstream data acquired by the acquisition unit 21 using the estimation model stored in the estimation model storage unit 31 (S12).

[0053] Next, the injection rate correction unit 41 acquires an evaluation index for the coagulation state after the coagulant is injected, and corrects the injection rate estimated by the injection rate estimation unit 22 according to the difference between the acquired evaluation index and the target value of the evaluation index stored in the target value storage unit 32 (S13). Next, the injection rate correction unit 41 outputs the corrected injection rate as a calculation result and stores it in the training data storage unit 33 as the target variable (S14). As a result, the pair of the upstream data, which is the explanatory variable, and the corrected injection rate, which is the target variable, is added to the training data group in the training data storage unit 33 as training data.

[0054] Next, the update unit 51 determines whether a predetermined period (e.g., 1 day, 1 week, etc.) has elapsed since the injection rate calculation process started (S15). If the predetermined period has not elapsed (NO in S15), the process returns to step S11.

[0055] On the other hand, if a predetermined period has elapsed (YES in S15), the update unit 51 performs machine learning using the training data set read from the training data storage unit 33, reconstructs the estimation model as a learning model, and updates the estimation model in the estimation model storage unit 31 (S16). After that, the injection rate calculation process is terminated.

[0056] (Additional notes) In this embodiment, the injection rate correction unit 41 outputs the corrected injection rate as the estimated result to the output device 15, but the injection rate estimation unit 22 may also output the estimated injection rate to the output device 15.

[0057] (Example 1) Figure 4 is a graph showing the time evolution of raw water turbidity, an example of the explanatory variable described above, and the coagulant injection rate, the objective variable described above. In Figure 4, the raw water turbidity is shown by a dashed line, and the injection rate in this embodiment is shown by a solid line. In addition, as a comparative example, Figure 4 shows the injection rate by feedforward control (FF control) only, i.e., the injection rate estimated only from the raw water turbidity using an estimation model, as a dashed line. In addition, as a comparative example, Figure 4 shows the injection rate by feedback control (FB control) only, i.e., the injection rate determined only based on the evaluation index and its target value, as a double-dashed line.

[0058] Referring to Figure 4, it can be seen that the injection rate controlled by FF alone has an extremely small time delay from the change in raw water turbidity, but it may differ from the value corresponding to the change in the coagulation state due to the change in raw water turbidity (it may be lower in the example in Figure 4). On the other hand, the injection rate controlled by FB alone corresponds to the value corresponding to the change in the coagulation state due to the change in raw water turbidity, but it can be seen that the time delay T1 from the change in raw water turbidity is large.

[0059] In contrast, the injection rate in this embodiment corresponds to the change in the coagulation state due to the change in raw water turbidity, compared to the injection rate using FF control alone. Furthermore, it can be seen that there is less time lag from the change in raw water turbidity compared to the injection rate using FB control alone.

[0060] (Variation 1) Incidentally, the raw water takes time to flow downstream from the location where the raw water turbidity is acquired in the upstream data (the location of the water quality meter 111 in the intake well 101) to the location where the evaluation index is acquired (the location of the camera 116 in the flocculation pond 103).

[0061] Therefore, in this modified example, the injection rate correction unit 41 stores the injection rate corrected based on the evaluation index at a time elapsed from the time the raw water turbidity, which is an explanatory variable in the above embodiment, is acquired, in the training data storage unit 33 as the target variable of the training data. With the above configuration, the causal relationship between the explanatory variable and the target variable in the training data is appropriately reflected. Accordingly, by using the estimation model updated with this training data, the injection rate of the coagulant can be estimated from the raw water turbidity with even greater accuracy.

[0062] Figure 5 is a graph showing the time change between raw water turbidity, an example of the explanatory variable mentioned above, and the coagulant injection rate, the objective variable. In Figure 5, the raw water turbidity is shown by a dashed line, and the coagulant injection rate of this modified example is shown by a solid line. Also, as with Figure 4, Figure 5 shows the coagulant injection rate with FF control only as a comparative example, shown by a dashed line, and the coagulant injection rate with FB control only as a double dashed line. Referring to Figure 5, it can be seen that the coagulant injection rate of this modified example has an extremely small time lag from the change in raw water turbidity compared to the coagulant injection rate of the example shown in Figure 4, and is about the same as the injection rate with FF control only.

[0063] The above time may also be the hydraulic residence time, i.e., the value obtained by dividing the flow path volume from the raw water turbidity acquisition point to the evaluation index acquisition point by the flow rate. Alternatively, the above time may be the time determined by tracer testing, numerical analysis, etc. Furthermore, the above time may include the response time from the acquisition of the evaluation index until the corrected injection rate is output.

