JP2002119956A - Turbidity predicting system, turbidity controlling system and turbidity managing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To predict with sufficient reliability turbidity of treated water obtained after a specified time, to determine feed rate of a coagulant by taking retardation time into consideration without using a complicated model and to automatically follow sudden change in quality of original water. SOLUTION: In a turbidity predicting system 20, from predicting variable data 41 among measured data 40 obtained from a water purification plant 50, by using an example base estimation model, the turbidity of the treated water formed in the water purification plant 50 after a specified time is estimated as an estimated turbidity 11. In a turbidity controlling system 30, from control variable data 43 among the measured data 40 obtained from the water purification plant 50 and the predicted turbidity 11 estimated by the turbidity predicting system 20, by using the example base estimation model, the feed rate of the coagulant to be fed in the water purification plant 50 is estimated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to a turbidity prediction system, a turbidity control system, and a turbidity management system.
In particular, turbidity prediction that predicts treated water turbidity after several hours at a water purification plant, determines the coagulant injection rate in response to turbidity and water quality fluctuations of raw water, and controls treated water turbidity within a desired range The present invention relates to a system, a turbidity control system, and a turbidity management system.

[0002]

2. Description of the Related Art In a general water purification plant, a series of purification treatments is performed in which impurities are formed into particles (floc) by adding a coagulant to raw water obtained from a river or the like, and further, sedimentation and filtration are performed. . In this purification process, the state of floc formation differs depending on the coagulant charging rate. For example, when the amount of the coagulant is too small, the impurities cannot be coagulated, or the coagulation of the impurities is insufficient to precipitate, which may adversely affect the subsequent filtration step. When the amount of the coagulant is too large, the processing cost due to the excessive addition of the coagulant increases.

[0003] In a conventional water purification plant, an attendant performs a beaker test using raw water to determine the input rate of the flocculant. Since the effect of the flocculant is affected by the pH and alkalinity of the raw water, the pH adjuster is supplied in consideration of chlorine supplied for sterilizing the raw water. Therefore, in the beaker test, the floc formation test was conducted by changing the combination of the input rate of the coagulant and the input rate of the pH adjuster stepwise using raw water as a sample. Was selected.

As a method for automating the determination of the coagulant introduction rate, a method using a multiple regression model defining a transfer function using an impulse response has been proposed.
In addition, focusing on the fact that there is a correlation between the particle size of the fine floc formed in the rapid mixing pond and the state of floc formation, the measured value of the particle size of the fine floc is fed back to obtain the desired particle size. There has been proposed a method in which a coagulant dosing ratio that becomes a set value is automatically determined by PI control (for example, see Japanese Patent Application No. 7-112103).

[0005]

However, in the conventional method of performing a beaker test, it takes a long time to determine the coagulant introduction rate. Therefore, when the raw water turbidity suddenly changes as in the case of rainfall. Cannot respond quickly and is difficult to automate due to the need for manual work. In addition, in the method using the multiple regression model, since a coagulant introduction rate, raw water quality and water temperature are in a non-linear relationship, a single internal model covering a wide range of change of these variables cannot be generated. For this reason, it is necessary to generate a plurality of internal models by dividing the range of change of each variable, and there has been a problem that a huge amount of work and time are required for model generation. Further, since the plurality of internal models are switched and used, the control may be discontinuous.

[0006] Further, the particle size of the minute flocs is measured to determine the PI.
In the method of controlling, the state of floc formation is predicted based on the particle size of the minute flocs, but it is difficult to obtain high prediction accuracy because other factors are not taken into consideration, and even if the flocs are fine, Since it takes a long time to form the feedback, a non-negligible error or time delay occurs in the feedback, and as a result, the PI control is not stable. The present invention has been made to solve such a problem, and an object of the present invention is to provide a turbidity prediction system capable of predicting turbidity of treated water obtained after a predetermined time with sufficient reliability. The purpose of the present invention is to provide a turbidity control system capable of determining a coagulant introduction rate in consideration of a delay time without using a simple model, and to further provide a turbidity control system capable of automatically following a sudden change in raw water quality. It is intended to be.

[0007]

In order to achieve the above object, a turbidity prediction system according to the present invention uses a set of predictor variable data obtained for each of a plurality of input variables indicating the behavior of a water treatment plant. A turbidity prediction system that performs an arithmetic process using a predetermined inference model to estimate and output a treated water turbidity of treated water generated after a predetermined time in a water purification plant according to a set of the predictive variable data. A plurality of historical data consisting of a set of predictor variable data obtained for each of a plurality of input variables necessary for the estimation and the actual treated water turbidity obtained from the water treatment plant using the predictor variable data as an input condition, The case-based input space is quantized according to a desired output tolerance to form a plurality of unit input spaces and each history data is arranged in the corresponding unit input space, and has one or more history data. A case base generated by creating a case representing history data of the unit input space for each position input space, and a unit input space corresponding to a new set of predictor variable data by searching the case base. A case retrieval unit for acquiring a case from one or more unit input spaces each having a case closest to the phase distance, and an output estimating unit for calculating and outputting a treatment water turbidity corresponding to a new set of predictor variable data And an output estimating unit.
The processing water turbidity corresponding to a new set of predictive variable data is calculated from the processing water turbidity of the case searched by the case search unit, and is output.

