JP7325320B2 - LEARNING MODEL GENERATOR AND ESTIMATION DEVICE - Google Patents

Description

TECHNICAL FIELD The present invention relates to a process for producing dewatered sludge by injecting a flocculant into sludge and dehydrating it.

In wastewater treatment facilities such as sewage treatment plants, there is a process (dehydration treatment) in which a coagulant is injected into suspension or sludge to form flocculated flocs, and then the sludge is dehydrated using a dehydrator to generate dehydrated sludge. It is done.

The condition of the dehydrated sludge (eg, water content) can vary depending on the condition of the sludge, the injection rate of the flocculant, the operating parameters of the dehydrator, and the like. Patent Document 1 below discloses a technique for predicting the state of dehydrated sludge (turbidity residue) obtained by dehydration treatment under current conditions. Specifically, it was constructed based on suspension data indicating the state of sludge, operation data regarding the operation of the dehydrator, and turbidity residual data indicating the state of the turbidity residual obtained in the dehydration process. A technique is disclosed for outputting turbidity residual data by inputting current values of suspension data and current values of operating parameters into a calculation model.

JP 2019-51458 A

By the way, it is known that when precipitation such as rainwater flows into the sewage flowing into the sewage treatment plant, the properties of the raw water change and the sludge generated at the sewage treatment plant becomes sludge that is easily dewatered. For this reason, if the operating parameters of the dehydrator and the injection rate of the flocculant are the same, the dehydrated sludge produced from sewage into which precipitation has flowed has a higher moisture content than the dehydrated sludge produced from sewage into which precipitation has not flowed. decreases. In other words, the moisture content of the dehydrated sludge becomes unstable depending on the presence or absence of inflow of precipitation.

This destabilization of water content becomes a problem when dehydrated sludge is incinerated or melted in a post-stage facility of a sewage treatment plant. That is, since the moisture content of the dehydrated sludge is not stable, the operation parameters (for example, furnace temperature and flue gas temperature) of the downstream facilities become unstable. Destabilization of this parameter causes malfunctions such as furnace failure.

However, simply measuring the solid concentration and pH of the sludge before dehydration does not accurately capture changes in the properties of the sludge. had to be controlled by feedback.

An object of one aspect of the present invention is to realize generation of dehydrated sludge having a substantially constant water content regardless of the presence or absence of precipitation.

In order to solve the above problems, a learning model generation device according to an aspect of the present invention is generated by injecting a flocculant into sludge generated at a sewage treatment plant into which sewage flows from a sewage system to dehydrate the sludge, The injection rate of the coagulant injected into the sludge when the moisture content of the dehydrated sludge that is incinerated or melted after generation falls within a predetermined range, and the region where precipitation flows into the sewage system during a predetermined period before the injection. A learning unit that generates a learning model that outputs the injection rate with the rain information in the target period as input by performing machine learning using the rain information including the amount of rain as teacher data.

Sewage into which precipitation has flowed is easier to dehydrate than sewage into which precipitation has not flowed. Therefore, if the dehydration conditions such as the injection rate of the coagulant are the same, the water content of the dewatered sludge produced from sewage into which precipitation has flowed is higher than the water content of the dewatered sludge produced from sewage into which precipitation has not flowed. descend. Here, according to the above configuration, a learning model is generated by machine learning using the injection rate and the rainfall information as teacher data. It is possible to estimate the flocculant injection rate at which the rate is within a given range of values. As a result, dehydrated sludge having a substantially constant water content can be produced regardless of the presence or absence of precipitation. As a result, the temperature inside the furnace that incinerates or melts the dehydrated sludge in the downstream facility of the sewage treatment plant can be kept constant, stabilizing the operation of the furnace and preventing problems such as breakdowns. be able to.

