JP2020025943A - Water treatment method and water treatment system - Google Patents

Abstract

To provide a water treatment method that enables a flocculated flock state in water treatment to be determined in real time during the water treatment.SOLUTION: A water treatment method forms a flocculated flock by injecting a flocculating agent into raw water. Image data on the flocculated flock are generated by an image acquisition device. At least the image data on the flocculated flock are input into a model that is constructed on the basis of a machine learning algorithm, and the result of determination indicating a state of the flocculated flock is output from the model.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a water treatment method and a water treatment system for determining a state of flocculated flocs from a model using a machine learning algorithm and a still image or a continuous image of flocculated flocs.

  In recent years, with the dramatic improvement in computer computing capacity, the improvement of communication infrastructure environments, and the evolution of analysis algorithms, the third artificial intelligence boom has emerged. The improvement of the communication infrastructure environment has made it possible to easily acquire a large amount of data, and unlike the artificial intelligence boom so far, expectations for its use in the real world are increasing.

  Machine learning / artificial intelligence is expected to be applied to various fields, and in the field of water treatment, a search for a utilization method is being made. One of the utilization methods of machine learning / artificial intelligence in the field of water treatment is to predict an objective variable in an unknown environment, for example, a temporal future. One of the prediction methods of the objective variable in the unknown environment is an analytical approach. This is to analyze a physical phenomenon theoretically and to formulate it. On the other hand, prediction of objective variables in unknown environments using machine learning / artificial intelligence is based on statistical analysis of the correlation between inputs and outputs without analyzing the theoretical relationship between the inputs and outputs in accordance with natural laws. In this case, a regression equation is created as an example of a relational equation.

  In the water treatment field, it is sometimes difficult to theoretically analyze the phenomena that are occurring, and machine learning / artificial intelligence that can predict the output from the input without elucidating the theoretical background is effective. It is expected that there are many aspects that can be used.

  As mentioned above, the artificial intelligence boom itself has reached its third boom this time. The prominent algorithm of the second artificial intelligence boom in the late 1980s was neural networks. During the first artificial intelligence boom, the addition of a hidden layer to the perceptron, which consisted of only the input and output layers, and the establishment of a learning method called the "back propagation method," Artificial intelligence boom. Eventually, the second artificial intelligence boom will be abolished, but it is said that one of the factors was the lack of satisfactory prediction accuracy depending on the task. In the third artificial intelligence boom, deep learning (deep learning), in which the hidden layers of the neural network are multi-layered, plays a leading role. Due to advances in learning algorithms, high accuracy can be achieved depending on the tasks to be applied. This is part of the boom. In this specification, machine learning using a neural network including an input layer, two or more hidden layers, and an output layer is referred to as deep learning. Artificial intelligence is a broader concept than machine learning, and there is machine learning as one of the means for creating artificial intelligence.

  Now, returning to the field of water treatment, in water treatment, aluminum or iron is used to efficiently remove suspended components such as turbidity in raw water and soluble components such as biopolymers and humic acid. In many cases, an aggregating treatment is performed in which a suspended component or a soluble component is agglomerated using an aggregating agent containing a polyvalent cation. The lump of dirt formed by agglomeration is called agglomerated floc. This flocculated floc is settled in a sedimentation tank provided at the subsequent stage of the flocculation tank, or is directly subjected to sand filtration or membrane filtration without sedimentation.

  At this time, the state of the flocculated floc has a great influence on the processing performance of each processing. For example, in the sedimentation treatment, the size and density of the flocculated flocs affect the sedimentation speed of flocculated flocs. Further, in the filtration treatment, minute components which are not sufficiently incorporated into the flocculated flocs due to insufficient formation of the flocculated flocs, such as picoplankton and soluble chromaticity components, leak into the filtered water. In addition, incomplete flocculation may block the filter sand and the pores of the membrane, increasing the rate of increase in filtration resistance and causing serious membrane fouling. Therefore, monitoring the state of flocculated flocs is extremely important in water treatment.

JP-A-3-175339 JP 2015-192960 A

  It is difficult to judge the state of the flocculated flocs in real time from the water quality. In other words, coagulation floc is formed after the coagulant is injected, and the treated water is finally obtained through precipitation and filtration.The coagulation operation was appropriate until the quality of the treated water was measured. It is difficult to determine whether it was good.

  Sometimes the raw water quality can fluctuate significantly in a short period of time due to typhoons or guerrilla downpours. For example, there are cases where the turbidity rises rapidly from 5 degrees to 1,000 degrees in 10 minutes. In such a situation, even if the quality of the treated water is checked as described above, and even if the injection rate of the flocculant is reviewed, it usually takes several tens of minutes from the injection of the flocculant to the obtaining of the treated water. Since it takes several minutes to several hours, it is not enough in time.

  Therefore, if raw water fluctuations actually occur at a water purification plant, etc., experienced operators determine the coagulant injection rate based on the raw water quality based on past knowledge, and visually check the floc condition of the coagulation tank. However, it is determined whether the determined injection rate is appropriate or not, and if necessary, the coagulant injection rate is reviewed again. Alternatively, the raw water is actually collected, dispensed in equal amounts into a plurality of beakers, and several kinds of flocculant injection rates are set to perform a flocculation test, a so-called jar test, to determine an appropriate flocculant injection rate for the current raw water. I have decided.