[0008] At least raw water turbidity, raw water flow rate, coagulant charging rate, temperature and treated water turbidity may be used as predictor variable data for the turbidity predicting system. The input rate and the pH of the treated water may be used.

[0009] The turbidity control system according to the present invention performs an arithmetic process on a set of control variable data obtained for each of a plurality of input variables indicating the behavior of the water purification plant using a predetermined inference model, thereby controlling the control variable. A turbidity control system for estimating and outputting an input rate of a flocculant to be charged in a water treatment plant according to a set of data, wherein a set of control variable data obtained for each of a plurality of input variables required for the estimation is provided. Using the control variable data as input conditions, a plurality of historical data consisting of the input rate of the coagulant actually injected at the water purification plant is fetched, and the case-based input space is quantized according to the desired output tolerance to generate multiple units. An example in which an input space is formed and each piece of history data is arranged in a corresponding unit input space, and each unit input space having one or more pieces of history data represents the history data of the unit input space. The case base generated by the creation and the case base are searched to obtain one or more unit input spaces having cases whose phases are closest to the unit input space corresponding to the new set of predictor variable data. A case search unit that obtains a case from each of the above, and an output estimating unit that calculates and outputs a coagulant dosing rate corresponding to the new set of predictor variable data, and the output estimating unit searches for the case. A coagulant input ratio corresponding to a new set of predictor variable data is calculated and output from the coagulant input ratio of the case.

[0010] As predicted variable data of the turbidity control system, a predicted turbidity, a raw water turbidity, a raw water flow rate, a temperature, and a treated water turbidity indicating the treated water turbidity after at least a predetermined time may be used. Raw water pH, pre-chlorine input rate, and treated water pH may be used as the prediction variable data.

[0011] The turbidity management system according to the present invention determines the charging rate of the coagulant to be charged in the water purification plant based on data obtained for each of a plurality of input variables indicating the behavior of the water purification plant. A turbidity management system that manages the turbidity of treated water generated at a treatment plant, wherein the treated water turbidity after a prescribed time is determined based on predicted variable data obtained for each prescribed input variable indicating the behavior of the water purification plant. A turbidity prediction system that estimates and outputs the predicted turbidity, and control variable data obtained for each predetermined input variable indicating the behavior of the water purification plant and the predicted turbidity obtained by the turbidity prediction system. And a turbidity control system for estimating and outputting the charging rate of the flocculant to be charged at the site.

As the turbidity prediction system of the turbidity management system, the turbidity prediction system according to the present invention described above may be used, and as the turbidity control system, the turbidity control system according to the present invention described above may be used. Good.

[0013]

Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a configuration diagram showing a turbidity management system according to one embodiment of the present invention. Hereinafter, a water purification plant that purifies raw water taken from a river and distributes the water will be described as an example. First, the water purification process in the water purification plant 50 will be described. Raw water taken from a river or the like by a pump 51 is introduced into a landing well 52, and a coagulant is injected in a mixing pond 53 to convert impurities into particles (flock). At this time, chlorine for sterilization (hereinafter referred to as pre-chlorine) is also supplied.

Thereafter, in the floc forming pond 54, the flocculant reacts over time to form flocs, and the sedimentation pond 5
At 5 the floc precipitates. The treated water obtained from the upper layer of the sedimentation basin 55 is filtered by a filter using sand or the like in the filtration basin 56 and stored in the pump well 57. Then, the water is pumped out from the pump well 57 according to the demand, and is distributed to a water consumer via a reservoir (not shown). Water treatment control device 58
In, the mixing rate of the coagulant to be charged into the raw water in the mixing basin 53 is adjusted based on the coagulant charging rate 12 instructed by the turbidity management system 10. Thereby, the turbidity of the treated water generated in the water purification plant 50 is managed within a predetermined range.

Next, the turbidity management system 10 will be described. This turbidity management system 10 predicts treated water turbidity after a predetermined time based on various measurement data 40 measured at the water purification plant 50, and based on the obtained predicted turbidity and various measurement data 40, This is a system that automatically and automatically determines the input rate of the coagulant. The turbidity management system 10 includes a turbidity prediction system 20 that estimates and outputs a treated water turbidity obtained after a predetermined time from the predicted variable data 41 of the measurement data 40 using the case-based inference model as a predicted turbidity 11,
And a turbidity control system 30 for determining the coagulant introduction rate 12 from the predicted turbidity 11 and the control variable data 43 of the measurement data 40 using a case-based inference model.

First, the turbidity prediction system 20 will be described. FIG. 2 shows a configuration example of the turbidity prediction system. The turbidity prediction system 20 includes history data 21, a case base generation unit 22, a case base 23, a case search unit 24, an output estimation unit 25, and an adaptive learning unit 26. The specific configuration of the turbidity prediction system 20 includes an information processing apparatus such as a personal computer as a whole, and the case base generation unit 22, the case search unit 24, the output estimation unit 25, and the adaptive learning unit 26 It is realized by software (application) executed on the device. The history data 21 and the case base 23 include data stored in an information storage device such as a hard disk or a memory.
Read and updated as needed.