[0064] (Example 2) Figure 6 is a graph showing the time evolution of the injection rate of powdered activated carbon, which is an example of the explanatory variable, and the injection rate of the coagulant, which is the objective variable. In Figure 6, the injection rate of powdered activated carbon is shown by a dashed line, and the injection rate of the coagulant in this embodiment is shown by a solid line. Also in Figure 6, as in Figure 4, the injection rate of the coagulant under FF control only is shown by a dashed line as a comparative example, and the injection rate of the coagulant under FB control only is shown by a double dashed line.

[0065] Generally, when powdered activated carbon is added to the intake well 101, etc., it is necessary to increase the injection rate of the coagulant. Therefore, the injection rate of powdered activated carbon is a variable with a high contribution as an explanatory variable in the above estimation model.

[0066] Referring to Figure 6, it can be seen that the coagulant injection rate under FB control alone corresponds to the change in the coagulation state due to the change in the injection rate of the powdered activated carbon, but the time lag T2 from the change in the injection rate of the powdered activated carbon is large. On the other hand, the coagulant injection rate under FF control alone has an extremely small time lag from the change in the injection rate of the powdered activated carbon, but it can be seen that it may differ from the value corresponding to the change in the coagulation state due to the change in the injection rate of the powdered activated carbon (it may be higher in the example in Figure 6). Therefore, under FF control alone, an excessive amount of coagulant is injected.

[0067] In contrast, it can be seen that the coagulant injection rate in this embodiment has less time delay from the change in the powdered activated carbon injection rate compared to the coagulant injection rate controlled by FB alone. Furthermore, it can be seen that, compared to the coagulant injection rate controlled by FF alone, the coagulant injection rate in this embodiment, after the time delay T2 has elapsed since the change in the powdered activated carbon injection rate, corresponds to the value corresponding to the change in the coagulation state due to the change in the powdered activated carbon injection rate.

[0068] (Modification 2) Incidentally, the raw water takes time to flow down from the injection point of the powdered activated carbon (a predetermined position in the water intake well 101) in the upstream data mentioned above to the acquisition point of the evaluation index (the position of the camera 116 in the floc formation pond 103).

[0069] Therefore, in this modified example, the injection rate correction unit 41 stores the injection rate corrected based on the evaluation index at a time elapsed since the acquisition of the injection rate of the powdered activated carbon, which is an explanatory variable in the above embodiment, in the training data storage unit 33 as the target variable of the training data. With the above configuration, the causal relationship between the explanatory variable and the target variable in the training data is appropriately reflected. Accordingly, by using the estimation model updated with this training data, the injection rate of the coagulant can be estimated with even greater accuracy from the injection rate of the powdered activated carbon.

[0070] The above time may be the hydraulic residence time, as in the modified example shown in Figure 5, or it may be the time determined by tracer testing, numerical analysis, etc. Furthermore, the response time from obtaining the evaluation index to outputting the corrected injection rate may be added to the above time.

[0071] [Embodiment 2] Another embodiment of the present invention will be described with reference to Figures 7 to 11.

[0072] Figure 7 shows an overview of the water treatment system according to this embodiment. The water treatment system 1a of this embodiment disinfects the raw water or treated water obtained by treating the raw water using a chlorine agent. Conventionally, the chlorine requirement is calculated using the concentration of substances that react with chlorine, such as ammonia nitrogen, in the raw water and theoretical formulas based on these reaction equations, and the calculated chlorine requirement is used as a reference when setting the chlorine agent injection rate.

[0073] As shown in Figure 7, the water treatment system 1a includes an intake well 101, a rapid mixing tank 102, a flocculation tank 103, a sedimentation tank 104, an intermediate chlorine mixing tank 105, a filtration tank 106, a downstream chlorine mixing tank 107, and a distribution reservoir 108 as equipment to realize the disinfection process described above. Note that the intake well 101, rapid mixing tank 102, flocculation tank 103, and sedimentation tank 104 shown in Figure 7 are the same as the intake well 101, rapid mixing tank 102, flocculation tank 103, and sedimentation tank 104 shown in Figure 1, respectively.

[0074] A flow meter 121 is installed in the water conduit upstream of the intake well 101. The flow meter 121 measures the flow rate of raw water sent to the intake well 101. The flow rate data measured by the flow meter 121 is input to the upstream chlorine injection device 122. In the intake well 101, the upstream chlorine injection device 122 injects the upstream chlorine. The upstream chlorine disinfects the raw water and oxidizes organic matter and other substances contained in the raw water.