  However, the former case largely depends on the experience and skills of the operators, and is not a wise method in the water treatment industry, which has the problem of technology transfer. In the latter case, although unevenness by the operator is reduced, it takes much time and it takes about 15 minutes until the result of the jar test is obtained.

Therefore, a technique for grasping the state of the flocculated flocs in real time and performing the flocculation treatment more appropriately is required.
Patent Literature 1 describes a floc monitoring device that captures an image of a floc under water, performs image processing, quantifies the particle size of each floc, and obtains a particle size distribution. There is also a description about performing operation control using the same device.

  Patent Document 2 discloses that a floc particle size is measured in a floc forming tank or in a mixing tank and a floc forming tank by using a floc particle size measuring device at a plurality of locations in a floc forming process, and the difference in the obtained floc particle size is measured. Or, there is a description of a water treatment system that detects abnormal floc formation from the ratio.

  In each case, the state of aggregation is determined based on the particle size of the aggregated floc. However, in reality, it is difficult to determine the quality of the aggregated floc only by the particle size of the floc. For example, when the flocculant injection rate is set to 10, 20, 30, 40, 50, and 60 mg / L for raw water for which the optimal injection rate of the flocculant is determined to be 40 mg / L, and each floc particle size is measured. If the coagulant injection rate exceeds 40 mg / L, the coagulant becomes excessive, charge neutralization does not occur, floc formation failure occurs, and the floc particle size decreases. Then, the floc particle diameter at the coagulant injection rate of 60 mg / L may be substantially the same as the floc particle diameter at the coagulant injection rate of 30 mg / L. In such a case, if the control based on the floc particle size is performed, the review of the coagulant injection rate may be erroneously performed.

  The present invention has been made in view of the above points, and an object of the present invention is to learn the state of agglomerated flocs by using machine learning algorithms such as deep learning, including various elements including the particle diameter, and to operate a veteran operation. To provide a water treatment method capable of determining the flocculation floc state in water treatment in real time during water treatment by constructing a model capable of determining flocculation floc equal to or greater than the number of members is there. Another object of the present invention is to provide a water treatment system that can execute such a water treatment method.

  In one embodiment, a flocculant is injected into raw water to form flocculated flocs, image data of the flocculated flocs is generated by an image acquisition device, and at least the flocculated floc image data is converted into a model constructed by a machine learning algorithm. There is provided a water treatment method, wherein the water treatment method is inputted and a determination result indicating the state of the flocculated floc is output from the model.

  As a machine learning algorithm, a deep learning method (deep learning method) is preferable. The deep learning method is a learning method based on a neural network in which hidden layers are multilayered. In this specification, machine learning using a neural network including an input layer, two or more hidden layers, and an output layer is referred to as deep learning. By using the deep learning method, the state of the flocculated floc, which has been determined to be good or bad based on human eyes and experience, can be determined by a computer based on the flocculated floc image data.

  For example, the floc particle diameter when the coagulant injection rate is 60 mg / L may be substantially the same as the floc particle diameter when the coagulant injection rate is 30 mg / L. However, comparing the actual state of floc at 30 mg / L and 60 mg / L, there are small differences in floc density, so-called tightness, color, shape, and the like. According to machine learning / deep learning, it is sometimes possible to discriminate an image with a precision higher than that of a human by repeatedly making a computer repeatedly learn a combination of image data and correct data corresponding to the image data.

  In the present specification, the water treatment refers to water purification treatment, water treatment, and wastewater treatment, and particularly, to water, polyaluminum chloride (PACl), sulfate, ferric chloride, polyiron, and polysilica iron (PSI). ), Water treatment involving a coagulation operation of injecting an inorganic coagulant such as an organic coagulant or a polymer or an organic coagulant.

  The current flocculated flocs in the water treatment system are imaged by the image acquisition device during the water treatment. The image of the flocculated flocs may be either a still image or a continuous image.

  In one aspect, the step of generating the image data of the flocculated floc by an image acquisition device includes extracting water containing the flocculated floc and introducing the water into a sampling vessel, and converting the image data of the flocculated floc in the sampling vessel into an image. This is a step of generating by the acquisition device.

  In the construction of a model using the deep learning method among the machine learning algorithms described below, since the image is characterized by a computer instead of a person, information other than the agglomerated flocs, such as the brightness of the background and the image, is also included in the model. May affect construction. Therefore, when an image acquisition device, for example, a camera is installed in an aquarium, the structure of the aquarium may be reflected in an image or a difference in brightness due to weather may occur. Therefore, as described above, water containing coagulated flocs is extracted from the water treatment area, introduced into a sampling container having a constant background and brightness, and the coagulated flocs in the sampling container are photographed for deep learning. Suitable image data can be obtained.

  The image data of the flocculated floc obtained by the above means is input to a model constructed by a machine learning algorithm, and the state of the current flocculated floc is evaluated as to whether or not the water treatment is appropriately performed. Judgment is made in real time from the state of the flock. The machine learning algorithm learns various factors including the particle size of the flocculated floc. Therefore, the model constructed by the machine learning algorithm (that is, the learned model) can determine the quality of the flocculated floc state with higher accuracy than the conventional flocculation determination technology based on the particle size.