The turbidity prediction system 20 uses the case-based inference model to calculate the treated water output from the sedimentation basin 55 after a lapse of a predetermined time required for water purification processing, for example, 1 to 6 hours, from the predicted variable data 41. Predict the turbidity of The history data 21 includes prediction variable data 41 used for prediction of turbidity in the measurement data 40 and the prediction data 4
1 is data that is a combination of the turbidity of treated water actually treated and generated in the water treatment plant 50, that is, the treated water turbidity (actual), and is measured in the past from the water treatment plant 50 in this case. A large number of data obtained by using is used.

The case base generation unit 22 uses the predicted variable data 41 as an input variable and the treated water turbidity (actual) as an output value, and sets the input space of each input variable to a desired value. The case base 23 is generated by quantizing with the prediction accuracy and forming a case. Case search section 2
In 4, one or more cases corresponding to the new predicted variable data 41 obtained from the water purification plant 50 are searched from the case base 23. The output estimating unit 25 derives an output value, that is, a predicted turbidity 11 corresponding to the new predicted variable data 41 from one or more cases searched by the case searching unit 24.

The predictive variable data 41 includes a water purification plant 50
Turbidity of raw water taken by the pump 51, raw water flow rate (or filtration flow rate), coagulant charging rate, air temperature around the floc formation pond 54 and sedimentation pond 55, turbidity of treated water output from the sedimentation pond 55 Is used. When considering the reaction effect of the flocculant due to the difference in pH value, in addition to the above variables, the pH value of raw water, the pre-chlorine input rate, and the pH value of treated water output from the sedimentation tank 55 are used. Is also good. As the prediction variable data 41, data relating to raw water (processed water) to be predicted is used, so that not only data at the time of prediction but also time-series data composed of data measured in the past are actually used.

In the adaptive learning unit 26, based on a combination of the new predicted variable data 41 and the treated water turbidity (actual) of the water actually processed and generated in the water purification plant 50, that is, the predicted actual data 42. , The corresponding cases in the case base 23 are sequentially updated. Note that the history data 21 and the case base generation unit 22 are not always necessary for the turbidity prediction system 20, and a separate arithmetic processing unit having the function of the case base generation unit 22 converts the case data 23 from the history data 21. It may be generated and incorporated into the turbidity prediction system 20. In addition, the adaptive learning unit 26 is not an essential component, and the necessity of the adaptive learning unit 26 may be determined according to a change in the behavior of the water purification plant 50.

Next, the turbidity control system 30 will be described. FIG. 3 shows a configuration example of the turbidity control system. The turbidity control system 30 includes history data 31, a case base generation unit 32, a case base 33, a case search unit 34, an output estimation unit 35, and an adaptive learning unit 36. As a specific configuration of the turbidity control system 30, it is composed of an information processing device such as a personal computer as a whole, and the case base generation unit 32, the case search unit 34, the output estimation unit 35 and the adaptive learning unit 36 It is realized by software (application) executed on the device. The history data 31 and the case base 33 are composed of data stored in an information storage device such as a hard disk or a memory.
Read and updated as needed.

The turbidity control system 30 uses the case-based inference model to calculate the treated water generated in the water treatment plant 50 after a lapse of a predetermined time required for the water purification process from the control variable data 43, for example, 1 to 6 hours. Of the coagulant required to control the turbidity of the water in the desired range. The history data 31 is data composed of a combination of the control variable data 43 used for estimating the coagulant input rate in the measurement data 40 and the coagulant input rate (actual) actually input in the water purification plant 50. A large number of data obtained in the past from the water purification plant 50 are used.

In the case base generation unit 32, the control variable data 43 of the history data 31 is used as an input variable, the coagulant introduction rate (actual) is used as an output value, and the input space of each input variable is set to a desired value. The case base 33 is generated by quantizing with the prediction accuracy to form a case. The case search unit 34 searches the case base 33 for one or more cases corresponding to the new control variable data 43 obtained from the water purification plant 50. The output estimating unit 35 generates new control variable data 43 from one or more cases searched by the case searching unit 34.
Is derived, that is, the coagulant introduction rate 12 is derived.

As the control variable data 43, in addition to the predicted turbidity after the above-mentioned predetermined time, the turbidity of the raw water taken by the pump 51 of the water purification plant 50, the raw water flow rate (or the filtration flow rate), the floc formation pond 54 And the temperature around the sedimentation basin 55,
The turbidity of the treated water output from the sedimentation basin 55 is used.
When considering the reaction effect of the flocculant due to the difference in pH value, in addition to the above variables, the pH value of raw water, the pre-chlorine input rate, and the pH value of treated water output from the sedimentation tank 55 are used. Is also good. As the control variable data 43, data on raw water (processed water) to be predicted is used, so that not only data at the time of prediction but also time-series data composed of data measured in the past is used.

In the adaptive learning unit 36, based on a combination of the new control variable data 43 and the input rate (actual) of the coagulant actually input in the water purification plant 50 in that case, that is, the control actual data 44, The 33 corresponding cases are updated sequentially. Note that the history data 31 and the case base generation unit 32 are not always necessary for the turbidity control system 30. The history data 31 and the case base 33 are separated by a separate processing unit having the function of the case base generation unit 32.
May be generated and incorporated into the turbidity control system 30. Also, the adaptive learning unit 36 is not an essential component,
The necessity of the adaptive learning unit 36 may be determined according to the change in the behavior of the water purification plant 50.