In order to solve the above problems, an estimating device according to an aspect of the present invention is generated by injecting a flocculant into sludge generated at a sewage treatment plant into which sewage from a sewage system flows and dehydrating the sludge. An estimating device for estimating the injection rate of a flocculant to be injected into the sludge in order to generate dewatered sludge to be incinerated or melted, wherein the sludge when the water content of the dewatered sludge is a value within a predetermined range. A learning model generated by performing machine learning using the injection rate of the coagulant injected into the sewage system and precipitation information including the amount of precipitation in the area where precipitation flows into the sewage system for a predetermined period before the injection as teacher data. and an estimating unit that estimates the injection rate using the rain information for the target period as an input.

Sewage into which precipitation has flowed is easier to dehydrate than sewage into which precipitation has not flowed. Therefore, if the dehydration conditions such as the injection rate of the coagulant are the same, the water content of the dewatered sludge produced from sewage into which precipitation has flowed is higher than the water content of the dewatered sludge produced from sewage into which precipitation has not flowed. descend. Here, according to the above configuration, using a learning model machine-learned using the injection rate and the rainfall information as teacher data, based on the rainfall information, the moisture content of the dewatered sludge is a value within a predetermined range. The infusion rate of the agent can be estimated. As a result, dehydrated sludge having a substantially constant water content can be produced regardless of the presence or absence of precipitation. As a result, the temperature inside the furnace that incinerates or melts the dehydrated sludge in the downstream facility of the sewage treatment plant can be kept constant, stabilizing the operation of the furnace and preventing problems such as breakdowns. be able to.

Further, in the estimating device according to an aspect of the present invention, the predetermined range of the moisture content is set when the amount of precipitation in the area where precipitation flows into the sewage system during the predetermined period before the injection of the coagulant is 0 It is preferably determined based on the water content of the dehydrated sludge produced in the above.

According to the above configuration, dewatered sludge having a water content within a predetermined range determined based on the water content of the dewatered sludge generated when the weather is fine (when the amount of precipitation is 0) is generated. Thereby, the temperature in the furnace for incinerating or melting the dewatered sludge in the downstream facility of the sewage treatment plant is set to the temperature for incinerating or melting the dehydrated sludge generated when the weather is fine during the predetermined period. By doing so, the frequency of subsequent temperature adjustments can be reduced.

In order to solve the above problems, a learning model generation device according to an aspect of the present invention is generated by injecting a flocculant into sludge generated at a sewage treatment plant into which sewage flows from a sewage system to dehydrate the sludge, Water content of dewatered sludge to be incinerated or melted after generation, injection rate of coagulant injected into the sludge, and precipitation information including precipitation in the region where precipitation flows into the sewage system for a predetermined period before the injection; and is used as teacher data to generate a learning model that receives the precipitation information and the injection rate in the target period and outputs the moisture content.

Sewage into which precipitation has flowed is easier to dehydrate than sewage into which precipitation has not flowed. Therefore, if the dehydration conditions such as the injection rate of the coagulant are the same, the water content of the dewatered sludge produced from sewage into which precipitation has flowed is higher than the water content of the dewatered sludge produced from sewage into which precipitation has not flowed. descend. Here, according to the above configuration, a learning model is generated by machine learning using the moisture content rate, the injection rate, and the rainfall information as teacher data. Based on this, it is possible to estimate the water content of the dewatered sludge when the coagulant is injected at the injection rate. As a result, if the actual injection rate of the coagulant is determined according to the estimated water content, it is possible to produce dehydrated sludge having a substantially constant water content regardless of the presence or absence of precipitation. As a result, the temperature inside the furnace that incinerates or melts the dehydrated sludge in the downstream facility of the sewage treatment plant can be kept constant, stabilizing the operation of the furnace and preventing problems such as breakdowns. be able to.

In order to solve the above problems, an estimating device according to an aspect of the present invention is generated by injecting a flocculant into sludge generated at a sewage treatment plant into which sewage from a sewage system flows and dehydrating the sludge. An estimating device for estimating the water content of dewatered sludge to be incinerated or melted, comprising: the water content of the dewatered sludge, the injection rate of the coagulant injected into the sludge, and the amount of water in the sewage system during a predetermined period before the injection. Using a learning model generated by performing machine learning using rainfall information including the amount of rainfall in an area where rainfall flows in as teacher data, the rainfall information and the predetermined injection rate in the target period are input, An estimation unit for estimating the moisture content is provided.