  In the construction of the model, the explanatory variables of the learning data are image data of the flocculated flocs, and the objective variables of the learning data are numerical values representing the quality of the flocculated flocs. For example, 1: very bad 2: bad 3: fairly good 4: Good, 5: Very good, which can be given as a five-grade evaluation. In one embodiment, a numerical value indicating the quality of the state of the flocculated floc may be given as a numerical value indicating an excess or deficiency of the coagulant injection rate. For example, a five-level evaluation of 1: little, 2: somewhat less, 3: optimal, 4: slightly more, 5: more can be used as a numerical value representing the excess or deficiency of the coagulant injection rate. However, the present invention is not limited to the five-level evaluation and may be, for example, a three-level evaluation or a ten-level evaluation.

  Which of the above five-level evaluation the actual coagulant injection rate corresponds to is determined by the turbidity and chromaticity of the treated water in the sedimentation process at the latter stage of the coagulation process, and the appearance of the flocculated floc by a veteran operator. Further, it can be determined from the quality of the sand filtration water at the subsequent stage, the rate of increase in the filtration resistance of sand filtration, and the like. A data set consisting of a combination of numerical values (1 to 5 in the above example) representing the excess or deficiency of the actual coagulant injection rate and the corresponding coagulation floc image data is used for building and updating the model. The numerical value representing the excess or deficiency of the actual coagulant injection rate is correct data, and is used for machine learning such as deep learning. In one embodiment, the correct answer data may be a numerical value (1 to 5 in the above example) indicating the quality of the actual state of the aggregated floc.

  FIG. 1 is a schematic diagram illustrating an embodiment of a model for determining the state of flocculated flocs. The model is a neural network having an input layer 301, a plurality of hidden layers (also called hidden layers) 302, and an output layer 303. Although the model shown in FIG. 1 has four hidden layers 302, the configuration of the model is not limited to the embodiment shown in FIG. Image data of the flocculated floc is input to the input layer 301 of the model. More specifically, the numerical value of each pixel constituting the image data of the aggregated floc is input to the input layer 301. In one example, when the image data of the agglomerated floc is grayscale image data, the gray level of each pixel (usually a numerical value within 0 to 255) is input to each node (neuron) of the input layer 301 of the model. When the image data of the aggregated floc is color image data, numerical values representing red, green, and blue of each pixel are input to corresponding nodes (neurons) of the input layer 301 of the model.

  In the example illustrated in FIG. 1, the output layer 303 of the model outputs a numerical value representing the determination result of the state of the flocculated floc. For example, in the case where the state of the flocculated floc is represented by a five-level evaluation of 1: very bad, 2: bad, 3: somewhat good, 4: good, 5: very good, the output layer 303 of the model Outputs a numerical value from 1 to 5. However, the configuration of the model shown in FIG. 1 is an example, and the present invention is not limited to the example shown in FIG.

  When the image data of the current flocculated floc is input to the model constructed as described above, the computer executes an operation according to the algorithm of the multilayer perceptron constituting the neural network, and the output layer 303 determines whether the state of the flocculated floc is good or bad. Is output, or a numerical value representing the determination result of the excess or deficiency of the coagulant injection rate. The operator can aim at optimizing the operation of the water treatment system based on the determination result.

  In one embodiment, information data related to water treatment is input to the model in addition to the image data of the flocculated flocs.

  FIG. 2 is a schematic diagram illustrating another embodiment of a model for determining the state of the flocculated floc. In this embodiment, the explanatory variables of the learning data include information data relating to water treatment in addition to the image data of the flocculated flocs. The input layer 301 of the model has a node (neuron) to which the image data of the aggregated floc is input and a node (neuron) to which the information data is input. According to the present embodiment, it is possible to improve the determination accuracy as compared with a model in which only the image data of the flocculated floc is used as an explanatory variable.

  The information data is data representing operating conditions of the water treatment system. More specifically, the information data includes raw water turbidity, chromaticity, pH, M-alkalinity, total organic carbon (TOC), chemical oxygen demand (COD), ultraviolet absorbance, water temperature, raw water flow, weather , Rainfall, and coagulant injection rate. In view of performing automatic control of the water treatment system using the model, it is desirable that items included in the information data be items that can be automatically acquired by the sensor, rather than items obtained by manual analysis.

  In particular, in the aggregation operation, pH and M-alkalinity are important for the quality of aggregation. For example, even when the coagulant injection rate is the same, if the pH and / or the M-alkalinity is different, the state of the flocculated floc differs. Therefore, as described above, a model is constructed in accordance with the machine learning algorithm using not only the image data of the flocculated floc but also information data including the pH of the raw water, the M-alkalinity, the coagulant injection rate, and the like. During actual water treatment, image data of flocculent flocs, information data such as pH of raw water, M-alkalinity, and flocculant injection rate are input to the trained model.

In one aspect, the determination result indicating the state of the flocculated floc is a determination result indicating an excess or deficiency of the coagulant injection rate.
In one aspect, the determination result is used for controlling the coagulant injection rate.