As described above, the present invention has practical operability, and according to the turbidity prediction system 20 described above, the turbidity of the treated water generated in the future in anticipation of the required time of the water purification treatment is determined. Can be predicted with sufficient reliability. Further, according to the turbidity control system described above, the coagulant introduction rate necessary for controlling the turbidity of the treated water generated in the future to a desired range in anticipation of the required time of the water purification treatment is sufficiently reliable. You can predict it. Therefore, in the turbidity management system 10, the turbidity control system 30 determines the coagulant introduction rate 12 using the predicted turbidity 11 predicted by the turbidity prediction system 20, and thus the water purification process requires some time. Even in this case, it is possible to quickly and automatically derive the predicted turbidity 11 and the coagulant charging rate 12 without the need for a manual operation by the attendant, and to carry out stable turbidity management without excess or deficiency in the coagulant to be charged.

In the turbidity prediction system 20 and the turbidity control system 30, the case bases 23 and 33 are configured using measurement data obtained by actually measuring the behavior of the water purification plant 50. The input / output relationship of the control system relating to turbidity prediction and turbidity control of the water purification plant 50 itself is included. This eliminates the need for a special internal model such as the multiple regression model used to represent the input-output relationship as in the past, provides continuity of prediction and control, and allows rapid changes in raw water quality due to rainfall and other factors. Can respond.

Further, in the turbidity prediction system 20 and the turbidity control system 30, existing cases having predicted variable data and control variable data similar to new predicted variable data and control variable data are searched from the case base. . At this time,
As described later, the input space is quantized using the input quantization number as a parameter to define the case base and the similarity, and the evaluation index value is calculated to determine the quantization number. For this reason, no convergence calculation is required, and furthermore, the perfection of the model can be evaluated immediately from the evaluation index value when the case base is generated, and there is no need to perform a model evaluation using separate test data.

The variables necessary for turbidity prediction and turbidity control in the turbidity prediction system 20 and the turbidity control system 30 include turbidity, flow rate, charging rate, temperature, and pH.
Turbidity can be accurately predicted with relatively few variables that are easy to measure. As a result, compared with the case of using a large number of various variables, the processing speed is improved by the small number of variables, the real-time performance of the entire system is improved, and it is necessary to install a large number of sensors for detecting these variables. And equipment costs can be reduced.

In the turbidity prediction system 20 and the turbidity control system 30 described above, although the data to be handled, that is, the contents of the history data 21 and 31 and the case bases 23 and 33 and the input / output variables are different, the configuration thereof is not limited. Is almost the same. Therefore, the case base generation units 22, 32,
The same software (application) that realizes the case search units 24 and 34, the output estimating units 25 and 35, and the adaptive learning units 26 and 36 can be used, and the software is executed in parallel by an information processing device such as a personal computer. do it.

Further, for the turbidity prediction system 20 and the turbidity control system 30, it is not necessary to use the case-based reasoning model as described above. For example,
If the coagulant feeding rate 12 can be determined from the predicted turbidity 11 by another method, only the turbidity prediction system 20 needs to be applied, and if the predicted turbidity 11 can be predicted by another method, Only the degree control system 30 may be used, and in any case, the operation and effect of the turbidity prediction system 20 or the turbidity control system 30 described above can be obtained.

Next, the operation of case-based reasoning used in the present embodiment will be described. Hereinafter, a case-based inference model of the turbidity prediction system 20 will be described as an example. However, this is applied not only to the turbidity prediction system 20 but also to the turbidity control system 30 in a similar manner. First, referring to FIGS. 4 to 6, the case base generation unit 2 of the turbidity prediction system 20
Operation 2 will be described. FIG. 4 is an explanatory diagram showing the concept of the phase used in the case-based inference model according to the present embodiment, FIG. 5 is an explanatory diagram showing the quantization process of the input space, and FIG. 6 is a flowchart showing the case-base generation process.

The case base 23 is a case base created by introducing the concept of topology (Topology), and the input space is quantized according to a desired output tolerance (required accuracy). The relationship between input and output is defined for each unit input space (hereinafter referred to as a mesh). The case search unit 24 selects a mesh corresponding to the new predictor variable data 41 by referring to the case base 23, and searches a case representing the respective meshes from the mesh or its neighboring meshes. The prediction variable data 41 is composed of the same input variable data as the history data 21 used for generating the case base 23.

The output estimating unit 25 calculates estimated output data corresponding to the new predicted variable data 41, here, the predicted turbidity 11, from the output data of one or more cases searched by the case searching unit 24. Output. In the adaptive learning unit 26,
The learning of the case base 23 is adaptively performed based on the new predicted performance data 42 actually obtained from the target behavior. The new predicted performance data 42 has the same configuration as the history data 21 used for generating the case base 23, but is composed of separate data that is not used as the history data 21. In this case, new data or the like obtained by actual measurement from the water purification plant 50 is used.

In the case-based reasoning model used in the present embodiment, the input space is quantized into a phase space based on the concept of continuous mapping in mathematics topology theory, so that an allowable width ε ( A general definition of the case base and similarity according to the output tolerance) is made.
The concept of continuous mapping in topological theory is, for example, the space X, Y
, The necessary and sufficient condition for the mapping f: X → Y to be continuous is an open set (near the output) O inverse mapping f in Y
^-1 (O) is equivalent to an open set of X (near the input). By using the concept of the continuous mapping, assuming that the mapping f from the input space to the output space is continuous, as shown in FIG. 4, the output neighborhood is determined by using the allowable width of the output error in the output space. The output neighborhood can be associated with the input neighborhood that satisfies the allowable range of the output error, and the input space can be quantized and regarded as a phase space.