[0075] A concentration meter 123 is installed upstream of the rapid mixing tank 102. The concentration meter 123 measures the residual chlorine concentration of the mixed water in the rapid mixing tank 102. The residual chlorine concentration measurement data measured by the concentration meter 123 is input into the injection rate calculation system 2a.

[0076] A flow meter 124 is installed in the water conduit between the sedimentation tank 104 and the intermediate chlorine mixing tank 105. The flow meter 124 measures the flow rate of treated water sent from the sedimentation tank 104 to the intermediate chlorine mixing tank 105. The flow rate data measured by the flow meter 124 is input to the intermediate chlorine injection device 125.

[0077] The intermediate chlorine mixing tank 105 is a mixing tank for injecting the intermediate chlorine agent into the sedimentation water sent from the sedimentation tank 104. The intermediate chlorine agent is injected into the intermediate chlorine mixing tank 105 by the intermediate chlorine agent injection device 125. The intermediate chlorine agent disinfects the raw water and removes dissolved manganese contained in the raw water in the subsequent filtration tank 106.

[0078] The filtration tank 106 is a tank equipped with filtration equipment that filters the settled water flowing in from the sedimentation tank 104 via the intermediate chlorine mixing tank 105. In the filtration tank 106, minute suspended solids remaining in the settled water are separated by filtration. The water filtered in the filtration tank 106 is sent to the subsequent chlorine mixing tank 107 as purified water.

[0079] A concentration meter 126 is installed upstream of the filtration tank 106. The concentration meter 126 measures the residual chlorine concentration of the sediment sent to the filtration tank 106. The residual chlorine concentration measurement data measured by the concentration meter 126 is input into the injection rate calculation system 2a.

[0080] A flow meter 127 is installed in the water conduit between the filtration tank 106 and the subsequent chlorine mixing tank 107. The flow meter 127 measures the flow rate of purified water sent from the filtration tank 106 to the subsequent chlorine mixing tank 107. The flow rate data measured by the flow meter 127 is input to the subsequent chlorine injection device 128.

[0081] The subsequent chlorine mixing tank 107 is a mixing tank for injecting the subsequent chlorine agent into the purified water sent from the filtration tank 106. The subsequent chlorine agent is injected into the subsequent chlorine mixing tank 107 by the subsequent chlorine agent injection device 128. The subsequent chlorine agent is used to adjust the residual chlorine concentration of the water distributed from the distribution reservoir 108.

[0082] The water distribution reservoir 108 is a reservoir equipped with control equipment that temporarily stores purified water flowing in from the filtration pond 106 via the downstream chlorine mixing pond 107 and controls the outflow according to the demand. Purified water is supplied from the water distribution reservoir 108 to houses and other buildings via water pipes.

[0083] A concentration meter 129 is installed upstream of the water reservoir 108. The concentration meter 129 measures the residual chlorine concentration of the purified water sent to the water reservoir 108. The residual chlorine concentration measurement data measured by the concentration meter 129 is input into the injection rate calculation system 2a.

[0084] (Injection rate estimation system) As shown in Figure 7, the water treatment system 1a includes an injection rate calculation system 2a. The injection rate calculation system 2a estimates the injection rates of the chlorine agent in the pre-, middle, and post-stages to be injected into the water, according to the attributes of the raw water, etc. Here, the injection rate is the ratio of the chlorine agent to be injected into the water to the water in question. Conventionally, the injection rate has been determined using the results of jar tests, past knowledge, predetermined calculation formulas, etc.

[0085] The injection rate calculation system 2a outputs the estimated injection rate of the preceding chlorine agent (hereinafter sometimes referred to as the "preceding injection rate") to the preceding chlorine agent injection device 122. The preceding chlorine agent injection device 122 determines the amount of preceding chlorine agent to be injected into the intake well 101 based on the flow rate data from the flow meter 121 and the preceding injection rate from the injection rate calculation system 2a.

[0086] Furthermore, the injection rate calculation system 2a outputs the estimated injection rate of chlorine in the intermediate stage (hereinafter sometimes referred to as the "intermediate stage injection rate") to the intermediate stage chlorine injection device 125. Based on the flow rate data from the flow meter 124 and the intermediate stage injection rate from the injection rate calculation system 2a, the intermediate stage chlorine injection device 125 determines the amount of chlorine to be injected into the intermediate stage and injects it into the intermediate stage chlorine mixing tank 105.