Sewage into which precipitation has flowed is easier to dehydrate than sewage into which precipitation has not flowed. Therefore, if the dehydration conditions such as the injection rate of the coagulant are the same, the water content of the dewatered sludge produced from sewage into which precipitation has flowed is higher than the water content of the dewatered sludge produced from sewage into which precipitation has not flowed. descend. Here, according to the above configuration, using a learning model machine-learned with the moisture content, the injection rate, and the rainfall information as teacher data, based on the injection rate and the rainfall information, the coagulant at the injection rate It is possible to estimate the moisture content of the dewatered sludge when As a result, if the actual injection rate of the coagulant is determined according to the estimated water content, it is possible to produce dehydrated sludge having a substantially constant water content regardless of the presence or absence of precipitation. As a result, the temperature inside the furnace that incinerates or melts the dehydrated sludge in the downstream facility of the sewage treatment plant can be maintained at a constant temperature, stabilizing the operation of the furnace and preventing troubles such as breakdowns. be able to.

Further, the estimating device according to an aspect of the present invention further includes a determination unit that determines the injection rate to be injected into the sludge based on the difference between the estimated water content and a target water content. is preferred.

According to the above configuration, it is possible to automatically determine the injection rate when the inputted injection rate is not suitable for producing dehydrated sludge having a target moisture content. As a result, the moisture content of the dehydrated sludge can be kept substantially constant. This makes it possible to maintain a constant temperature inside the furnace, which incinerates or melts dehydrated sludge in the downstream facility of the sewage treatment plant, stabilizes the operation of the furnace, and prevents malfunctions such as breakdowns. be able to.

Further, in the estimating device according to one aspect of the present invention, the target moisture content is set when the amount of precipitation in the area where precipitation flows into the sewage system in the predetermined period before the injection of the coagulant is 0 It is preferable to determine based on the moisture content of the generated dehydrated sludge.

According to the above configuration, dewatered sludge having a water content within a predetermined range determined based on the water content of the dewatered sludge generated when the weather is fine (when the amount of precipitation is 0) is generated. Thereby, the temperature in the furnace for incinerating or melting the dewatered sludge in the downstream facility of the sewage treatment plant is set to the temperature for incinerating or melting the dehydrated sludge generated when the weather is fine during the predetermined period. By doing so, the frequency of subsequent temperature adjustments can be reduced.

Further, in the learning model generation device according to the aspect of the present invention, the amount of precipitation included in the precipitation information as the teacher data and the precipitation information input to the estimating unit is the area where precipitation flows into the sewage system. is the total amount of precipitation from the time of injection of the coagulant to the past by the time required for the sewage to reach the sewage treatment plant from the area.

According to the above configuration, rainfall information including the amount of rainfall at the time of rainfall in each region is used as training data, so highly accurate estimation is realized considering the time lag of arrival from each region to the sewage treatment plant. be able to.

Further, in the estimation device according to the aspect of the present invention, the amount of precipitation included in the precipitation information as the training data and the precipitation information input to the estimation unit is determined for each of the areas where precipitation flows into the sewage system. and is the total amount of precipitation from the injection of the coagulant to the past by the time required for the sewage to reach the sewage treatment plant from the area.

According to the above configuration, the precipitation information including the amount of precipitation at the time of precipitation in each area is used as the teacher data and the precipitation information to be input to the estimating unit. It is possible to realize a highly accurate estimation that takes into consideration.

The learning model generating device and the estimating device according to each aspect of the present invention may be realized by a computer. In this case, the computer is operated as each part (software element) included in the learning model generating device and the estimating device. Accordingly, a control program for the learning model generating device and the estimating device, which causes the learning model generating device and the estimating device to be implemented by a computer, and a computer-readable recording medium recording it are also included in the scope of the present invention.

