JP2002045882A - Water quality controller for sewage treatment plant - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a water quality controller of a sewage treatment plant capable of sufficiently grasping the characteristics of the water quality process of the sewage treatment plant and accurately controlling water quality. SOLUTION: Flow-in sewage data from a disturbance sensor 16, treating water data from water quality state sensors 171-175 and a manipulated variables from actuators 181-186 are inputted to a process value storage means 2 and stored. A process model is constructed in a sewage treatment process model identification means 3 based on information from the process value storage means 2. In an optimum manipulated variable decision means 4, the optimum manipulated variable of the actuators is obtained based on the flow-in sewage data and the treating water data by using the process model from the sewage treatment process model identification means 3.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a water quality control device for a sewage treatment plant having an actuator such as a pump or a blower, and more particularly to a water quality control device capable of performing water quality control with high accuracy.

[0002]

2. Description of the Related Art The most important role of a sewage treatment facility such as a sewage treatment plant is to treat incoming sewage and to maintain the quality of effluent discharged to rivers and the like as good as possible. By the way, in recent years,
For example, the quality of incoming sewage is changing due to phosphorus compounds and the like contained in synthetic detergents. Along with this, a relatively sophisticated process, such as the conventional decomposition of an organic substance, is required to be more advanced.

[0003] Here, the advanced treatment is a treatment for a specific substance such as nitrogen or phosphorus. It is known that when these nitrogen and phosphorus and other substances are released into closed water areas such as lakes and marshes, they become nutrient sources for microorganisms such as plankton present in these water areas and cause damage such as red tide. For this reason, there has been an active movement to legally set a numerical standard for the total amount of nitrogen and phosphorus discharged into natural waters.

[0004] In response, attempts have been made to develop various techniques for removing nitrogen and phosphorus (denitrification and dephosphorization) in the technical aspect. Among them, the control of so-called operation amount, which performs denitrification and dephosphorization by appropriately adjusting multiple actuators such as the aeration air volume that sends air to the biological reaction tank used for sewage treatment and the return sludge pump that circulates sludge Is one of the promising approaches.

[0005] Such a method based on water quality control is economically inexpensive because no new construction or remodeling is performed, and can be easily realized only by installing new software on existing equipment. In the sense, it is the most powerful approach.

Conventionally, the operation amount control for denitrification and dephosphorization has been performed by the following method.

As the most widely used method, there is a qualitative method of determining the operation amount based on the experience and intuition of the operator of the sewage treatment plant. In recent years, universities and municipalities have attempted to use the International Associat
ASM2 (Activated Sludge Model No.) that simulates the temporal dynamics of denitrification and dephosphorization published by the largest society in the field of sewage treatment called ion on Water Quality.
Some control is performed using 2). Further, as a method used by some manufacturers or the like, there is a method of directly determining the operation amount by modeling the correlation between the concentration of nitrogen or phosphorus and the operation amount using a neural network.

[0008]

Among such water quality control methods, the qualitative method of determining the operation amount based on the intuition and experience of the operator is a conventional method in which the conventional operation or control is continued. This seems to be an effective approach. However, when regulations with legally specific numerical values are introduced, operators cannot take prompt action because they do not have operating experience in such cases.

[0009] In such a qualitative method, when denitrification and dephosphorization are not performed well, it is found that the denitrification and dephosphorization are not performed, and then, for example, a chemical such as PAC (polyaluminum chloride) is introduced. Then, forcible removal of nitrogen and phosphorus is performed. In other words, the qualitative control method has a drawback that the control is followed when an event that has not been experienced so far or that cannot be easily predicted occurs.

On the other hand, in the control method based on the ASM2 model capable of simulating the dynamics, the parameters of the model are sufficiently calibrated in advance, and the change in water quality is predicted by using the calibrated model. By setting the operation amount for performing control in advance to an appropriate value, it is possible to prevent the control from following up.

However, the ASM2 model itself does not have an actuator for feeding a substance (= acetic acid) serving as a nutrient source of microorganisms. Such substance input is often performed in actual sewage treatment plant processes, and has a significant effect on nitrogen and phosphorus removal. Therefore, in the ASM2 model, the relationship between some actuators that perform water quality control in an actual sewage treatment plant and water quality cannot be simulated well, and the operation amount of the actuator cannot be determined in principle.

In particular, the input of acetic acid or the input of a chemical such as PAC is a factor that has a strong effect on the removal of (nitrogen or) phosphorus, and the inability to simulate this actuator requires the purpose of denitrification and dephosphorization. This is a fatal drawback when considering the control. In addition, the effect of water temperature, which largely controls the reaction rate of microorganisms, cannot be directly captured.
That is, the ASM2 model is an imperfect model for use in controlling a real sewage treatment process.

In the neural network approach, the manipulated variable is directly determined from the concentration of nitrogen or phosphorus. However, in the determination, the temporal dynamics of nitrogen or phosphorus is not sufficiently considered. That is, the current operation amount is determined from the current nitrogen and phosphorus concentrations, and the relationship between the nitrogen and phosphorus concentrations from the predetermined past to the present and the operation amount is not considered. .

In general, the sewage water quality treatment process is a short one and a few hours long after the operation quantity is determined (fixed) to a certain value and the actual water quality change ends (settles in a certain steady state). Things can take weeks, but empirically,
Further, it is known from a simulation using the ASM2 model or the like. Therefore, it is not enough to determine the relationship between the concentration of water quality (nitrogen or phosphorus) at a certain moment and the manipulated variable at a certain moment by a neural network. That is, the method using the neural network
It is incomplete in that it does not fully capture the characteristics of the sewage water quality process.