  The operation of the water treatment system can be automatically optimized based on the determination result of the coagulant injection rate being excessive or insufficient. For example, the determination result is output at a predetermined cycle, for example, every minute. At a certain time T, if the determination result is 1: "less", the current coagulant injection rate is increased by + 50% in 0.5 minutes, and if 2: "slightly less", the current coagulant injection rate is increased. The rate is increased by + 25% in 0.5 minutes, 3: if "optimal", the current coagulant injection rate is unchanged, and 4: if "slightly high", the current coagulant injection rate is 0. .5 drop by -25% in 5 minutes, 5: if "high", decrease current flocculant injection rate by -50% in 0.5 minute. Subsequently, at time T + 1, the determination result is output again, and the coagulant injection rate is similarly adjusted. This step is repeated until an optimal determination result is output.

  In the above example, the numerical value representing the determination result of the excess or deficiency of the coagulant injection rate is any one of 1 to 5, but the present invention is not limited to the above example. In one embodiment, the numerical values representing the determination results of the excess or deficiency of the coagulant injection rate are + 50% (corresponding to “small”), + 25% (corresponding to “slightly low”), and 0% (corresponding to “optimum”). , -25% (corresponding to "slightly many"), or -50% (corresponding to "many").

  Here, at time 0, that is, when switching from the existing control in the water treatment system to control based on the above model, or when newly starting the operation of the water treatment system, the coagulant injection rate at the time of starting the control is as follows: The value may be determined based on past knowledge of the water treatment system, judgment of a skilled person, batch tests such as a jar test, and the like. Control of the water treatment system is started based on the coagulant injection rate determined here. After that, after the coagulant is injected, taking into account the water flow time up to the point where the image is captured by the image acquisition device, when the control for optimizing the coagulant injection rate using the above model is started. good.

  In one embodiment, the water treatment method determines a reference value of the coagulant injection rate, which is irrelevant to the state of the coagulant floc, by a reference value calculation formula, and calculates a correction value of the reference value by the coagulant injection rate. And a step of correcting the coagulant injection rate based on the reference value and the correction value.

  When the quality of the raw water flowing into the water treatment system changes rapidly in a short time, such as when a guerrilla downpour occurs, control may be delayed and the quality of the treated water may be deteriorated. Therefore, the computer has a reference value calculation formula for determining the reference value of the coagulant injection rate independently of the state of the flocculant floc. To determine. Further, the computer determines the correction value of the reference value from the determination result indicating the excess or deficiency of the coagulant injection rate.

The reference value calculation formula is, for example, a reference value (D ₀ ) of a coagulant injection rate from a value of raw water quality (turbidity, chromaticity, pH, M-alkalinity, etc.) constantly monitored by a raw water quality meter. Is an expression for determining. This reference value calculation formula is, for example, from a data set of past operation results in a water treatment system, that is, a value of raw water quality and a corresponding coagulant injection rate, from a value of raw water quality by a statistical analysis method, It is possible to use a regression equation to calculate the corresponding coagulant injection rate. The flocculant is injected based on the reference value of the flocculant injection rate thus obtained, and flocculants are formed.

This coagulated floc is imaged by the above-described method, and image data of the coagulated floc is input to the above-described model, thereby determining whether the coagulant injection rate is excessive or insufficient. Here, assuming that D ₀ is a reference value (mg / L) of the coagulant injection rate, and d is a correction value represented by a positive or negative percentage (%) indicating a judgment result of excess or deficiency of the coagulant injection rate, The corrected flocculant injection rate (D, mg / L) is represented by the following correction formula (1).
D = D ₀ × (1 + d / 100) (1)

The correction value d represents the determination result of the excess or deficiency of the coagulant injection rate, which is + 50% (“low”), + 25% (“slightly low”), 0% (“optimum”), −25% (In the case of "slightly many") or -50% (in the case of "largely"). For _example, D 0 = 20mg / _L, when the d = -25%, corrected coagulant injection rate D is 15 mg / L. According to the embodiment described above, it is possible to more quickly control the coagulant injection rate based on the coagulated floc image data.

  In one aspect, the water treatment method further includes a step of updating the model using a data set including a combination of the image data and a numerical value indicating whether the actual state of the flocculated floc is good or bad.

  The above model using the machine learning algorithm is continuously updated based on newly obtained data in a water treatment system that is operating using the model. As a result, even if the properties of raw water in the water treatment system have changed significantly or the operation method has changed since the model was first constructed, the model that is updated according to the actual situation is not appropriate. It is possible to continuously output an appropriate determination result.

  In one embodiment, a stirring tank that stirs the raw water into which the coagulant has been injected to form a flocculated floc, an image acquisition device that generates image data of the flocculated floc, and a computer having a model constructed by a machine learning algorithm. A computer that stores the model and a processing device that executes an operation for inputting at least the image data to the model and outputting a determination result indicating the state of the cohesive floc from the model A water treatment system is provided.