[0087] Furthermore, the injection rate calculation system 2a outputs the estimated injection rate of downstream chlorine (hereinafter sometimes referred to as the "downstream injection rate") to the downstream chlorine injection device 128. Based on the flow rate data from the flow meter 127 and the downstream injection rate from the injection rate calculation system 2a, the downstream chlorine injection device 128 determines the amount of downstream chlorine to be injected and injects it into the downstream chlorine mixing tank 107. The injection rate calculation system 2a may be composed of a single information processing device equipped with a set of hardware and software components, or it may be composed of multiple information processing devices.

[0088] Figure 8 is a block diagram showing an example of the main components of the injection rate calculation system 2a. The injection rate calculation system 2a of this embodiment differs from the injection rate calculation system 2 shown in Figure 2 in that the data stored in the storage device 16 is different, but the other components are the same.

[0089] The estimation model storage unit 31 of the storage device 16 stores an estimation model for the preceding stage to estimate the injection rate of the chlorine agent, an estimation model for the intermediate stage to estimate the injection rate of the chlorine agent, and an estimation model for the final stage to estimate the injection rate of the chlorine agent. The target value storage unit 32 of the storage device 16 stores the target value of the residual chlorine concentration for the preceding stage, the target value of the residual chlorine concentration for the intermediate stage, and the target value of the residual chlorine concentration for the final stage as target values ​​for the evaluation index.

[0090] The training data storage unit 33 of the storage device 16 stores the training data for the preceding stage to construct the preceding estimation model, the training data for the middle stage to construct the preceding estimation model, and the training data for the latter stage to construct the preceding estimation model.

[0091] The training data described above includes the upstream data mentioned above as explanatory variables. The training data described above includes the corrected injection rate from the previous stage as the dependent variable. The upstream data described above is data acquired upstream of the chlorine injection location described above. The upstream data described above includes attribute data indicating the attributes of the raw water.

[0092] The training data in the middle section above includes the upstream data mentioned above, the injection rate of the chlorine agent at the preceding stage (first position), and the residual chlorine concentration after injection as explanatory variables. The training data in the middle section above includes the corrected injection rate at the middle section (second position downstream of the first position) as the dependent variable.

[0093] The training data for the latter stage described above includes, as explanatory variables, the upstream data described above, the injection rate of chlorine agent at the preceding stage (first position) and the residual chlorine concentration after injection, and the injection rate of chlorine agent at the intermediate stage (first position) and the residual chlorine concentration after injection. The training data for the latter stage described above includes, as the dependent variable, the corrected injection rate at the later stage (second position downstream of the first position).

[0094] According to the above configuration, the chlorine injection rate estimated from upstream data using the estimation model is corrected based on the residual chlorine concentration, which evaluates the state of the mixed water after the chlorine injection. Then, the estimation model is updated using training data in which the corrected chlorine injection rate is the dependent variable and the upstream data is the independent variable. By using the updated estimation model, the chlorine injection rate can be estimated with high accuracy from the upstream data. As a result, the accuracy of the chlorine injection rate can be improved.

[0095] (Previous step: Injection rate calculation process) Figure 9 is a flowchart showing an example of the flow of the upstream injection rate calculation process in the water treatment system 1a with the above configuration. First, the acquisition unit 21 acquires upstream data obtained at least one of the injection position of the chlorine agent in the upstream stage and a position upstream of the injection position, and stores it in the training data storage unit 33 as an explanatory variable (S21). Next, the injection rate estimation unit 22 estimates the injection rate of the chlorine agent in the upstream stage from the upstream data acquired by the acquisition unit 21, using the estimation model of the upstream stage stored in the estimation model storage unit 31 (S22).

[0096] Next, the injection rate correction unit 41 acquires the residual chlorine concentration after the injection of the preceding chlorine agent, that is, the preceding residual chlorine concentration measured by the concentration meter 123 of the rapid mixing tank 102, and corrects the injection rate of the preceding chlorine agent estimated by the injection rate estimation unit 22 according to the difference between the acquired preceding residual chlorine concentration and the target value of the preceding residual chlorine concentration stored in the target value storage unit 32 (S23). Next, the injection rate correction unit 41 outputs the corrected injection rate of the preceding chlorine agent as an estimation result and stores it in the training data storage unit 33 as the target variable of the preceding training data (S24). As a result, the pair of the upstream data, which is the explanatory variable, and the corrected preceding injection rate, which is the target variable, is added to the preceding training data group in the training data storage unit 33 as the preceding training data.