The quantization process of the input space is performed according to a procedure as shown in FIG. The history data 21 is composed of a set of input data and output data obtained in the past, and in FIG. 5, is composed of inputs x ₁ and x ₂ and an output y. These history data are distributed in the input space x ₁ -x ₂ as shown in the upper right of FIG. When this is quantized by equally-spaced meshes having predetermined widths in the x ₁ and x ₂ directions as shown in the lower right of FIG. 5, the allowable width ε of the output error is considered as shown in the lower left of FIG. The size of the mesh, that is, the input quantization number is determined.

The allowable range ε of the output error is a value indicating to what extent an error between the output obtained by the estimation and the unknown unknown value with respect to the new input data is allowed, and is set in advance as a modeling condition. Therefore, this allowable width ε
Is used to determine the size of the mesh, the input neighborhood corresponding to the size of the output neighborhood, that is, a case can be defined, and the error of the output data estimated from all the input data belonging to the case is the output error of the output error. This satisfies the allowable width ε.

The case base generation unit 22 generates the case base 23 by using such a quantization process of the input space. In FIG. 6, first, the history data 21 is read (step 100), modeling conditions such as an allowable range ε of the output error are set (step 101), and various evaluation indices are calculated based on the allowable range ε. An input quantization number is selected for each input variable based on the index (step 102). Then, each case constituting the case base 23 is generated from the history data 21 allocated to each mesh (step 103).

Here, the process of determining the number of input quantizations using the evaluation index will be described with reference to FIGS. FIG. 7 is a flowchart showing a process of determining the number of input quantizations, FIG. 8 is an explanatory diagram showing an output distribution condition which is one of the evaluation indexes,
FIG. 9 is an explanatory diagram showing a continuity condition which is one of the evaluation indices;
FIG. 10 is an explanatory diagram showing the relationship between the satisfaction rate of each evaluation index and the input quantization number.

The output distribution condition is, as shown in FIG. 8, the output distribution of the output y of the history data belonging to the arbitrary mesh obtained by quantizing the input space with the selected input quantization number. The condition is that the width is smaller than the allowable width ε of the output error. Thus, it is checked whether one mesh, that is, the input neighborhood, satisfies the condition defined for the corresponding output neighborhood, that is, the allowable range of output error ε. As shown in FIG. 9, the continuity condition is, for an arbitrary mesh obtained by quantizing an input space with a selected input quantization number, an output value y of a case generated by the mesh, and the case The difference between the average output value y ′ of the surrounding cases existing around the similarity r is the allowable width ε (r +
1) The condition is smaller.

Thus, it is checked whether or not the difference between the output values between the cases, that is, between the input neighborhoods, satisfies the condition defined between the corresponding output neighborhoods, that is, the allowable width ε of the output error. By satisfying this continuity condition, it can be determined that each case continuously covers the input space so as to satisfy the desired accuracy. When inspecting this continuity condition, it is necessary to consider the distance between the case of the inspection target mesh and the surrounding cases, that is, the similarity r described later. This is to correctly reflect the concept of continuous mapping in the above-described topology. Here, since the output error allowable width in the mesh is within ε, when the similarity between the two cases is r, the output error allowable width is within ε (r + 1).
Therefore, the continuity condition is that the difference between the output value y of a case generated by an arbitrary mesh and the average output value y ′ of the case and surrounding cases of similarity r is the allowable range ε of the output error.
The condition is smaller than (r + 1).

In the process of determining the input quantization number, as shown in FIG. 7, first, an evaluation criterion (threshold) is set as a criterion for judging the quality of the evaluation index (step 11).
0). Then, each evaluation index is calculated for each input quantization number (step 111), the obtained evaluation index is compared with the evaluation criterion, and the input quantization number for which an evaluation index satisfying the evaluation criterion is obtained is calculated. Select (step 112). As an evaluation criterion, it is desirable to select an input quantization number at which 90% or more cases satisfy both the output distribution condition and the continuity condition, and the system displays a 90% or 95% division number. ing. This is because the values of 90% and 95% are considered to be statistically appropriate values.

This input quantization number is determined in order for each input variable. For example, the input variables _{x 1, x 2, ‥,} x n
Cases, will determine the number of input quantization in order from x ₁ to x _n. Here, when calculating the evaluation index, it is necessary to assign the input quantization numbers to all the input variables. Therefore,
when obtaining an evaluation index related to x _i, for x ₁ ~x _i-1, using the already number of input quantization is determined at that time, x _{i + 1} of the subsequent x _i, ‥, for x _n is , X _i are used.

Among the above-mentioned conditions, for the output distribution condition and the continuity condition, the ratio of cases satisfying the condition to all cases, that is, the evaluation index satisfaction rate is used as the evaluation index. For example, the evaluation index value of the input quantization number m with respect to x _{i is obtained by} quantizing the input range widths of x ₁ , x ₂ , ‥, and x _n with the respective input quantization numbers, and in all cases generated by quantization. , The ratio of cases satisfying the evaluation index condition, that is, the output distribution condition satisfaction rate and the continuity condition satisfaction rate. Then, for the input variable x _i ,
All evaluation index value they select one from the input quantization number who meet criteria is determined as an input the quantization number of the input variables x _i.