The present invention has been made in view of the above points, and provides a water quality control device of a sewage treatment plant capable of performing a water quality control accurately by sufficiently grasping characteristics of a sewage water quality process. The purpose is to do.

[0016]

SUMMARY OF THE INVENTION The present invention relates to a water quality control device for a sewage treatment plant that has a plurality of actuators and treats inflow sewage. A treated water data sensor that measures treated water data related to treated water treated at the treatment plant, and a process that stores inflowed sewage data from the inflowed sewage data sensor, treated water data from the treated water data sensor, and the operation amount of the actuator Value storage means, based on the information from the process value storage means, using a sewage treatment process model identification means for constructing a process model, and a process model from the sewage treatment process model identification means, to the inflow sewage data and treated water data Optimum operation amount determining means for obtaining an optimum operation amount of the actuator based on the It is quality control system of the sewage treatment plant according to claim.

According to the present invention, a substance such as acetic acid or P
Using a model close to the actual sewage treatment process taking into account the introduction of chemicals such as AC, quantitatively simulating water quality fluctuations in the actual sewage treatment process, On the other hand, an optimal command value can be given.

According to the present invention, the sewage treatment process model identification means includes: a process characteristic analysis means for analyzing characteristics of the sewage treatment process from information recorded in the process value recording means;
A sewage treatment system comprising: a structure identification unit configured to determine a structure of a process model from an analysis result obtained by the process characteristic analysis unit; and a parameter identification unit configured to estimate parameters of the process model determined by the structure identification unit. It is a water quality control device of the site.

According to the present invention, it is possible to construct a system that captures the characteristics of temporal and spatial dynamics of sewage quality using only data previously stored in a data server of a sewage treatment plant. By using this model, it is possible to construct a control system using water quality prediction values obtained by simulating the temporal and spatial dynamics of sewage water quality, or a control system incorporating this model itself. Optimal command values for a plurality of manipulated variables can be given only from the data obtained.

According to the present invention, the sewage treatment process model identification means constructs a process model in advance by a physicochemical method, and further comprises an actuator for feeding a substance serving as a nutrient source of microorganisms, and a substance feeding having a coagulating sedimentation effect. A water quality control device for a sewage treatment plant, further comprising a process model enhancing means for further enhancing a process model based on a correlation with an actuator to be performed.

According to the present invention, a correlation between a sewage quality model capable of simulating physical and chemical dynamics of sewage quality and an actuator used in an actual sewage treatment plant is described.
It is possible to express quantitatively using actual data,
Therefore, it is possible to simulate the actual dynamics of the sewage treatment process and determine the optimal operation amount. Also,
By using such a physical / chemical process, a function as a software sensor or a state observer for a water quality item that cannot be sensed can be provided, and the optimal operation amount can be further ensured.

According to the present invention, the optimum manipulated variable determining means includes a control parameter determining means for determining parameters related to the calculation of the optimal manipulated variable from a step response waveform of the model obtained by the sewage treatment process model identifying means, And a manipulated variable command value calculating means for actually calculating a manipulated variable command value using the parameters determined by the use parameter determining means.

According to the present invention, in determining the specific optimal operation amount, the characteristics of the sewage treatment process are determined by one to several parameters that characterize the sewage treatment process, such as the time constant, gain, and peak value of each water quality item. This makes it possible to express, so that when determining the optimal operation amount, the calculation can be performed with a small number of parameters, and a numerically stable optimal operation amount determination algorithm can be established. Also,
By incorporating physically significant parameters such as time constants and gains into the optimal manipulated variable calculation, it is possible to easily tune the optimal manipulated variable calculation when the time constant or gain of each water quality item changes. .

According to the present invention, the optimum operation amount determination means determines the operation amount so as to cancel the influence of disturbance of the model obtained by the sewage treatment process model identification means and to achieve a predetermined target water quality. Feedforward-type optimal manipulated variable determining means, and feedforward-type corrected manipulated variable determining means for obtaining an optimal manipulated variable by correcting the manipulated variable so as to correct an error between the sewage treatment water quality and the target water quality. It is a water quality control device of a sewage treatment plant.

According to the present invention, in determining the optimum operation amount, fluctuations in the treated water quality due to fluctuations in the quality and quantity of the inflowed sewage are predicted using a model representing the sewage treatment process, and this is fed-forward. The pre-compensation of the mold can keep the treated water quality stable, and the minimum necessary feedback compensation only when the treated water quality fluctuates due to errors due to model imperfections and the effects of unexpected disturbances , The stability of the treated water quality can be further increased.

According to the present invention, a plurality of inflow sewage data sensors of the same type are provided, and the optimum manipulated variable determining means includes data preprocessing means for averaging information from a plurality of inflow sewage data sensors of the same type. This is a characteristic water quality control device for sewage treatment plants.

According to the present invention, it is possible to correct the deviation of the water quality depending on the location in the tank of the sewage treatment plant, and to calculate the water quality value of the liquid having a spatial distribution in which the sewage treatment is actually performed. It can be represented by water quality values at points that do not have a distribution. Therefore, it is possible to accurately calculate the “error” between the output value from the identification model and the sensor measurement value, which is often created without considering the spatial distribution. As a result, a control system that performs a feedback operation on the “error” can be improved in accuracy.