  In one aspect, inputting at least image data of flocculated flocs to the input layer of a model comprising a neural network having an input layer, a plurality of hidden layers, and an output layer; and a multilayer perceptron constituting the neural network. A computer-readable recording medium on which a program for causing a computer to execute a step of outputting a numerical value indicating the state of the flocculated floc from the output layer by performing an operation in accordance with the above algorithm is provided.

  According to the water treatment method according to the present invention, from the model constructed by the machine learning algorithm and the image data of the flocculated flocs obtained by the image acquisition device such as a small camera, the flocculation conventionally determined by the human eye The state of floc can be automatically determined in real time with high accuracy, and more optimal operation of the water treatment system becomes possible.

It is a schematic diagram which shows one Embodiment of the model which determines the state of aggregation floc. It is a mimetic diagram showing other embodiments of a model which judges a state of a flocculation floc. It is a figure showing one embodiment of a water treatment system. It is a figure showing other embodiments of a water treatment system. It is a front view which shows a sampling container and an image acquisition apparatus typically. It is a side view which shows a sampling container and an image acquisition apparatus typically. FIG. 5 is a schematic diagram illustrating an embodiment of a computer constituting at least a part of the arithmetic system illustrated in FIGS. 3 and 4.

  FIG. 3 is a diagram showing one embodiment of the water treatment system. In this water treatment system, the state of the flocculant floc (excess or insufficient coagulant injection rate) is determined using the learned model in the water treatment system using the rapid filtration method, and the coagulant injection rate is controlled. All the descriptions including the construction of the model, the configuration of the model, and the like described in “Problems to be Solved by the Invention” are applied to each embodiment of the water treatment system described below.

  The water treatment system includes a landing well 1 for receiving raw water, a rapid stirring tank 2 and a slow stirring tank 3 as stirring tanks connected to the landing well 1, and a coagulant injection pump 5 for injecting a coagulant into raw water. ing. In the present embodiment, the flocculant injection pump 5 is connected to the rapid stirring tank 2 and is arranged so as to inject the flocculant into the raw water in the rapid stirring tank 2. In one embodiment, the flocculant injection pump 5 may be connected to a pipe connecting the landing well 1 and the rapid stirring tank 2.

  The rapid stirring tank 2 is connected to the landing well 1, and the slow stirring tank 3 is connected to the rapid stirring tank 2. The rapid stirring tank 2 is a stirring tank that stirs the raw water into which the coagulant has been injected, and forms a flocculated floc by coagulating a suspended substance in the raw water. The slow stirring tank 3 is a stirring tank that stirs water containing flocculated flocs and coarsens flocculated flocs.

  A settling tank 7 is connected to the slow stirring tank 3. The water containing the flocculated flocks grown in the slow stirring tank 3 is sent to the sedimentation tank 7, where the flocculated flocs are precipitated in the sedimentation tank 7. The structure of the sedimentation tank 7 is not particularly limited. However, in the case where an inclined plate or an inclined tube is provided inside the sedimentation tank 7 and the flocculated floc is settled on the inclined plate or the inclined tube to remove the flocculated floc. There is also. A filtration tank 8 is connected to the precipitation tank 7, and the supernatant water of the precipitation tank 7 is sent to the filtration tank 8. In the filtration tank 8, impurities such as coagulant residues and minute flocs are removed from the supernatant water, and the water is purified. The configuration of the filtration tank 8 is not particularly limited, but is generally configured to remove impurities by sand filtration or membrane filtration. The purified water is sent from the sedimentation tank 7 to the water purification tank 9. In one embodiment, the precipitation tank 7 may be omitted, and the slow stirring tank 3 may be omitted. In this case, the water containing the flocculated floc is sent to the filtration tank 8, where impurities such as flocculated floc are removed.

  The water treatment system inputs an image acquisition device 12 that generates image data of the flocculated floc formed in the rapid stirring tank 2 and at least the image data to a model constructed by a machine learning algorithm, and sets a state of the flocculated floc. The system further includes an arithmetic system 100 that outputs a pass / fail determination result from the model. The arithmetic system 100 includes a program for constructing and updating a model, a storage device 110 in which the model is stored, and a processing device 120 (such as a GPU or a CPU) that executes an operation according to the program. The processing device 120 inputs at least the image data to the model, and executes an operation for outputting a determination result of the state of the state of the flocculated floc from the model.

  The arithmetic system 100 is composed of at least one computer. The at least one computer may be a single server or a plurality of servers. The above-described model for determining whether the coagulant injection rate is excessive or insufficient (that is, the state of the flocculant floc) based on at least the flocculated floc image data is stored in the storage device 110 of the arithmetic system 100. Further, the above-described reference value calculation formula is also stored in the storage device 110 of the arithmetic system 100.

  The arithmetic system 100 may be an edge server connected to the image acquisition device 12 via a communication line, a cloud server connected to the image acquisition device 12 via a network such as the Internet, or an image acquisition device. 12 may be a fog computing device (gateway, fog server, router, etc.) installed in a network. The arithmetic system 100 may be a plurality of servers connected by a network such as the Internet. For example, the arithmetic system 100 may be a combination of an edge server and a cloud server.