ADVANTAGE OF THE INVENTION According to one aspect of the present invention, dewatered sludge having a substantially constant water content can be produced regardless of the presence or absence of precipitation.

BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram which shows one structural example of the sewage treatment plant which concerns on this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a functional block diagram which shows an example of a principal part structure of the chemical-feeding-rate estimation system which concerns on Embodiment 1 of this invention. 3 is a diagram showing a specific example of teacher data stored in the storage device shown in FIG. 2; FIG. FIG. 2 is a schematic diagram showing an example of a sewer connected to the sewage treatment plant shown in FIG. 1; 4 is a diagram showing a calculation example of total precipitation included in the teacher data shown in FIG. 3; FIG. FIG. 4 is a table showing an example of changes in moisture content of dehydrated cake when dehydrator parameters and chemical dosing rates are changed, included in the teacher data shown in FIG. 3 ; FIG. 4 is a diagram showing an example of the relationship between the dehydrator parameters and the chemical feeding rate, and the moisture content of the dehydrated cake, which are included in the teacher data shown in FIG. 3. FIG. 3 is a flowchart showing an example of the flow of learning model generation processing executed by the learning model generation device shown in FIG. 2; It is a flow chart which shows an example of the flow of the chemical-feeding-rate estimation process which the estimation apparatus shown in FIG. 2 performs. It is a functional block diagram which shows an example of a principal part structure of the chemical-feeding-rate estimation system which concerns on Embodiment 2 of this invention. 11 is a diagram showing a specific example of teacher data stored in the storage device shown in FIG. 10; FIG. 11 is a flow chart showing an example of the flow of learning model generation processing executed by the learning model generation device shown in FIG. 10; 11 is a flow chart showing an example of the flow of chemical injection rate estimation processing executed by the estimation device shown in FIG. 10;

[Embodiment 1]
<Summary of Sewage Treatment Plant>
FIG. 1 is a schematic diagram showing a configuration example of a sewage treatment plant 90 according to this embodiment. As shown in FIG. 1, the sewage treatment plant 90 comprises a settling basin 91, a primary sedimentation basin 92, a reaction tank 93, a final sedimentation basin 94, a gravity thickener 95, a mechanical thickener 96, a dehydrator 97 and an incinerator 98. ing. In addition, the equipment provided in the sewage treatment plant 90 is not limited to this example.

Sewage flows into the sewage treatment plant 90 from the sewage system. In addition, the sewage system according to each embodiment of the present disclosure may be of a combined type or a separate type.

The settling basin 91 is a facility for removing readily sedimentary matter such as sand in the sewage that has flowed into the sewage treatment plant 90 (hereinafter sometimes referred to as inflow sewage). The influent sewage discharged from the settling basin 91 is first sent to the settling basin 92 .

The primary sedimentation tank 92 is equipment for separating and removing sedimentary organic matter in the inflow sewage. The influent sewage from which sedimentary organic matter has been removed by the first sedimentation tank 92 is sent to the reaction tank 93 . Also, settled organic matter removed from the influent sewage is sent to the gravity thickener 95 .

The reaction tank 93 is a facility for biologically treating contaminants in inflowing sewage. As an example, in the reaction tank 93, pollutants are treated by using the activated sludge method as food for microorganisms. Activated sludge containing microorganisms grown by the activated sludge method is sent to a final sedimentation tank 94 . Note that the reaction tank 93 is also called an aeration tank.

The final sedimentation tank 94 is equipment for solid-liquid separation of the activated sludge. The treated water from which the activated sludge has been separated is discharged to rivers or the like through subsequent equipment (filtration equipment, disinfection equipment, advanced treatment equipment, etc.) (not shown) or as it is. On the other hand, part of the sedimented and separated activated sludge is sent to the reaction tank 93 again, and the rest is sent to the mechanical thickener 96 .