[0097] Next, the update unit 51 determines whether a predetermined period (e.g., 1 day, 1 week, etc.) has elapsed since the start of the injection rate calculation process in the previous stage (S25). If the predetermined period has not elapsed (NO in S25), the process returns to step S21.

[0098] On the other hand, if a predetermined period has elapsed (YES in S25), the update unit 51 performs machine learning using the previous stage's training data set read from the training data storage unit 33 and reconstructs the previous stage's estimation model as a learning model (S26). The update unit 51 updates the previous stage's estimation model stored in the estimation model storage unit 31 with the reconstructed previous stage's estimation model. After that, the injection rate calculation process for the previous stage described above is terminated.

[0099] (Injection rate calculation process in the middle section) Figure 10 is a flowchart showing an example of the flow of the process for calculating the injection rate of the intermediate stage in the water treatment system 1a with the above configuration. First, the acquisition unit 21 acquires the upstream data, the injection rate of the chlorine agent in the preceding stage, and the residual chlorine concentration after injection, and stores them as explanatory variables in the training data storage unit 33 (S31). Next, the injection rate estimation unit 22 estimates the injection rate of the chlorine agent in the intermediate stage using the intermediate stage estimation model stored in the estimation model storage unit 31, based on the upstream data acquired by the acquisition unit 21, the injection rate of the chlorine agent in the preceding stage, and the residual chlorine concentration after injection (S32).

[0100] Next, the injection rate correction unit 41 acquires the residual chlorine concentration after the injection of the intermediate chlorine agent, that is, the intermediate residual chlorine concentration measured by the concentration meter 126 of the filter pond 106, and corrects the injection rate of the intermediate chlorine agent estimated by the injection rate estimation unit 22 according to the difference between the acquired intermediate residual chlorine concentration and the target value of the intermediate residual chlorine concentration stored in the target value storage unit 32 (S33). Next, the injection rate correction unit 41 outputs the corrected injection rate of the intermediate chlorine agent as an estimation result and stores it in the training data storage unit 33 as the target variable of the intermediate training data (S34). As a result, the pair of the explanatory variables, namely the upstream data and the injection rate and residual chlorine concentration of the preceding chlorine agent, and the target variable, namely the corrected injection rate of the intermediate chlorine agent, is added to the intermediate training data group in the training data storage unit 33 as intermediate training data.

[0101] Next, the update unit 51 determines whether a predetermined period has elapsed since the start of the intermediate injection rate calculation process (S35). If the predetermined period has not elapsed (NO in S35), the process returns to step S31. On the other hand, if the predetermined period has elapsed (YES in S35), the update unit 51 performs machine learning using the intermediate training data set read from the training data storage unit 33 and reconstructs the intermediate estimation model as a learning model (S36). The update unit 51 updates the intermediate estimation model stored in the estimation model storage unit 31 with the reconstructed intermediate estimation model. After that, the injection rate calculation process described above is terminated.

[0102] (Subsequent injection rate calculation process) Figure 11 is a flowchart showing an example of the flow of the subsequent injection rate calculation process in the water treatment system 1a with the above configuration. First, the acquisition unit 21 acquires the upstream data, the injection rate of the chlorine agent in the preceding stage and the residual chlorine concentration after injection, and the injection rate of the chlorine agent in the middle stage and the residual chlorine concentration after injection, and stores them as explanatory variables in the training data storage unit 33 (S41). Next, the injection rate estimation unit 22 estimates the injection rate of the chlorine agent in the subsequent stage using the estimation model stored in the estimation model storage unit 31, based on the upstream data acquired by the acquisition unit 21, the injection rate of the chlorine agent in the preceding stage and the residual chlorine concentration after injection, and the injection rate of the chlorine agent in the middle stage and the residual chlorine concentration after injection (S42).

[0103] Next, the injection rate correction unit 41 acquires the residual chlorine concentration after the injection of the downstream chlorine agent, that is, the downstream residual chlorine concentration measured by the concentration meter 129 of the water reservoir 108, and corrects the injection rate of the downstream chlorine agent estimated by the injection rate estimation unit 22 according to the difference between the acquired downstream residual chlorine concentration and the target value of the downstream residual chlorine concentration stored in the target value storage unit 32 (S43). Next, the injection rate correction unit 41 outputs the corrected injection rate of the downstream chlorine agent as an estimation result and stores it in the training data storage unit 33 as the target variable of the downstream training data (S44). As a result, the set of the explanatory variables, namely the upstream data, the injection rate of the preceding chlorine agent and the residual chlorine concentration after injection, and the injection rate of the intermediate chlorine agent and the residual chlorine concentration after injection, and the target variable, namely the corrected injection rate of the downstream chlorine agent, is added to the downstream training data group in the training data storage unit 33 as training data for the downstream.