  In the present embodiment, the flocculated floc in the rapid stirring tank 2 after the flocculant is injected is photographed by the image acquisition device 12. The image acquisition device 12 generates image data of the flocculated floc, and sends the image data to the arithmetic system 100. The arithmetic system 100 acquires image data from the image acquisition device 12 and stores it in the storage device 110. The image acquisition device 12 is installed above the rapid stirring tank 2 (above the water surface). An illuminator such as an LED light illuminates the flocculated floc in the rapid stirring tank 2 so that the image acquisition device 12 can clearly image the flocculated floc. The installation position of the image acquisition device 12 may be inside the rapid stirring tank 2 (below the water surface) or in a pipe connecting the rapid stirring tank 2 and the slow stirring tank 3 as long as the flocculated floc can be clearly imaged. Absent.

  The image acquisition device 12 is a digital camera including an image sensor (for example, a CCD image sensor or a CMOS image sensor) that can generate a still image or a continuous image. As the specifications of the digital camera, those having a manual focus function are desirable in order to always keep the focal length constant, and those having a number of pixels of 5 million pixels or more, preferably 10 million pixels or more are suitable.

  In the present embodiment, the flocculated flocs to be imaged are flocculated flocs in rapid stirring. This is because the control delay can be shortened, which is preferable from the viewpoint of controlling the coagulant injection rate. In the case of a water purification plant where fluctuations in raw water are poor, coarse flocs in the slow stirring tank 3 may be taken as an object of imaging.

  The arithmetic system 100 issues a command to the coagulant injection pump 5 to adjust the coagulant injection rate based on the determination result of the coagulant injection rate output from the model. Specifically, when the determination result indicates that the current coagulant injection rate is low, the arithmetic system 100 issues a command to the coagulant injection pump 5 to increase the coagulant injection rate. If the determination result indicates that the current coagulant injection rate is high, the arithmetic system 100 issues a command to the coagulant injection pump 5 to reduce the coagulant injection rate. As described above, the arithmetic system 100 can optimize the coagulant injection rate by controlling the operation of the coagulant injection pump 5 based on the determination result output from the model. The coagulant injection rate at the start of the optimization of the coagulant injection rate using the model is determined from knowledge based on past operation of the water treatment system and jar test results.

  As shown in FIG. 3, the landing well 1 is provided with a turbidity sensor 21, a pH sensor 22, and an M-alkalinity measuring device 23 for acquiring information data related to water treatment. The turbidity sensor 21, the pH sensor 22, and the M-alkalinity measuring device 23 are connected to the arithmetic system 100, and the turbidity measured value, the pH measured value, and the M-alkalinity measured value are sent to the arithmetic system 100. It is supposed to be. Further, at least one of total organic carbon (TOC), chemical oxygen demand (COD), ultraviolet absorbance, water temperature, raw water flow rate, weather, rainfall, and coagulant injection rate is input to the arithmetic system 100 as information data. May be done. The numerical value input to the arithmetic system 100 as the information data may be a measured value transmitted from a sensor or a value input to the arithmetic system 100 by an operator.

  In one embodiment, the information data described above is input to the input layer of the learned model as an explanatory variable together with the image data of the flocculated floc. The model outputs the determination result of the excess or deficiency of the coagulant injection rate as the objective variable. Since the quality of the raw water flowing into the image acquisition device 12 at the time of capturing the flocculated flocs and the quality of the raw water at the landing well 1 at that time may be different, the raw water quality corresponding to the acquired flocculated flocs image data It is desirable to adopt a value in consideration of the time required for the water to reach the image acquisition device 12 from the landing well 1.

  The model is constructed by acquiring various data using a water treatment system shown in FIG. 3 before operation using the model by using a deep learning method or the like. Here, in the model using the machine learning algorithm, the prediction accuracy for input data that has not been experienced is basically low, so the data in the construction of the model includes changes in raw water quality due to seasonal variations. desirable. Therefore, it is desirable to construct a model using data acquired over a period of preferably one year or more. However, if the raw water quality is stable, the data acquisition period can be shortened. Further, when a model has already been constructed in a nearby water treatment system that treats raw water in the same manner, or a water treatment system that treats similar raw water, the model can be used.

  The flocculated floc information output from the model relates to the quality of flocculated floc, that is, the excess or deficiency of the flocculant injection rate. The operator of the water treatment system may review the coagulant injection rate with reference to this information. However, as shown in FIG. 3, if this information is incorporated into the control of the coagulant, the coagulant injection rate can be continuously adjusted. Optimization can be achieved. In this way, the present water treatment system can continuously operate stably.

  However, if the operation of the water treatment system is prolonged, the model may not be able to output an appropriate determination result due to a change in the quality of the raw water or a change in a part of the operation flow. Therefore, it is necessary to appropriately update the model. Whether or not the determination result of the excess or deficiency of the coagulant injection rate was correct can be determined based on the turbidity of the supernatant water (precipitation water turbidity) at the outlet of the precipitation tank 7. If the flocculation treatment is not performed properly, the sedimentation of flocculated floc may be deteriorated, or the suspended matter remaining in the supernatant water without flocculation may increase, resulting in high turbidity of the precipitated water.