The present invention has a plurality of actuators,
In a sewage treatment plant water quality control device that treats inflow sewage, an inflow sewage data sensor that measures inflow sewage data related to inflow sewage and a treated water data sensor that measures effluent data about treated water that is treated in the sewage treatment plant When,
Process value storage means for storing the inflow sewage data from the inflow sewage data sensor, the treated water data from the treated water data sensor, and the operation amount of the actuator, and a sewage for constructing a process model based on information from the process value storage means. A treatment process model identification means, and a process monitoring means for using the process model from the sewage treatment process model identification means to obtain and display a predicted value of the treated water data based on the inflow sewage data and the operation amount of the actuator. A water quality control device for a sewage treatment plant, characterized in that:

According to the present invention, by combining the sensor value of the current water quality state and the predicted value of the water quality, visualizing and presenting the same to the operator, the operation amount determined by the optimum operation amount determination means can be determined by the operator. The correction means may be a correction means for performing correction based on experience or the like.

[0030]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a sewage treatment system showing one embodiment of a water quality control device of a sewage treatment plant according to the present invention.

In FIG. 1, a water quality control device of a sewage treatment plant controls a sewage treatment process 1 as a control object.

That is, the sewage treatment process 1 includes a first sedimentation basin 11, an anaerobic tank 12, an anoxic tank 13, and an aerobic tank 1.
4 and a final sedimentation basin 15. Anaerobic tank 12
Is provided with an acetic acid input device 181 for inputting a nutrient source of microorganisms in a sewage treatment process, and a final sedimentation basin 15 is provided with a PAC injector 182 for coagulating and precipitating phosphorus using a chemical. Further, a circulation pump 183 for circulating digestive fluid is connected between the aerobic tank 14 and the anoxic tank 13.

The aerobic tank 14 is provided with a blower 184 for supplying oxygen as a respiratory source of microorganisms, and a return pump 185 for circulating microorganisms is provided between the final sedimentation tank 15 and the anaerobic tank 12. Have been. The final sedimentation basin 1
5 Excess sludge extraction pump 1 for extracting excess sludge
86 are provided.

The acetic acid input device 181 and the PAC injection device 1
82, circulation pump 183, blower 184, return pump 1
85 and the excess sludge withdrawal pump 186 each constitute an actuator.

The first settling basin 11 of the sewage treatment process 1
At the entrance, a disturbance sensor 160 for measuring the amount and quality of the inflow sewage is provided, and further, a water temperature sensor 19 is provided at the first settling basin 11.

The disturbance sensor 16 and the water temperature sensor 1
Reference numeral 9 denotes an inflow sewage data sensor that measures inflow sewage data relating to inflow sewage.

Further, in the sewage treatment process 1, a process state sensor (treatment) for measuring various water quality values such as nitrate concentration, ammonia concentration, phosphoric acid concentration, dissolved oxygen concentration, MLSS concentration, etc. (Water data sensors) 171 to 175 are provided.

This sewage treatment process 1 comprises A2O (Anox
ic-Anaerobic-Oxic) is a sewage treatment process called the AOAO method, AO method, step inflow method,
Alternatively, it can be applied to various processes such as a standard activated sludge method.

The water quality control device according to the present invention is a process for storing inflow sewage data from the disturbance sensor 16 and the water temperature sensor 19, treated water data from the process state sensors 171 to 175, and an operation amount from the actuators 181 to 186. Value storage means 2, sewage treatment process model identification means 3 for constructing a process model based on information from process value storage means 2, and actuators 181 to 186 using the process model from sewage treatment process model identification means 3. An optimal operation amount determining means for obtaining an optimal operation amount.

Next, the operation of the present embodiment having the above configuration will be described. First, process value recording means 2
, The disturbance sensor 16, the water temperature sensor 19,
An area for recording items related to input data and output data measured by the process state sensors 171 to 175 and an area for recording an operation amount command value determined by the optimum operation amount determination means 4 are secured, and these areas are initialized. I do.

Next, the state values of various water qualities such as nitrate concentration, ammonia concentration, phosphoric acid concentration, dissolved oxygen concentration, and MLSS concentration, which are observed at predetermined intervals from the process state sensors 171 to 175 of the sewage treatment process 1 are shown. The time series data (output data) is recorded in the process value recording means 2. At the same time, the operation amount values of the actuators 181 to 186, the values of the disturbance sensor 16, and the water temperature sensor 19 over a predetermined period of time operated by a previously given operation method.
Is recorded in the process value recording means 2.

Next, from the data stored in the process value recording means 2, various water quality data of the same period as the time series data of the manipulated variable over a predetermined period is extracted, and the sewage treatment process shown in FIG. The data is delivered to the process characteristic analysis means 31 of the model identification means 3. FIG.
In the process characteristic analyzing means 31, first, the data of each operation amount of the actuators 181 to 186 shown in FIG. 1, the water quality data from the disturbance sensor 16 and the inflowing sewage data relating to the inflowing sewage from the water temperature sensor 19, and the process state sensor The correlation analysis with the treated water data on the quality of the treated water measured at 171 to 175 is performed.

In this case, for example, a periodogram method that is often used when performing correlation analysis between time-series data is used. By using such a periodogram method, correlation as time-series data can be examined, and correlation analysis in consideration of temporal dynamics can be performed. Here, the operation amount data and the inflow sewage data become input data, and the treated water data becomes output data. In this manner, the strength of the correlation between each input data and each output data can be quantitatively grasped.
A list as shown in FIG.