[0045] At this time, the evaluation index of the continuity condition satisfaction index S _C output distribution condition satisfaction index S _D, rather than increases monotonically with an increase in the input quantization number m, as shown in FIG. 10 Since the input quantization number m falls below the evaluation criterion again for some input quantization number m, the evaluation criterion is satisfied as m ₂ in some cases because the input quantization number m is parabolic. For this, among the maximum value m _max of the test input quantization number that is set in advance, the smallest input each evaluation index continuity condition satisfaction index S _C output distribution condition satisfaction index S _D satisfies the evaluation criteria By selecting the quantization number m ₁ , the optimal input quantization number can be selected without depending on the maximum value m _max , the number of meshes can be suppressed to a minimum, and the size of the case base can be reduced.

In the case base generation unit 22, the input quantization number is selected as described above, the input space quantized by the input quantization number, each history data is allocated to each mesh here, and the case is generated. Generated. FIG. 11 is an explanatory diagram showing the case generation processing, and FIG. 12 is a flowchart showing the case generation processing.

First, each input variable is quantized (divided) based on the selected input quantization number, and a mesh is generated (step 120). In Figure 11, the input variable x ₂ is divided into six along with input variable x ₁ is 10 divided. Then, each piece of history data is allocated to each mesh (step 121), a mesh in which the history data exists is selected as a case, and its input / output values and output values are calculated (step 122). As shown in the upper right of FIG. 11, when three pieces of history data are allocated to the same mesh, FIG.
1 As shown in the lower right, these are integrated as one case. At this time, the average value of the outputs y of the three pieces of history data is used as the output value representing the case, and the median value of the mesh is used as the input value representing the case.

In the turbidity prediction system of FIG. 2, the predicted turbidity 11 is estimated from the newly input predicted variable data 41 using the case base 23 generated in this manner.
First, the case search unit 24 searches for a similar case from the case base 23 using the value of each input variable of the predictive variable data 41 and the degree of similarity. FIG. 13 is an explanatory diagram showing the definition of similarity,
FIG. 14 is a flowchart showing a similar case search process in the case search unit 24. The similarity is the case base 23
Is a scale indicating the degree of similarity of each case to the mesh corresponding to the new input data among the meshes provided in the input space possessed by.

In FIG. 13, if a case exists in the central mesh corresponding to the input data, here, the predictor variable data 41, the case and the input data are defined as "similarity r = 0". In addition, the similarity to the case existing immediately next to the central mesh is “similarity r = 1”, and thereafter, the similarity increases by one each time the central mesh is separated by one mesh. Therefore, when the estimation is performed, the estimated value of the similarity r in the case has an accuracy within (r + 1) × the output error allowable width ε. At this time, if both cases are successfully used for the input value to be estimated,
It is expected that the output value is better than (r + 1) × ε. When only one case is used for the value to be estimated, it is expected that the accuracy is about (r + 1) × ε based on the continuity of input and output.

As shown in FIG. 14, the case search section 24 first takes in input data (step 130).
A mesh corresponding to the input data is selected from the input space of the case base 23 (step 13).
1) The similarity used as the case search range is initialized to 0 (step 132), and similar cases are searched from the case search range indicated by the similarity (step 133). If a case exists in the mesh corresponding to the input data (step 134: YES), the case is output as a similar case (step 136). On the other hand, step 134
In the case where there is no case in the mesh corresponding to the input data (step 134: NO), the similarity is increased by 1 to expand the case search range (step 13).
5) Return to step 133 and search for a similar case again.

As described above, in the case search unit 24, the similar case corresponding to the newly input data is stored in the case base 2
3, the output estimating unit 25 estimates output data Y corresponding to the new input data A based on these similar cases. For example, as shown in FIG. 15, when a case exists in the mesh 150 corresponding to the input data A (22.1, 58.4), the output value y = 70.2 of the case is selected as the estimated output data. You. As shown in FIG. 16, when no case exists in the mesh 151 corresponding to the input data A (23.8, 62.3), the case search unit 24 expands the search range 152 to search for similar cases. Search for. Then, the output estimating unit 25 calculates estimated output data from the searched cases. At this time, when a plurality of cases are searched, the average value of the output values of each case is used as the estimated output data. In this way, the output estimating unit 25 estimates and outputs the estimated output data Y corresponding to the new input data A.

Next, the operation of the adaptive learning unit will be described with reference to FIGS. As illustrated in FIG. 2, the adaptive learning unit 26 uses the case base 2 based on the new predicted performance data 42 obtained by actually measuring the target, that is, the water purification plant 50.
Update 3 At this time, the new predicted performance data 42 may be automatically obtained, for example, every hour, so that the adaptive learning can be automated. First, a case corresponding to new data is searched from the input space of the case base 23. Here, if there is a case corresponding to the new data, only the case is revised.

FIG. 17 is an explanatory diagram showing an adaptive learning operation when a corresponding case exists. Here, since there is a case 160 corresponding to the new data B (23.9, 66.8, 48.2), the output value y of the new data B = 4
From 8.2 and the output value 49.7 of Case 160 before revision,
A new output value y = 49.0 of the case is calculated.
As the output revision calculation formula, a forgetting factor C _Forget is provided, and the output value Y _old before revision and the output value Y of the new data B are added at the ratio indicated by the forgetting factor to obtain the output value after revision of the case. .