The gravity concentration tank 95 is a facility for concentrating the sedimented organic matter removed in the first sedimentation tank 92 into solids by the difference in specific gravity and gravity. The sedimentary organic matter from which the solid matter has been concentrated by the gravity concentration tank 95 is sent to the dehydrator 97 .

The mechanical thickener 96 is equipment for thickening the sedimentation sludge sent from the final sedimentation basin 94 . The mechanical thickener 96 thickens the activated sludge using, for example, gravitational methods, centrifugal methods, or filtration methods. The sedimentation sludge thickened by the mechanical thickener 96 is sent to the dehydrator 97 . Hereinafter, concentrated sedimentary organic matter and sedimentation sludge are collectively referred to as concentrated sludge.

The dehydrator 97 is equipment for further dehydrating the thickened sludge. As an example, the dehydration treatment is performed by injecting a coagulant into thickened sludge to form flocs and then dehydrating the flocs. Specifically, flocs are formed by injecting a flocculant into thickened sludge fed into a flocculator (also referred to as a flocculator) and agitating the thickened sludge. The thickened sludge in which flocs are formed is put into a dehydrator, and the water is separated and removed by the dehydrator. That is, the dehydrator 97 according to each embodiment of the present disclosure is a generic term for a flocculator and a dehydrator. The dewatered thickened sludge is sent to the incinerator 98 . Hereinafter, dehydrated sludge, which is dewatered thickened sludge, will be referred to as dehydrated sludge.

Although the dehydrator included in the dehydrator 97 according to the present disclosure is described as being a screw press dehydrator, it is not so limited. By way of example, the dehydrator may be a centrifugal dehydrator, a belt press dehydrator or a filter press dehydrator.

The incinerator 98 is equipment for incinerating the cake. In place of the incinerator 98, the sewage treatment plant 90 may be equipped with a melting furnace that melts the cake. Also, the incinerator 98 or melting furnace may be provided in a facility separate from the sewage treatment plant 90 .
<Main configuration of the chemical dosing rate estimation system>
FIG. 2 is a functional block diagram showing an example of the main configuration of the chemical-feeding rate estimation system 100. As shown in FIG. The chemical dosing rate estimation system 100 is a system for estimating the dosing rate (hereinafter referred to as chemical dosing rate) of a coagulant injected into thickened sludge in the dehydrator 97 of the sewage treatment plant 90 . Specifically, the chemical addition rate estimation system 100 is a system for estimating the chemical addition rate in which the moisture content of the cake produced by the dehydration process is within an appropriate range.

The dosing rate estimation system 100 includes a learning model generation device 1 , an estimation device 2 , an input device 3 , an output device 4 and a storage device 5 .

The input device 3 receives an input from the user and outputs an input signal based on the input to the learning model generation device 1 or the estimation device 2 .

The output device 4 outputs the information generated by the estimation device 2 . The output method by the output device 4 is not particularly limited. The output device 4 may be, for example, a display device that displays the information as an image, or an audio output device that outputs the information as audio. Also, the output device 4 may be a control signal output device that outputs the information as an electric signal to a coagulant injection device (not shown).

Storage device 5 stores programs and data used in chemical injection rate estimation system 100 . The data held (stored) by the storage device 5 will be described later.

(Principal configuration of learning model generation device 1)
The learning model generation device 1 generates a learning model for estimating the chemical dosing rate to thickened sludge. The learning model generation device 1 includes a control section 10 . The control unit 10 controls all the units of the learning model generation device 1, and is realized by a processor and a memory, for example. In this example, the processor accesses storage (not shown), loads a program (not shown) stored in the storage into memory, and executes a series of instructions contained in the program. Each unit of the control unit 10 is thus configured.

As such units, the control unit 10 includes a teacher data acquisition unit 101 and a learning unit 102 .