Next, an analysis of the strength of the nonlinearity for further analyzing the characteristics of the process in detail is performed. As a method of checking the strength of the nonlinearity, for example, the following method is available. First, each input / output data is classified into several data sets. The number of data sets to be classified is appropriately selected within the following range.

Number of data in one data set × data collection cycle> Slowest time constant of various water quality data That is, the length of time series data included in each data set is longer than that of the slowest response in the sewage treatment process. Choose to be longer. The number of data sets is set to at least two or more.

Next, for each of the data sets relating to the input data and each of the data sets relating to the output data among the classified data sets, for example, deconvolution or so-called linear system identification is performed. If the result of the deconvolution, which describes the relationship between a particular input and output, is not very dependent on the selected dataset and produces similar results, then the input-output relationship is not very nonlinear Can be concluded. Conversely, if they produce significantly different results, it can be concluded that the input-output relationship has significantly strong nonlinearity. This operation is shown in FIG.

The following method is conceivable as another method for detecting nonlinearity. First, difference data of each cycle of the input time series data and difference data of the output time series data are created. Then, the difference data of the output time-series data is divided by the difference data of the input time-series data. At this time, since the time-series data is composed of a plurality of values (here, N), data at the same time are divided. If the N pieces of data generated in this way are significantly different, the nonlinearity is considered to be strong. If the N pieces of data have similar values, the output data is similarly divided by the data with the time shifted by one toward the future with respect to the input data, and the values are compared. At this time, if the values of N-1 (by shifting the time of data, one uncorresponding data is generated and the number of data is reduced by one) are significantly different, the nonlinearity is considered to be strong. If the N-1 data have values close to each other, the same procedure is repeated. By repeating this procedure N times, the strength of the nonlinearity can be detected.

Next, the sewage treatment process model identification means 3
Then, based on the relationship between each input and each output analyzed by the process characteristic analyzing unit 31, the structure identifying unit 32 identifies the structure of each input / output characteristic. First, those whose correlation shown in FIG. 3 is equal to or less than a certain value, for example, 0.5 or less, are ignored. After that, the structure of each input / output characteristic is identified.

For example, as shown in FIG. 5 (a), the relationship between the aeration amount and the ammonia concentration is such that the ammonia concentration does not change gradually as the aeration amount is increased. It is determined by the analyzing means 31. Therefore, a model having a non-linear function model representing a saturation characteristic on the input side and a dynamics represented by a linear transfer function is selected as a model structure for describing a relational expression between aeration flow rate and ammonia concentration.

As shown in FIG. 5 (b), the nitric acid concentration with respect to the return rate is opposite when the return rate is low and when the return rate is high. The model has a structure with a switching mechanism.

Further, as shown in FIG. 5C, since the relationship between the PAC injection and the phosphoric acid concentration has an effect proportional to the PAC injection amount, for example, a simple proportional model is adopted. Further, since the relationship between the acetic acid input amount and the phosphoric acid concentration is expressed by a relationship having a certain degree of saturation characteristics, the relationship is expressed by a nonlinear model in which the saturation characteristics are applied to the input similarly to the relationship between the aeration air flow and the ammonia concentration.

As shown in FIG. 5 (d), the reaction is generally accelerated when the water temperature rises. For example, a nonlinear function model in which the change becomes more intense when the water temperature rises as the water temperature increases is inputted. On the side, and represent the dynamics by a model represented by a linear transfer function.

Thus, the process characteristic analyzing means 31
The structure of the model is determined by the structure identification means 32 by reflecting the characteristics of the process obtained by the process. The structure identification means 32 is performed for all combinations of all inputs and all outputs.

Next, each parameter of the structure determined by the structure identification means 32 is identified by the parameter identification means 33. In this case, the parameters are estimated using the time series data of each input and each output stored in the process value storage means 2, for example, using the least squares method.

Further, for example, referring to FIGS.
In the case of a nonlinear function and a linear transfer function as in the structure shown in (d), L. Sun et al. "New Approach to
Paramter Estimation of A Hammerstein Model, "Tran
s of SICE, Vol. 34, No. 11, pp. 1-9, (1998).

Thus, the process model identification means 3
When the process model is constructed in the above, the optimum operation amount of the actuators 181 to 186 is obtained by the optimum operation amount determining means 4 based on the process model as follows.

Variations in the quality of the treated water depend on the actuator 181
When the manipulated variable of １８186 is not operated at all, it is caused by the amount of inflow sewage and the quality of inflow sewage which are disturbances. Therefore, basically, the optimum operation amount can be determined by the following procedure. First, assuming that there is no fluctuation in the inflow sewage water amount, the inflow sewage water quality, and the water temperature which are disturbances, an optimal operation amount for stably treating the water quality is determined. Then, if the manipulated variable is corrected so that the fluctuation of the treated water quality due to the fluctuation of the amount of inflow sewage, which is disturbance, and the water quality or the water temperature becomes zero, this becomes the optimum manipulated variable.

This configuration procedure will be specifically described. First, the treated water quality as output and the actuators 181-1 as hydraulic power
The relationship between the manipulated variable of 86, inflow / sewage / water quality and water temperature
Give as follows.

Y = f (u, d) (1) Here, y is a vector composed of the treated water quality of output, u is a vector composed of the manipulated variable, and d is a vector composed of the inflow sewage water quality, the inflow sewage water amount and the water temperature. Vector.