On the other hand, if there is no case corresponding to the new data, a new case is generated based on the new data. FIG. 18 is an explanatory diagram showing an adaptive learning operation when there is no corresponding case. Here, new data B
Since there is no case in the mesh 161 corresponding to (23.7, 62.3, 43.8), the median value of the mesh corresponding to the new data B is set as the input value, and the output value y of the new data B is represented as the representative value. Is newly generated and added to the case base 23.

As described above, the inference model according to the present embodiment is obtained by applying the case-based inference framework to the modeling, and based on the concept of Topology.
It can be said that this is a modeling technique applicable to general objects where the continuity of the input / output relationship of the system is satisfied. Therefore, the data is stored as a case in the identified input space, and at the time of output estimation, the reliability of the estimated output data can be indicated by the phase distance (similarity) between the input and the previously stored input case. In the present embodiment, the output data is estimated using such a model. Therefore, the following operational effects can be obtained as compared with a conventional inference model such as a neural network or a multiple regression model.

In the conventional inference model, 1) Since a special model structure is used to define the relationship between the entire input and output areas, much effort is required to find an optimum structure for the system. 2) When learning the history data, it is necessary to perform a convergence calculation for identifying a plurality of parameters of the model structure, and this process takes an enormous amount of time. 3) Even when updating the model based on new data, it is necessary to identify parameters, and it is actually difficult to perform adaptive learning. 4) It is difficult to grasp how reliable the model output value is with respect to the input value to be estimated.

On the other hand, according to the present embodiment, 1) Cases (questions and answers) experienced in the past are accumulated as a case base, and input / output cases including the input / output relation of the system are used. Therefore, a special model for expressing the input / output relationship is not required. 2) When generating the case base, the input space is quantized using the input quantization number as a parameter to define the similarity with the case base, and the evaluation index value is calculated to determine the quantization number. Therefore, no convergence calculation is required, and the degree of completion of the model can be evaluated from the evaluation index value. Thus, there is no need to perform model evaluation using separate test data as in the related art.

According to the present embodiment, 3) the answer to the newly input question is obtained by searching for similar cases. Therefore, the degree of similarity of the case searched for the question can be determined, and this similarity can be used for evaluating the reliability of the answer. 4) Since the case base is composed of individual cases,
The case base can be partially revised based on new data, and it is not necessary to identify parameters as in the conventional case, so that adaptive learning can be easily performed.

As for the problem of learning and convergence calculation in the conventional model, a case-based reasoning (Case-Based Rea
soning: CBR), there is a problem of defining the case base structure and the similarity. This is a major engineering problem that conventional case-based reasoning cannot be defined without sufficient knowledge of the object. In the case-based reasoning model according to the present embodiment, based on the concept of continuous mapping in mathematics topology theory, a unique definition of a case base and similarity according to output tolerance, that is, required accuracy,
This is done by quantizing the input space to a phase space.
Therefore, the input / output model can be determined without requiring sufficient knowledge of the target, that is, identification of the input / output structure.

[0060]

As described above, according to the present invention, as a turbidity prediction system, a water purification plant is used after a predetermined time by using a case-based inference model from prediction variable data of measurement data obtained from a water purification plant. Since the turbidity of the treated water generated in (1) is estimated as the predicted turbidity, the turbidity of the treated water to be generated in the future can be predicted with sufficient reliability in anticipation of the required time of the water purification treatment. In addition, as a turbidity control system, the control variable data of the measurement data obtained from the water treatment plant and the predicted turbidity of the treated water generated after a predetermined time are used at the water treatment plant using a case-based inference model. Since the input rate of the coagulant to be added is estimated, the coagulant input rate necessary to control the turbidity of the treated water generated in the future to a desired range in anticipation of the required time of the water purification treatment is sufficient. Predictable with reliability.

Furthermore, as the turbidity management system, the predictive turbidity predicted by the turbidity predicting system is used to determine the coagulant feeding rate by the turbidity control system, so that it takes some time for the water purification treatment. Even in this case, it is possible to quickly and automatically derive the predicted turbidity and the coagulant charging rate without the need for a manual operation by a staff member, and to carry out stable turbidity management without excess or deficiency of the coagulant to be charged. In addition, there is no need for a special internal model such as a multiple regression model to represent the input-output relationship as in the past, and continuity of prediction and control is obtained.
It can respond quickly to sudden changes in raw water quality due to rainfall.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a turbidity management system according to an embodiment of the present invention.

FIG. 2 is a configuration example of a turbidity prediction system.

FIG. 3 is a configuration example of a turbidity control system.

FIG. 4 is an explanatory diagram showing a concept of a phase used in a case-based inference model.

FIG. 5 is an explanatory diagram showing quantization processing of an input space.

FIG. 6 is a flowchart showing a case base generation process.

FIG. 7 is a flowchart illustrating a process of determining an input quantization number.

FIG. 8 is an explanatory diagram showing an output distribution condition.

FIG. 9 is an explanatory diagram showing continuity conditions.

FIG. 10 is an explanatory diagram showing a relationship between a satisfaction rate of each evaluation index and an input quantization number.

FIG. 11 is an explanatory diagram showing a case generation process.

FIG. 12 is a flowchart illustrating a case generation process.