[0104] Next, the update unit 51 determines whether a predetermined period has elapsed since the start of the subsequent injection rate calculation process (S45). If the predetermined period has not elapsed (NO in S45), the process returns to step S41. On the other hand, if the predetermined period has elapsed (YES in S45), the update unit 51 performs machine learning using the subsequent training data set read from the training data storage unit 33 and reconstructs the subsequent estimation model as a learning model (S46). The update unit 51 updates the subsequent estimation model stored in the estimation model storage unit 31 with the reconstructed subsequent estimation model. After that, the preceding injection rate calculation process is terminated.

[0105] (Additional notes) In this embodiment, the preceding chlorine agent, the middle chlorine agent, and the subsequent chlorine agent are used in the water treatment system 1a, but the system is not limited to this. Depending on the water purification facility, any two of the preceding chlorine agent, the middle chlorine agent, and the subsequent chlorine agent may be used, or any one of them may be used. The configuration and treatment of the chlorine agents that are not used among the preceding chlorine agent, the middle chlorine agent, and the subsequent chlorine agent can be omitted. Furthermore, in this embodiment, residual chlorine concentration is used as an evaluation index, but the system is not limited to this, and other indicators such as electrical conductivity can be used as evaluation indicators.

[0106] Furthermore, as described above, the injection rate calculation system 2a of this embodiment has the same configuration as the injection rate calculation system 2 shown in Figure 2. Therefore, by incorporating the injection rate calculation system 2a of this embodiment into the injection rate calculation system 2 shown in Figure 2, the water treatment system 1a of this embodiment can be easily incorporated into the water treatment system 1 shown in Figures 1 to 3.

[0107] [Embodiment 3] Another embodiment of the present invention will be described with reference to Figures 12 and 13. The water treatment system 1 of this embodiment differs from the water treatment system 1 shown in Figures 1 to 3 in that an injection rate calculation system 3 is provided instead of the injection rate calculation system 2, but the other configurations are the same.

[0108] Figure 12 is a block diagram showing an example of the main components of the injection rate calculation system 3 of this embodiment. The injection rate calculation system 3 of this embodiment differs from the injection rate calculation system 2 shown in Figure 2 in that an optimization device 17 is provided instead of the estimation device 11 and correction device 12, a model construction device 18 is provided instead of the model update device 13, and in the storage device 16, an estimation model storage unit 34 and a training data storage unit 35 are provided instead of the estimation model storage unit 31 and training data storage unit 33. Otherwise, it is the same.

[0109] The training data storage unit 35 pre-stores a set of training data, which is a collection of training data where the upstream data and the injection rate of the coagulant are explanatory variables, and the evaluation index of the coagulation state is the objective variable.

[0110] The model construction device 18 constructs an estimation model for estimating an evaluation index for the flocculation state in the floc formation pond 103 after the injection of a flocculant. The model construction device 18 stores the constructed estimation model in the estimation model storage unit 31. The model construction device 18 is equipped with a control unit 70. The control unit 70 controls all parts of the model construction device 18. Note that the control unit 70 has the same hardware configuration as the control unit 50 shown in Figure 2, so its explanation is omitted.

[0111] The control unit 70 includes a construction unit 71. The construction unit 71 performs machine learning using the training data set read from the training data storage unit 35 and constructs an estimation model as a learning model. When the upstream data and the injection rate are input to the estimation model, the evaluation index corresponding to the upstream data and the injection rate is output from the estimation model. The construction unit 71 generates the regression model as the estimation model. The construction unit 71 stores the constructed estimation model in the estimation model storage unit 34.

[0112] The estimation model storage unit 34 stores an estimation model for estimating an evaluation index for the coagulation state from the upstream data and the injection rate of the coagulant.

[0113] The optimization device 17 optimizes the injection rate of the coagulant so that the evaluation index estimated from the upstream data and the injection rate of the coagulant using the estimation model described above reaches the target value. The optimization device 17 is equipped with a control unit 60. The control unit 60 controls all parts of the optimization device 17. Note that the control unit 60 has the same hardware configuration as the control unit 20 and control unit 40 shown in Figure 2, so its explanation is omitted.