F (.) Is a function representing the relationship between the manipulated variable and disturbance and the treated water quality in the steady state (assuming that the steady state exists).

The function f (.) Can be easily derived, for example, by obtaining a relational expression in a steady state of the process model constituted by the above-described process model identification means.

[0062] Now, (1) the left side of y expression, if equal to the target value y _R which is a measure of the advance gave stable processing quality, it is possible to obtain a stable quality of treated water as a result.

[0063] Thus, by appropriate manipulation of the u in equation (1), it suffices to left becomes y _R. This is (1) was placed and left y _R of the formula, also fixed to the amount it is assumed that the inflow sewage water quality is assumed not to vary the flow sewage water and water temperature d (which is referred to as d _E.), This may be solved for u. However, if the solution does not exist, a number of elements of the vector y _R, as a value within any or a range given in advance, can be solved. In this way an optimum operation amount when there is no flow into sewage water quality and variability of inflow sewage water and water temperature found to u _E.

Next, in order to consider fluctuations in the inflow sewage water quality, the inflow sewage water amount, and the water temperature, the relationship between the treated water quality as the output, the operation amount as the input, and the relationship between the inflow sewage amount / water quality and the water temperature is taken into consideration. Give as follows.

Y = f (u, d) + bΔu + aΔd (2) Here, aΔd is a term caused by fluctuations of the inflow sewage water quality, the amount of inflow sewage water, and the water temperature, and bΔu is associated with this fluctuation. This is a term based on the amount of operation to be corrected. Also, a or b
Is a term representing the influence on the fluctuation of the treated water quality due to the fluctuation of the inflow sewage water quality / water amount and the water temperature or the operation amount. For example, a linear approximation transfer function model is derived from the process model constructed by the process model identification means. Can be built by Here, it should be noted that a and b can express temporal dynamics. Further, Δ such as Δd and Δu means a variation of each vector value variable. (2) from the equation, in order to suppress the variation of Δd is, Delta] u a Δu = b ^{- 1} aΔd Tosureba can completely erases the variation of the inflow sewage water. Where b ⁻¹
If does not exist, an approximate inverse model is substituted. Therefore, the final feedforward optimal operation amount is as follows.

U = u _E + b ⁻¹ aΔd (3) The above process model is used when obtaining the manipulated variable of the equation (3). Therefore, due to the error between the actual process and the process model, with the manipulated variable in equation (3),
Quality of treated water does not become y _R, there is a case in which cause the error. Therefore, in order to correct the error, it is necessary to use feedback compensation such as a PI controller. Therefore, for example, the following optimal operation amount can be obtained in the following form.

[0067]

(Equation 1) Here, Δy = y _R −y. In the equation (4), the first item on the right side is the optimum manipulated variable (= constant value) when there is no change in the inflow sewage quality or the amount of inflow sewage, and the second item is the change in the inflow sewage quality, the inflow sewage amount or the water temperature. And the third and fourth terms are feedback correction terms that compensate for the actual fluctuations in the treated water quality with feedback.

Incidentally, the treated water quality y is usually measured by the process state sensors 171 to 175. However, the area of the treatment tank in the sewage treatment plant is relatively large, and the value varies greatly depending on the measurement place. It can be something. Therefore, the value of Δy is also changed by the process state sensor 171.
~ 175, which is different. Then,
The correction amount of the feedback correction term in equation (4) greatly differs depending on the process state sensors 171 to 175. In order to avoid this, for example, FIG.
As shown in the figure, the final sedimentation basin 15 and the vertical bisector are divided into a plurality of locations, the same type of process state sensors 171 to 175 are installed in each location, and the data preprocessing means 4a
To obtain the average value. Alternatively, countermeasures such as using a weighted average value according to the dominant region of the process state sensor using a Voronoi diagram (Thesen method) can be taken. For example, equation (4) y can be obtained as follows.

Y = (value of sensor 1 × area of region 1 + value of sensor 2 × area of region 2 + value of sensor 3 × area of region 3) ÷ total area (5) Thereby, the reliability of the feedback correction term of the equation (4) can be improved.

The data preprocessing means 4 a for averaging information from the process state sensor is built in the optimum manipulated variable determining means 4.

According to the present embodiment, acetic acid injection and PAC
Including operations often performed in sewage treatment plants, such as injection
The temporal dynamics of the actual sewage treatment water quality process can be quantitatively expressed in consideration of the effect of fluctuations in water temperature. Then, using a process model, which is a means for quantitatively expressing temporal dynamics, an optimal operation amount of the actuator for maintaining the sewage treatment water quality at a desired value is quantitatively determined in advance by a feedforward type controller. be able to. In addition, by using it together with a feedback type controller, a highly accurate and highly reliable water quality control device for a sewage treatment plant can be constructed. In addition, by installing a plurality of process state sensors for the quality of the treated water and performing appropriate treatment on the sensors, a more accurate water quality control system can be constructed.

Second Embodiment Next, a second embodiment of the present invention will be described. First, data is stored in the process value storage unit 2 in the same manner as in the first embodiment. Next, the sewage treatment process model identification means 3 will be described.

In the sewage treatment process model identification means 3,
As a water quality model of the sewage treatment process 1, a model structure is constructed using a physical model or a chemical model as described below.

That is, the first sedimentation basin 11 and the final sedimentation basin 15
On the other hand, a physical equation representing a phenomenon in which contaminants physically settle is used. For example, using Hazen's theory, for a given influent water quality, the outflow water quality concentration is determined as follows.