FIG. 13 is an explanatory diagram showing a definition of similarity.

FIG. 14 is a flowchart illustrating a similar case search process.

FIG. 15: Output estimation operation (when a similar case exists)
FIG.

FIG. 16 is an explanatory diagram showing an output estimation operation (when no similar case exists).

FIG. 17: Adaptive learning operation (when a corresponding case exists)
FIG.

FIG. 18 is an explanatory diagram showing an adaptive learning operation (when no corresponding case exists).

[Explanation of symbols]

Reference Signs List 10: turbidity management system, 11: predicted turbidity, 12: coagulant charging rate, 20: turbidity prediction system, 21: history data, 22: case base generation unit, 23: case base, 24
... Case search unit, 25 ... Output estimation unit, 26 ... Adaptive learning unit,
30 turbidity control system, 31 history data, 32 case base generation unit, 33 case base, 34 case search unit, 35 output estimation unit, 36 adaptive learning unit, 40 measurement data, 41 prediction Variable data, 42: predicted result data, 43: control variable data, 44: control result data, 5
0: water purification plant, 51: pump, 52: landing well, 53: mixing pond, 54: floc formation pond, 55: sedimentation pond, 56: filtration pond, 57: pump well, 58: water treatment control device.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 4D015 BA19 BA21 BB05 CA14 EA03 EA32 FA02 FA16 4D062 BA19 BA21 BB05 CA14 EA03 EA32 FA02 FA16 5B049 BB00 CC31 DD05 EE01 EE05 EE12 EE41 FF00 FF09

Claims (9)

[Claims] 1. A set of predictive variable data obtained for each of a plurality of input variables indicating the behavior of a water purification plant is subjected to arithmetic processing using a predetermined inference model, whereby the water purification is performed in accordance with the set of predictive variable data. A turbidity prediction system for estimating and outputting treated water turbidity generated after a predetermined time at a site, comprising a set of predictor variable data obtained for each of a plurality of input variables required for the estimation and its prediction Plural historical data consisting of actual treated water turbidity obtained from a water treatment plant with variable data as input conditions are fetched, and the case-based input space is quantized according to the desired output tolerance to form multiple unit input spaces. Forming and arranging each of the history data in the corresponding unit input space, and creating a case representative of the history data of the unit input space for each unit input space having one or more history data. By searching the case base generated by the above, the case base is retrieved from one or more unit input spaces each having a case whose phase distance is closest to the unit input space corresponding to the new set of predictor variable data. And an output estimating unit that calculates and outputs a treated water turbidity corresponding to the new set of predictive variable data from the treated water turbidity of the case searched by the case searched unit. A turbidity prediction system comprising: 2. The turbidity prediction system according to claim 1, wherein at least raw water turbidity, raw water flow rate, coagulant charging rate, temperature, and treated water turbidity are used as the predictive variable data. Forecasting system. 3. The turbidity prediction system according to claim 2, wherein at least raw water pH, pre-chlorine input rate, and treated water pH are further used as the prediction variable data. 4. A set of control variable data obtained for each of a plurality of input variables indicating the behavior of the water purification plant is subjected to arithmetic processing using a predetermined inference model, whereby the water purification is performed in accordance with the set of control variable data. This is a turbidity control system that estimates and outputs the input rate of coagulant to be charged at the site, and sets a set of control variable data obtained for each of a plurality of input variables required for estimation and the control variable data as input conditions. Multiple pieces of historical data consisting of the input rate of the coagulant actually injected at the water purification plant are captured, and the case-based input space is quantized according to the desired output tolerance to form multiple unit input spaces and respond By arranging each of the history data in a unit input space, and creating a case representing the history data of the unit input space for each unit input space having one or more history data. By searching the selected case base and this case base, cases are obtained from one or more unit input spaces each having a case with the closest phase distance to the unit input space corresponding to the new set of predictor variable data. And an output estimating unit that calculates and outputs a coagulant dosing rate corresponding to the new set of predictive variable data from the coagulant dosing rate of the case searched by the case searching unit. A turbidity control system. 5. The turbidity control system according to claim 4, wherein, as the predictive variable data, a predicted turbidity, a raw water turbidity, a raw water flow rate, a temperature, and a treated water turbidity indicating a treated water turbidity after at least a predetermined time. A turbidity control system characterized by using: 6. The turbidity control system according to claim 5, wherein at least raw water pH, pre-chlorine charging rate, and treated water pH are further used as the prediction variable data. 7. A process generated in the water purification plant by determining an input rate of a coagulant to be injected in the water purification plant based on data obtained for each of a plurality of input variables indicating the behavior of the water purification plant. A turbidity management system for managing turbidity of water, comprising: estimating treated water turbidity after a predetermined time based on predicted variable data obtained for each predetermined input variable indicating the behavior of the water purification plant; Turbidity prediction system to output as a degree, Based on control variable data obtained for each predetermined input variable indicating the behavior of the water purification plant and the predicted turbidity obtained by the turbidity prediction system, in the water purification plant A turbidity control system for estimating and outputting a charging rate of a flocculant to be charged. 8. The turbidity management system according to claim 7, wherein the turbidity prediction system according to claim 1 is used as the turbidity prediction system. 9. The turbidity management system according to claim 7, wherein the turbidity control system according to claim 4 is used as the turbidity control system.