  Therefore, in the present embodiment, during operation of the model, the value of the sedimentation water turbidity is continuously acquired by the turbidity sensor 30 and the data set of the above-described explanatory variables and the objective variables is taken into consideration in consideration of the time of arrival. As additional information. This supplemental information is compared with the output result of the model, and the accuracy of the model is periodically verified. When the accuracy of the model is lower than that at the beginning of the operation, correct data is re-assigned to the objective variable of the data set based on the supplementary information, and the model is updated by a deep learning method or the like.

  When the computing system 100 includes an edge server at the edge (site) of the water treatment system and a cloud server connected to the edge server by a network (for example, the Internet), the model updating operation is performed by the cloud server. It is desirable to send the updated model (ie, the trained model) to the edge server and then store it in the edge server.

  FIG. 4 is a diagram showing another embodiment of the water treatment system. The configuration of this embodiment, which is not particularly described, is the same as that of the embodiment shown in FIG. In the present embodiment, similarly to the above-described embodiment, the excess / deficiency of the coagulant injection rate is determined by using a model that has been learned in the processing system using the rapid filtration method, and the coagulant injection rate is controlled. It is an example of a method.

  In the present embodiment, the imaging of the flocculated flocs is not performed in the water treatment area of the water treatment system, but is performed in the sampling container 41 arranged in the light shielding housing 40. That is, the water containing the flocculated flocs is extracted from the water treatment area and introduced into the sampling vessel 41 through the extraction pipe 43 branched from the water treatment area. Then, the image acquisition device 12 images the flocculated flocs in the sampling container 41. In the present embodiment, one end of the extraction pipe 43 is connected to a pipe connecting the rapid stirring tank 2 and the slow stirring tank 3. In one embodiment, one end of the extraction pipe 43 may be connected to a rapid stirring tank 2, a slow stirring tank 3, or a pipe connecting the slow stirring tank 3 and the precipitation tank 7.

  FIG. 5 is a front view schematically illustrating the sampling container 41 and the image acquisition device 12, and FIG. 6 is a side view schematically illustrating the sampling container 41 and the image acquisition device 12. The sampling container 41 and the image acquisition device 12 are arranged in the light shielding housing 40. A dark room is formed inside the light shielding housing 40. The sampling container 41 has a transparent side wall 41a. As a material of the transparent side wall 41a, quartz, transparent acrylic resin, transparent vinyl chloride, or the like can be used. As such a sampling container 41, a flow cell can be used.

  An illuminator 46 such as an LED light is arranged in the light shielding housing 40. The illuminator 46 illuminates the flocculated flocs in the sampling container 41 so that the image acquisition device 12 can clearly photograph the flocculated flocs. The illuminator 46 is arranged above the sampling container 41 and illuminates the sampling container 41 from above. The image acquisition device 12 is arranged facing the transparent side wall 41 a of the sampling container 41, and images the aggregated flocs in the sampling container 41.

  As a power for introducing the water containing the flocculated floc from the water treatment area to the sampling vessel 41, the method using the water level difference is preferable because it is the gentlest and the floc is not damaged. , A mono pump, a positive displacement pump, a peristaltic pump, or the like may be used.

  The water containing the flocculated flocks flows in from the lower part of the sampling vessel 41 and flows out of the sampling vessel 41 over the overflow weir 41b. Further, the flow rate of the water sent to the sampling container 41 may be such that the flocculated flocs do not settle in the sampling container 41. The water containing coagulated flocs which flowed out of the sampling vessel 41 may be returned to the rapid stirring tank 2 or the slow stirring tank 3, or may be discarded.

  According to the present embodiment, it is possible to eliminate the influence of sunlight and the reflection of a structure such as a stirring tank. In addition, by keeping the brightness in the dark room constant by the illuminator 46, it is possible to always image the flocculated flocs under the same conditions. Further, maintenance such as lens cleaning of the image acquisition device 12 can be easily performed.

  Although it depends on the raw water quality, dirt adheres to the transparent side wall 41a of the sampling container 41 within one to several days, which hinders the imaging of the aggregated floc image. To prevent this, cleaning or replacement of the sampling container 41 is manually performed once a day. Alternatively, an agent for removing dirt, such as a sulfuric acid solution or a sodium hypochlorite solution, is automatically flown into the sampling container 41 at a frequency (for example, once every three hours) according to the dirt, and the transparent side wall 41a is removed. Dirt may be removed. Further, a mechanism for automatically cleaning the transparent side wall 41a by a physical means such as a sponge or a wiper may be provided.

  FIG. 7 is a schematic diagram showing one embodiment of a computer constituting at least a part of the arithmetic system 100 shown in FIGS. The computer includes a storage device 110 in which programs and data are stored, a processing device 120 such as a CPU (Central Processing Unit) or a GPU (Graphic Processing Unit) that performs operations in accordance with the programs stored in the storage device 110, , A program, and various information to the storage device 110, an output device 140 for outputting a processing result and processed data, and a communication device 150 for connecting to a network such as the Internet. Have.

  The storage device 110 includes a main storage device 111 accessible by the processing device 120 and an auxiliary storage device 112 for storing data and programs. The main storage device 111 is, for example, a random access memory (RAM), and the auxiliary storage device 112 is a storage device such as a hard disk drive (HDD) or a solid state drive (SSD).