X _out = X _in / (1 / u / (Q / A))) (6) X _out : Outflow water concentration, X _in : Inflow water concentration u: Sludge settling velocity, Q: Inflow amount = Outflow, A: surface area of sedimentation basin. Here, it is assumed that everything except X _out is known.

Equation (6) shows that the sludge settling velocity is known,
It is derived based on the material balance under the assumption that the liquid in the sedimentation basin is completely mixed.

Anaerobic tank 12, anoxic tank 13 and aerobic tank 1
In No. 4, a chemical formula for decomposing organic substances by a microbial reaction is used. The chemical formula representing the reaction rate is generally given in the following format.

[0078]

(Equation 2) S: tank water quality concentration, S _in : inflow water quality concentration, F
(•): reaction rate, Q: inflow = outflow, V: water in tank The water quality variable X _out of (6) and the water quality variable S of (7) are:
Generally, there are a plurality of IAWQ activated sludge model Nos.
In 2 (ASM2), 17 water quality variables are defined. In addition, the reaction rate F (·) is often represented by a so-called mono-form when expressing the decomposition of a substrate by a microorganism. F (•) includes a plurality (for example, ASM2
(Several 60 parameters) are included.

The amount of water and the quality of water between the facilities are mixed simply by averaging or averaging the concentration, and the model is obtained by connecting Q, Xin, Sin, etc. in (6) and (7). Can be

Further, since the circulation pump 183, the return pump 185, and the surplus sludge extraction pump 186 as actuators merely adjust the amount, they can be modeled as follows, for example.

Q _out = L (Q _in ) (8) Q _in : flow rate before pump, Q _out : flow rate after pump, L:
Function Regarding Pump Pumping Amount For the blower 184, a relational expression between the air flow from the blower 184 and the dissolved oxygen concentration in the aerobic tank 14 can be modeled as follows, for example.

[0082]

(Equation 3) It is. Here, the dissolved oxygen concentration S ₀₂ is an element of the tank water quality variable S (that is, S is a vector value, S ₀₂ is an element in the vector), and the respiration rate Rr
(•) corresponds to the element of the reaction rate F (•) in equation (7).

Using the above equations (6) to (9), a model of the sewage treatment process can be constructed on a computer.

However, when the above equations (6) to (9) are used for actual control, the following three problems may occur.

1) In so-called advanced treatment for the purpose of removing nitrogen and phosphorus, important actuator elements such as acetic acid injection and PAC injection, which are often performed, cannot be expressed.

2) It is not clear how to set many parameters mainly included in F (•) in equation (7).

3) Fluctuations in water temperature cannot be captured explicitly.

Therefore, in order to solve these problems,
The sewage treatment process model identification means 3 may have the following process model functionalization means 3a.

The following functions are performed in the process model enhancing means 3a. First, the introduction of acetic acid is intended mainly to replenish nutrient sources of microorganisms involved in phosphorus removal. For example, when ASM2 is used as a chemical formula for decomposing organic matter by a microbial reaction of the formula (7), such a nutrient is used. the source is represented by the variable S _a. Therefore, it is considered that there is a strong correlation input amount of acetic acid and the variable S _a (S _a is
This is an element of the vector S in the equation (7). ). Therefore, the relationship of acetic input amount Q _S and S _a for example, expressed as follows.

S _a = K _S Q
_S
......... (10) where, _{K S} is a coefficient parameter representing a proportional relationship.
Alternatively, slightly more as a complex relationship, providing a _{S a = K 1 Q S +} K 2 Q S 2 ... + K N Q S N ......... nonlinear structure as that (11). Here, K _i , i =
1,..., N are coefficient parameters.

Similarly, it is known that there is a strong correlation between the PAC injection amount and the concentration of phosphoric acid. However, since a chemical reaction is interposed between PAC and phosphoric acid, there is a temporal dynamic due to this. So, for example, PAC
Injection volume _{Q PAC} and phosphoric acid concentration _S PO4 _{(S PO4} is an element of the vector S in equation (7).) The relationship expressed as follows.

[0092]

(Equation 4)

(Equation 5) Now, in the equation (13), H (•) and Y can be calculated by taking out the process data recorded by the process value storage means 2 in FIG. Therefore, for example, using an estimation method such as the least square method, (1
The estimated value of the parameter θ in the expression 3) can be calculated as follows.

[0093]

(Equation 6) By the above operations, it is possible to construct a process model that can simulate an actual sewage treatment process by incorporating the dynamics of water quality change by an actuator such as acetic acid injection or PAC injection into a physical / chemical model.

The construction of such a process model is performed by the process model identification means 3. In such a method, a large amount of data is not required unlike the first embodiment because the process data stored in the process value storage means 2 is used only when obtaining θ in the equation (14).

When the process model is constructed in the process model identification means 3 in this manner, the actuators 181 to 18 are next determined by the optimum manipulated variable determination means 4.
6 are obtained. First, for the process model identified by the process model identification means 3, the inflow sewage water quality and the amount of inflow sewage water, which are disturbances, and the feedforward type optimal operation amount determination means are determined by the same method as in the first embodiment. To obtain the operation amount u _E and input these.