[0114] The control unit 60 includes an acquisition unit 61 and an injection rate optimization unit 62 (optimization unit). The acquisition unit 61 acquires the upstream data from the input device 14. The acquisition unit 61 stores the acquired upstream data in the training data storage unit 35. By storing the upstream data in the training data storage unit 35 at predetermined intervals, the training data storage unit 35 stores a collection of the upstream data.

[0115] The injection rate optimization unit 62 uses the upstream data acquired by the acquisition unit 61, the estimated model read from the estimated model storage unit 34, and the target value of the evaluation index read from the target value storage unit 32 to optimize the injection rate of the coagulant so that the evaluation index estimated using the estimated model from the upstream data and the injection rate of the coagulant becomes the target value. The injection rate optimization unit 62 outputs the optimized injection rate to the output device 15. If the output device 15 is the control signal output device described above, the coagulant injection device 113 can inject the coagulant into the raw water based on the injection rate.

[0116] According to the above configuration, the injection rate of the coagulant is optimized so that the evaluation index for evaluating the state of the mixed water after the coagulant is injected, which is estimated from the upstream data and the injection rate of the coagulant using an estimation model, reaches a target value. Therefore, the accuracy of the injection rate of the coagulant can be improved using the upstream data, the estimation model, and the target value of the evaluation index.

[0117] (Injection rate calculation process) Figure 13 is a flowchart showing an example of the flow of the injection rate calculation process in the water treatment system 1 with the above configuration. Before executing the injection rate calculation process, the training data storage unit 35 stores the training data set, and the model construction device 18 constructs an estimation model using the training data and stores it in the estimation model storage unit 34.

[0118] First, the acquisition unit 61 acquires upstream data obtained at least one of the injection position of the coagulant and a position upstream of the injection position (S51). Next, the injection rate optimization unit 62 optimizes the injection rate of the coagulant so that the evaluation index estimated using the estimation model in the estimation model storage unit 34 from the upstream data and the injection rate of the coagulant becomes the target value in the target value storage unit 32 (S52). Next, the injection rate optimization unit 62 outputs the optimized injection rate as the calculation result (S53). After that, the injection rate calculation process is terminated.

[0119] (Additional notes) The injection rate optimization unit 62 may add the above-mentioned upstream data, the optimized injection rate, and the evaluation index estimated using the estimation model from the above-mentioned upstream data and the optimized injection rate to the training data group in the training data storage unit 35 as new training data. The model building device 18 may then update the estimation model stored in the estimation model storage unit 34 using the training data group to which the new training data has been added.

[0120] Furthermore, although the water treatment system 1 of the above embodiment is used for controlling the injection of a coagulant or chlorine agent, it is not limited to this and can be used for controlling the injection of any chemical agent such as a pH adjuster or powdered activated carbon that is injected into the water treatment system.

[0121] [Examples of implementation using software] The functions of the estimation device 11, the correction device 12, the model update device 13, the optimization device 17, and the model construction device 18 (hereinafter referred to as "devices") can be realized by programs that cause a computer to function as the device, and by programs that cause a computer to function as each control block of the device (particularly each part included in the control units 20, 40, 50, 60, and 70).

[0122] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing the program. By executing the program using this control device and storage device, the functions described in each of the embodiments are realized.

[0123] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0124] Furthermore, some or all of the functions of each of the above control blocks can also be realized by logic circuits. For example, an integrated circuit in which logic circuits functioning as each of the above control blocks are formed is also included in the scope of the present invention. In addition, it is also possible to realize the functions of each of the above control blocks by, for example, a quantum computer.

[0125] Furthermore, each process described in the above embodiments may be performed by AI (Artificial Intelligence). In this case, the AI ​​may operate on the control device described above, or it may operate on other devices (for example, an edge computer or a cloud server).

[0126] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of Symbols]

[0127] 1.1a Water treatment system 2, 2a, 3 Injection Rate Calculation System 11 Estimation device 12 Correction device 13 Model update device 14 Input devices 15 Output device 16 Storage device 17 Optimization device 18 Model building device 20, 40, 50, 60, 70 Control Unit 21, 61 Acquisition Department 22 Injection rate estimation section (estimation section) 34 Estimated Model Memory Unit 32 Target value storage unit 33, 35 Training data storage unit 41 Injection rate correction section (correction section) 51 Update section 62. Injection Rate Optimization Unit (Optimization Unit) 71 Construction Department 101 Water landing well 102 Rapid mixing tank 103 Flocculation pond 104 Sedimentation pond 105, 107 Chlorine mixing tank 106 Filtration pond 108 Reservoir 111 Water quality meter 112, 121, 124, 127 flowmeter 113 Coagulant injection device 114 Rapid stirrer 115 Slow-speed agitator 116 Cameras 122, 125, 128 Chlorine injection device 123, 126, 129 Densitometer