The teacher data acquisition unit 101 acquires a teacher data group 51 based on an input signal from the input device 3, specifically, an input signal indicating an instruction to start learning (hereinafter referred to as a learning start signal). Specifically, the teacher data acquiring unit 101 reads the teacher data group 51 from the storage device 5 upon acquiring the learning start signal from the input device 3 .

(Details of teacher data group 51)
FIG. 3 is a diagram showing a specific example of the training data group 51. As shown in FIG. In FIG. 3, the teacher data group 51 includes a plurality of teacher data (data sets), each of which consists of total precipitation (precipitation information), dehydrator parameters, chemical dosing rate and cake moisture content.

Note that the teaching data group 51 according to each embodiment of the present disclosure is described as being generated by a teaching data generation device (not shown) and stored in the storage device 5, but is not limited to this example. For example, the learning model generation device 1 may generate the teacher data group 51 .

The chemical dosing rate and the dehydrator parameter are the chemical dosing rate to the thickened sludge generated from the inflowing sewage and the parameters for the operation of the dehydrator 97 during dehydration of the thickened sludge. In other words, the chemical dosing rate and the dehydrator parameters are parameters relating to the dosing rate and the operation of the dehydrator 97 in the dehydration treatment of thickened sludge produced from inflowing sewage. The moisture content of the cake is the mass ratio of moisture contained in the cake produced by the dehydration treatment, that is, "(mass of moisture in cake)/(mass of cake)". In addition, the chemical feeding rate according to each embodiment of the present disclosure is assumed to be "(concentration of dissolved coagulant×injection amount of coagulant)/(concentration of thickened sludge×supply amount of thickened sludge)".

Further, in each embodiment of the present disclosure, the dehydrator 97 consists of a flocculator and a screw press dehydrator, so the dehydrator parameters are, as shown in FIG. Including pressure. Details of these will be described later.

As an example, the worker of the sewage treatment plant 90 inputs the moisture content of the produced cake, the chemical dosing rate and the dehydrator parameters in the production of the cake into the teaching data generation device after finishing the dehydration treatment of the thickened sludge. . As a result, the chemical dosing rate, dehydrator parameter, and cake moisture content in the dehydration process are associated with each other.

The total amount of precipitation is the total amount of precipitation in each region where precipitation flows into the sewage system before the inflowing sewage flows into the sewage treatment plant 90, that is, during the period (predetermined period) during which the inflowing sewage flows through the sewage system. In each embodiment of the present disclosure, the total amount of precipitation is described as the total amount of precipitation per hour, but the definition of the total amount of precipitation is not limited to this example. For example, total precipitation may be the total amount of precipitation per day.

(details of total precipitation)
FIG. 4 is a schematic diagram showing an example of a sewage system 99 connected to the sewage treatment plant 90, and shows areas where precipitation flows into the sewage system 99. As shown in FIG. An area where precipitation flows into the sewage system is specified by being divided into meshes of 1 km square, for example. That is, the total precipitation amount is the total precipitation amount in the identified 1-km square mesh.

In addition, in FIG. 4, in order to make the explanation easier to understand, it is assumed that there are three regions (regions R1, R2, and R3) where precipitation flows into the sewage system 99 . As an example, the training data generation device communicates with a weather information server (not shown) based on an operator's operation input, and acquires hourly rainfall amounts in areas R1, R2, and R3. The weather information server is a server of the Japan Meteorological Agency or a company that provides weather forecasts.

FIG. 5 is a diagram showing an example of calculation of total precipitation. The example of FIG. 5 shows a calculation example of the total amount of precipitation contained in the sewage that flowed into the sewage treatment plant 90 at 10:30 on a certain day. The time when the sewage was flowing through regions R1, R2 and R3 is before 10:30. In other words, there is a time lag between the time the sewage entered the sewage treatment plant 90 and the time it flowed through regions R1, R2 and R3. The training data generation device calculates the total rainfall in consideration of this time lag.

Specifically, in each of the areas R1, R2 and R3, the time required for the sewage to reach the sewage treatment plant 90 from each area (that is, the time lag) is specified in advance. In the example of FIG. 5, the time is 10 minutes for region R1, 20 minutes for region R2, and 15 minutes for region R3.