Thereafter, the feedback-type correction operation means applies the process model to the process model and the fluctuation range of the inflow sewage water quality and the inflow sewage water amount within a predetermined time, and a predetermined predetermined value. A variation range of the operation amount within a range allowed to be operated within a time is obtained. Then, the stepwise change in the fluctuation range obtained in this manner is sequentially given to each of the manipulated variables, the inflow sewage amount, and the inflow sewage quality.

When a step-like change is given in this way, a change in treated water quality as an output of the process model is observed (not a measured value). From this observed waveform,
Time constant, gain, etc. of treated water quality for each manipulated variable and each disturbance
Peaks and the like can be read. When the control parameters are determined in this manner, for example, by looking at the gain, it is possible to see which manipulated variable significantly changes the treated water quality with respect to which operation amount. Similarly, for example, by looking at the time constant, it is possible to see which manipulated variable can change a certain treated water quality most quickly. When this is quantitatively expressed, for example, Δy = AΔd + BΔu (15)

(Equation 7) Can be written in the form Where A _ij and B _ij
Are ij elements of A and B, and are given by a so-called transfer function. (16) and (17) of the parameters _{a i} j to h
Note that it includes information on the time constant and the gain of _ij , and can be easily converted from the gain and the time constant. And
For the formulas (15) to (17), for example, an ordinary multivariable P
ID control, model matching theory, H∞ optimal control theory,
Alternatively, generalized predictive control or the like can be applied. Here, it should be noted that equation (16) and (1)
Since the parameters a _{ij to} h _{ij in the} equation 7) can be easily converted from the time constant and the gain, when the above control method is applied using these parameters, the time constant and the gain are changed. ((6) is used to (12), the time constant and the gain when the value of u _E differs greatly changes), a point which is to reflect the easy control of this change. Equations (15) to (17) are:
Since it is derived from the process model of 2), it can be configured for water quality that cannot always be measured. For this reason,
Consideration of unmeasurable water quality can be reflected in the operation amount.

In this way, the optimum value of the correction value Δu can be calculated by the feedback-type correction operation means, and finally the operation amount u can be obtained as u = u _E + Δu (18) it can.

According to the present embodiment, even when data on the water quality and operation amount is not sufficiently accumulated in the sewage treatment plant, advanced sewage treatment such as acetic acid injection and PAC injection, for example, removal of nitrogen and phosphorus. A process model that simulates the temporal dynamics of a real sewage treatment process incorporating an indispensable actuator can be constructed. Then, with such a process model, it is possible to determine an optimal operation amount incorporating the characteristics of the temporal dynamics of the treated water quality with respect to the operation amount change in the sewage treatment process. In determining the optimal amount of operation, consideration can also be given to water quality items that cannot be measured.In addition, the inflow sewage water quality, the amount of inflow sewage water, or the temporal change of the treated water Changes in dynamics can be easily reflected in determining the optimal manipulated variable.

Third Embodiment Next, a third embodiment of the present invention will be described with reference to FIG. The third embodiment shown in FIG.
In addition to the embodiment, a process monitoring means 5 for monitoring a process is provided between the sewage treatment process model identifying means 3 and the optimum manipulated variable determining means 4.

First, the operation of the process model identification means 3 is substantially the same as in the first or second embodiment. Next, while inputting the inflow sewage amount, the inflow sewage water quality, and the water temperature, which are disturbances, to the process model constructed by the process model identification means 3, and inputting the current manipulated variables, the process model of the process model identification means 3 The treated water data (output data) is displayed on the process monitoring means 4. Here, the time series data may be simply displayed as a graph, but in order to make it more visually appealing,
Do the following:

The water quality of the sewage treatment is composed of floating substances mainly composed of microorganisms, which are constituents of sludge, and soluble substances which are dissolved in water. Therefore, for example, a floating substance (microorganism) is expressed as a set of color-coded circular particles. The density and the particle radius are expressed in proportion to each other. Thus, when the concentration of the floating substance is high, it is represented by large particles, and when the concentration is low, it is represented by small particles. By doing so, the concentration of the floating substance can be expressed quantitatively and visually appealing.

For the soluble water quality, a model such as a sedimentation basin or an aeration tank is displayed, and the soluble water quality is represented by the color of the liquid in the model. Then, the correspondence between the concentration of the soluble water quality and the shade of the color is provided. In this way, the concentration of soluble water quality can be expressed quantitatively and visually. This conceptual diagram is shown in FIG. Also,
In order to represent time series data of water quality, the above principle can be achieved by displaying it as an animation.

By presenting the process monitoring means 5 having such a configuration to the operator of the sewage treatment plant, the operator of the sewage treatment plant inputs quantitative changes in the quality of the treated water into the head as visual information. can do.
Then, based on the visual information, the operator determines the optimal operation amount.

Here, the important point different from the conventional driving based on intuition and experience is that the visual information obtained by the operator is quantitative information on the change in the quality of the treated water, and the operation determined by the operator to be optimal. The result of the quantity is also fed back as visual information of the operator through the process monitoring means 5.

Then, based on the fed-back information, the operator can learn a new and better operation amount determination method by himself. That is, the control system shown in FIG. 7 is a feedback control system with a human system interposed. If the operator can properly learn the information obtained through vision and determine the optimal operation amount, it can be expected that the man-machine cooperative optimal feedback control system will be self-organized.

If the driver cannot continue driving for some reason, such as when the operator is replaced or retires, the driving method learned up to that point is constructed as, for example, an expert system. This makes it possible to construct a fully automated control system.