  The input device 130 includes a keyboard and a mouse, and further includes a recording medium reading device 132 for reading data from a recording medium, and a recording medium port 134 to which the recording medium is connected. The recording medium is a non-transitory tangible computer-readable recording medium such as an optical disk (for example, a CD-ROM or a DVD-ROM) or a semiconductor memory (for example, a USB flash drive or a memory card). is there. Examples of the recording medium reading device 132 include an optical drive such as a CD-ROM drive and a DVD-ROM drive, and a card reader. An example of the recording medium port 134 is a USB port. The programs and / or data stored in the recording medium are introduced into the computer via the input device 130 and stored in the auxiliary storage device 112 of the storage device 110. The output device 140 includes a display device 141 and a printing device 142.

  The learned model constructed by the machine learning algorithm is stored in the storage device 110. This trained model is a neural network having an input layer 301, a plurality of hidden layers (also called hidden layers) 302, and an output layer 303, as shown in FIG. 1 or FIG. The computer operates according to a program electrically stored in the storage device 110. That is, the computer obtains the image data of the flocculated floc from the image obtaining device 12, inputs at least the image data to the learned model constructed by the machine learning algorithm, and executes the processing of the multilayer perceptron constituting the neural network. By executing an operation in accordance with the algorithm, a step of outputting a numerical value representing a judgment result of the state of the flocculation floc or a judgment result of the excess or deficiency of the flocculant from the output layer 303 is executed. Further, the computer executes the steps of constructing the model and updating the model. A program for causing a computer to execute these steps is recorded on a non-transitory tangible computer-readable recording medium, and provided to the computer via the recording medium. Alternatively, the program and the model may be input to the computer from the communication device 150 via a communication network such as the Internet.

  The above embodiments have been described for the purpose of enabling a person having ordinary knowledge in the technical field to which the present invention pertains to implement the present invention. Various modifications of the above-described embodiment can be naturally made by those skilled in the art, and the technical idea of the present invention can be applied to other embodiments. Therefore, the present invention is not limited to the embodiments described, but is to be construed in its broadest scope in accordance with the spirit defined by the appended claims.

REFERENCE SIGNS LIST 1 landing well 2 rapid stirring tank 3 slow stirring tank 5 flocculant injection pump 7 sedimentation tank 8 filtration tank 9 water purification tank 12 image acquisition device 21 turbidity sensor 22 pH sensor 23 M-alkalinity meter 30 turbidity sensor 40 Housing 41 Sampling container 43 Extraction tube 46 Illuminator 100 Operation system 110 Storage device 111 Main storage device 112 Auxiliary storage device 120 Processing device 130 Input device 132 Recording medium reading device 134 Recording medium port 140 Output device 141 Display device 142 Printing device 150 Communication Device 301 Input layer 302 Hidden layer 303 Output layer

Claims (9)

Inject flocculant into raw water to form floc,
The image data of the flocculated floc is generated by an image acquisition device,
At least the image data of the flocculated flocks, input to a model constructed by a machine learning algorithm,
A water treatment method, wherein a determination result indicating a state of the flocculated floc is output from the model. The water treatment method according to claim 1,
The step of generating the image data of the flocculated floc by an image acquiring device includes extracting water including the flocculated floc and introducing the water into a sampling container, and generating the image data of the flocculated floc in the sampling container by an image acquiring device. A water treatment method. The water treatment method according to claim 1 or 2,
A water treatment method, wherein information data relating to water treatment is input to the model in addition to the image data of the flocculated flocs. The water treatment method according to any one of claims 1 to 3,
The water treatment method, wherein the determination result indicating the state of the flocculation floc is a determination result indicating an excess or deficiency of the coagulant injection rate. The water treatment method according to claim 4,
The water treatment method, wherein the determination result is used for controlling a coagulant injection rate. The water treatment method according to claim 5, wherein
Irrespective of the state of the flocculated floc, the reference value of the flocculant injection rate is determined by a reference value calculation formula,
The correction value of the reference value is determined based on the determination result indicating the excess or deficiency of the coagulant injection rate,
The water treatment method further comprising a step of correcting the coagulant injection rate from the reference value and the correction value. The water treatment method according to any one of claims 1 to 3,
The water treatment method further comprising a step of updating the model using a data set including a combination of the image data and a numerical value indicating whether the state of the actual flocculated floc is good or bad. A stirring tank that forms the flocculated floc by stirring the raw water into which the flocculant has been injected,
An image acquisition device that generates image data of the agglomerated floc,
Comprising a computer having a model constructed by a machine learning algorithm,
The computer includes a storage device in which the model is stored, and a processing device that executes an operation for inputting at least the image data to the model and outputting a determination result indicating the state of the cohesive floc from the model. Have a water treatment system. Inputting image data of at least agglomerated flocs to the input layer of a model comprising a neural network having an input layer, a plurality of hidden layers, and an output layer;
A computer-readable recording that records a program that causes a computer to execute a step of outputting a numerical value indicating the state of the state of the flocculated floc from the output layer by performing an operation according to an algorithm of a multilayer perceptron constituting the neural network. Medium.

Images