The training data generation device calculates the amount of rainfall in each region from 10:30, the time when sewage flowed into the sewage treatment plant 90, going back in time by the time lag. In FIG. 5, this point in time is referred to as "precipitation confirmation time". In the example of FIG. 5, the training data generation device acquires the amount of precipitation in region R1 at 10:20. Also, the training data generation device acquires the amount of rainfall in region R2 at 10:10. Also, the training data generation device acquires the amount of precipitation in region R3 at 10:15. In FIG. 5, these precipitation amounts are referred to as "regional precipitation amounts". In the example of FIG. 5, the regional precipitation amount for region R1 is 6 (mm/h), the regional precipitation amount for region R2 is 5 (mm/h), and the regional precipitation amount for region R3 is 4 (mm/h). was assumed to be

The training data generation device sums up the obtained regional rainfall amounts to calculate the total rainfall amount. In the example of FIG. 5, the total precipitation is 6+5+4=15 (mm/h).

The training data generation device associates the calculated total rainfall with the chemical dosing rate, dehydrator parameter, and cake moisture content obtained as a result of the dehydration treatment of the sewage that flowed in at 10:30. As a result, teacher data included in the teacher data group 51 shown in FIG. 3 is generated.

Note that the training data group 51 includes only training data relating to dehydration processes in which the moisture content of the cake was within the proper range (predetermined range). In the example of FIG. 3, the appropriate range is 74 to 80%, but the appropriate range of cake moisture content is not limited to this example. In addition, the appropriate range of the moisture content of the cake is determined, for example, based on the moisture content of the cake produced in fine weather. In other words, the appropriate range is determined based on the moisture content of the cake produced when the total amount of precipitation calculated by the method described above is zero.

As an example, the teacher data generation device determines whether the moisture content of the cake in the generated teacher data is within the appropriate range each time the dehydration process is completed, and determines that the moisture content of the cake is within the appropriate range. The obtained teacher data is added to the teacher data group 51, and the teacher data whose cake water content is determined to be out of the proper range is discarded. As another example, an operator of the sewage treatment plant 90 may determine whether the moisture content of the cake in the training data is within the appropriate range. Specifically, the operator may input a combination of chemical feeding rate, dehydrator parameter, and total rainfall amount related to the dehydration process in which the moisture content of the cake is within the appropriate range, to the teaching data generation device. Alternatively, the worker inputs the chemical dosing rate and dehydrator parameters into the teacher data generation device, and the teacher data generation device calculates the total precipitation corresponding to the input chemical dosing rate and dehydrator parameters. good too.

In the example of FIG. 3, the teacher data includes the moisture content of the cake in order to make the explanation easier to understand, but the teacher data may not include the moisture content of the cake.

(details of dehydrator parameters)
As described above, dehydrator parameters according to embodiments of the present disclosure include screw speed, flocculator speed, input pressure and presser pressure.

The number of screw revolutions is the number of revolutions per minute of the screw provided in the screw press dehydrator. The number of rotations of the flocculator is the number of rotations per minute of the agitating part (for example, agitating blade) of the flocculator that agitates the thickened sludge into which the flocculant has been added.

Input pressure is the pressure applied when the thickened sludge is input to the screw press dehydrator. The presser pressure is the pressure (back pressure) exerted by the presser on the cake discharged from the outlet of the screw press dehydrator. The cake is further dewatered by applying presser pressure to the cake.

FIG. 6 is a table showing an example of changes in the moisture content of the dehydrated cake when the dehydrator parameters and the dosing rate are changed. FIG. 7 is a diagram showing an example of the relationship between the dehydrator parameters, the chemical feeding rate, and the moisture content of the dehydrated cake. As shown in the relationship diagram G1 of FIGS. 6 and 7, the moisture content of the cake increases as the screw rotation speed increases. On the other hand, when the screw rotation speed is lowered, the moisture content of the cake is lowered.