As described above, according to the present embodiment, quantitative information is given to the operator, and the judgment is left to the operator.
A method in which the result is reflected quantitatively can be used, and a control system having both quantification of the result and flexibility in judgment can be constructed.

[0109]

According to the present invention, in the sewage treatment, especially in the advanced sewage treatment for the purpose of removing nitrogen and phosphorus, the characteristics of the water quality process in the sewage treatment plant can be sufficiently grasped, and the water quality can be controlled accurately. .

[Brief description of the drawings]

FIG. 1 is an overall schematic diagram showing a water quality control device of a sewage treatment plant according to the present invention.

FIG. 2 is a diagram illustrating a process model identification unit.

FIG. 3 is a diagram showing an example of input / output correlation by a process characteristic analysis unit.

FIG. 4 is a diagram showing a method for detecting non-linearity by a process characteristic analyzing means.

FIG. 5 is a diagram showing an example of a structure determined by a structure identification unit.

FIG. 6 is a diagram showing a method of determining a representative value when a plurality of water quality sensors are installed.

FIG. 7 is an overall schematic diagram showing a control device of a sewage treatment plant when a human system is interposed by adding a process monitoring means.

FIG. 8 is a diagram showing a display method of a process monitoring unit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Sewerage process 2 Process value storage means 3 Process model identification means 4 Optimum operation amount determination means 5 Process monitoring means 11 First sedimentation tank 12 Anaerobic tank 13 Oxygen tank 14 Aerobic tank 15 Final sedimentation tank 16 Disturbance sensor 171-175 Water quality state Sensor 19 Water temperature sensor 31 Process characteristic analysis means 32 Structure identification means 33 Parameter identification means 181 Acetic acid input machine 182 PAC injection machine 183 Circulation pump 184 Blower 185 Return pump 186 Excess sludge removal pump

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Masahiko Tsutsumi 1 Toshiba-cho, Fuchu-shi, Tokyo Inside the Toshiba Fuchu Works Co., Ltd. (72) Inventor Yasuhiko Nagamori 1-Toshiba-cho Fuchu-shi, Tokyo Fuchu Toshiba Corporation In-house (72) Inventor Yukio Hatsuka 1-1-1, Shibaura, Minato-ku, Tokyo F-term in Toshiba head office (reference) 4D028 AA08 AC01 AC06 AC09 BC18 BD16 CA00 CA09 CA11 CA12 CB03 CC01 CC04 CC14 CD01 CE02 CE03

Claims (7)

[Claims] 1. A water quality control device for a sewage treatment plant having a plurality of actuators and treating inflow sewage, comprising: an inflow sewage data sensor for measuring inflow sewage data relating to the inflow sewage; Process data sensor for measuring the process water data related to the process, process value storage means for storing the inflow sewage data from the inflow sewage data sensor, the process water data from the process water data sensor and the operation amount of the actuator, and process value storage Based on the information from the means, a sewage treatment process model identification means for constructing a process model and a process model from the sewage treatment process model identification means are used to determine an optimal operation amount of the actuator based on inflow sewage data and treated water data. Sewage characterized by comprising an optimum operation amount determining means to be obtained Management field of water quality control equipment. 2. A wastewater treatment process model identification means comprising: a process characteristic analysis means for analyzing characteristics of a wastewater treatment process from information recorded in a process value recording means; and a structure of a process model based on an analysis result by the process characteristic analysis means. And a parameter identification unit for estimating a parameter of the process model determined by the structure identification unit.
The water quality control device of the sewage treatment plant described. 3. A sewage treatment process model identification means for constructing a process model in advance by a physicochemical method and for feeding a substance serving as a nutrient source for microorganisms and an actuator for feeding a substance having a coagulation sedimentation effect. 2. The water quality control device for a sewage treatment plant according to claim 1, further comprising a process model enhancing means for further enhancing the process model based on the correlation with the process model. 4. A control parameter determining means for determining parameters related to the calculation of the optimum manipulated variable from a step response waveform of the model obtained by the sewage treatment process model identifying means, and a control parameter determining means. 2. The water quality control device for a sewage treatment plant according to claim 1, further comprising an operation amount command value calculating unit that actually calculates the operation amount command value using the parameter determined by the determination unit. 5. An optimum operation amount determining means for canceling an influence of a disturbance of a model obtained by the sewage treatment process model identifying means and determining an operation amount so as to achieve a predetermined target water quality. And a feedforward-type corrected manipulated variable determining means for correcting the manipulated variable so as to correct an error between the sewage treatment water quality and the target water quality to obtain an optimal manipulated variable. Item 1. A water quality control device for a sewage treatment plant according to item 1. 6. A plurality of inflow sewage data sensors of the same type are provided, and the optimum manipulated variable determining means includes data preprocessing means for averaging information from a plurality of inflow sewage data sensors of the same type. The water quality control device for a sewage treatment plant according to claim 1. 7. A sewage treatment plant water quality control device having a plurality of actuators for treating inflow sewage, comprising: an inflow sewage data sensor for measuring inflow sewage data relating to inflow sewage; Process data sensor for measuring the process water data related to the process, process value storage means for storing the inflow sewage data from the inflow sewage data sensor, the process water data from the process water data sensor and the operation amount of the actuator, and process value storage Sewage treatment process model identification means for constructing a process model based on information from the means, and prediction of treated water data based on inflow sewage data and operation amount of an actuator using a process model from the sewage treatment process model identification means Process monitoring means for obtaining and displaying a value. Sewage treatment plants of water quality control device that.