Claims

1. An injection rate calculation system for calculating the injection rate of chemicals to be injected to treat raw water flowing into a water purification facility, An acquisition unit that acquires upstream data measured at at least one of the injection position of the agent and a position upstream of the injection position, and an evaluation index for evaluating the state of the mixed water after the agent has been injected. An estimation unit that estimates the drug injection rate from the upstream data using an estimation model, A correction unit corrects the injection rate estimated by the estimation unit based on the evaluation index, An injection rate calculation system comprising: an update unit that updates the estimation model using training data in which the upstream data is used as explanatory variables and the injection rate corrected by the correction unit is used as the objective variable.

2. An injection rate calculation system for calculating the injection rate of chemicals to be injected to treat raw water flowing into a water purification facility, An acquisition unit that acquires upstream data measured at at least one of the injection position of the drug and a position upstream of the injection position, An injection rate calculation system comprising: an optimization unit that optimizes the injection rate of the drug so that the evaluation index estimated from the upstream data and the injection rate of the drug becomes a target value, using an estimation model constructed with training data in which the upstream data and the injection rate of the drug are explanatory variables and an evaluation index for evaluating the state of the mixed water after the drug has been injected is the dependent variable.

3. The upstream data includes raw water turbidity, which is the turbidity of the raw water flowing into the water treatment facility. The injection rate calculation system according to claim 1 or 2, wherein the evaluation index is the evaluation index at the time elapsed from the raw water turbidity acquisition point to the mixed water state acquisition point, where the time spent flowing from the raw water turbidity acquisition point to the mixed water state acquisition point is the time elapsed from the raw water turbidity acquisition point.

4. The injection rate calculation system according to claim 1 or 2, wherein the upstream data includes at least one of attribute data indicating the attributes of the raw water, the injection rate of a different agent injected at a position upstream of the injection position of the agent, and the amount of water that has been treated with the raw water and is mixed with the raw water.

5. The aforementioned agent is a flocculant, The injection rate calculation system according to claim 1 or 2, wherein the evaluation index is obtained from a captured image of the floc.

6. The aforementioned drug is a chlorine agent, The injection rate calculation system according to claim 1 or 2, wherein the evaluation index is residual chlorine concentration.

7. The chlorine agent is injected at a first position and at a second position downstream of the first position. The injection rate calculation system according to claim 6, wherein the estimation model, which includes the injection rate of the chlorine agent injected at the second position as an explanatory variable or dependent variable, includes the injection rate of the chlorine agent injected at the first position and the residual chlorine concentration after injection as explanatory variables.

8. An injection rate calculation system according to claim 1 or 2, A water treatment system comprising: a control device that controls the injection of the chemical agent according to the injection rate calculated by the injection rate calculation system.

9. An injection rate calculation program for causing a computer to function as an injection rate calculation system according to claim 1, wherein the computer functions as the estimation unit, the correction unit, and the update unit.

10. An injection rate calculation program for causing a computer to function as the injection rate calculation system described in claim 2, wherein the computer functions as the optimization unit.

11. An acquisition step of obtaining upstream data measured at at least one of the injection position of a chemical agent injected to treat raw water flowing into a water treatment plant and a position upstream of the injection position, and an evaluation index for evaluating the state of the mixed water after the chemical agent has been injected, An estimation step in which the infusion rate of the drug is estimated from the upstream data using an estimation model, A correction step in which the injection rate estimated in the estimation step is corrected based on the evaluation index, An injection rate calculation method, comprising: an update step, which updates the estimation model using training data in which the upstream data is used as explanatory variables and the injection rate corrected in the correction step is used as the objective variable.

12. An acquisition step of acquiring upstream data measured at at least one of the injection position of a chemical agent injected to treat raw water flowing into a water treatment plant and a position upstream of the injection position, An injection rate calculation method, comprising: an optimization step of optimizing the injection rate of the drug, using an estimation model in which the upstream data and the injection rate of the drug are explanatory variables and an evaluation index for evaluating the state of the mixed water after the drug has been injected is the dependent variable, so that the evaluation index estimated from the upstream data and the injection rate of the drug becomes a target value.

Images