Further, as shown in the relationship diagram G2 of FIGS. 6 and 7, when the flocculator rotation speed is set to the optimum point (specifically, the optimum numerical range), the cake water content decreases. On the other hand, if the flocculator rotation speed is set to a value other than the optimum point, the moisture content of the cake increases.

Moreover, as shown in the relational diagram G3 of FIGS. 6 and 7, the moisture content of the cake decreases as the injection pressure is increased. On the other hand, when the injection pressure is lowered, the moisture content of the cake increases.

Further, as shown in the relational diagram G4 of FIGS. 6 and 7, the moisture content of the cake decreases when the presser pressure is increased. On the other hand, decreasing the presser pressure increases the moisture content of the cake.

Moreover, as shown in the relational diagram G5 of FIGS. 6 and 7, the moisture content of the cake decreases when the chemical feeding rate is increased (the amount of coagulant is increased). However, overdosing, ie, too much flocculant, will increase the moisture content of the cake. On the other hand, decreasing the chemical feeding rate (reducing the amount of flocculant) increases the moisture content of the cake.

Thus, by changing (adjusting) each dehydrator parameter and chemical feeding rate of the dehydrator 97, the moisture content of the cake can be changed. In other words, when the chemical charging rate estimation system 100 estimates the chemical charging rate that makes the moisture content of the cake within an appropriate range, the dehydrator parameters are teacher data at the time of learning model generation, and at the time of chemical charging rate estimation It can be said that it is important data as input data for

Returning to the description of the learning model generation device 1 with reference to FIG. 2 again.

The learning unit 102 performs machine learning based on the teacher data group 51 acquired from the teacher data acquisition unit 101 to generate a learning model 52 . The learning unit 102 stores the generated learning model 52 in the storage device 5 . In the teacher data included in the teacher data group 51, the total precipitation and dehydrator parameters are input data, and the chemical feeding rate is correct data. The learning model 52 learns the sewage currently flowing into the sewage treatment plant 90 (hereinafter referred to as current inflowing sewage) using the total amount of precipitation during the period when it was flowing through the sewage system and the current dehydrator parameters as input data. The input to model 52 is the dosing rate. Since the teacher data is data including the total precipitation amount, the dehydrator parameters, and the chemical dosing rate when the moisture content of the cake falls within the appropriate range, the chemical dosing rate output from the learning model 52 is The chemical dosing rate becomes an appropriate range.

Note that the learning unit 102 generates the learning model 52 using a known machine learning method such as a neural network (NN), for example.

(Principal configuration of estimation device 2)
The estimating device 2 estimates the chemical dosing rate to thickened sludge generated from the current influent sewage. The estimating device 2 has a control unit 20 . The control unit 20 controls each unit of the estimation device 2 in an integrated manner, and is realized by a processor and a memory, for example. In this example, the processor accesses storage (not shown), loads a program (not shown) stored in the storage into memory, and executes a series of instructions contained in the program. Each unit of the control unit 20 is thus configured.

As such units, the control unit 20 includes an input data acquisition unit 201 and an estimation unit 202 .

The input data acquisition unit 201 acquires an input signal indicating an instruction to start estimation (hereinafter referred to as an estimation start signal) from the input device 3, and the input data input to the input device 3, that is, the current inflow sewage Get the total amount of precipitation during the time it was in the sewer and the current dehydrator parameters. The total rainfall amount is calculated, for example, by a method similar to that for the total rainfall amount included in the teacher data. Also, these data are input to the input device 3 by, for example, a worker of the sewage treatment plant 90 . The input data acquisition unit 201 outputs the acquired input data to the estimation unit 22 .

Based on the input data and the learning model 52, the estimating unit 202 estimates the chemical dosing rate to thickened sludge generated from the current inflowing sewage. The estimating unit 202 reads the learning model 52 from the storage device 5 after acquiring the input data from the input data acquiring unit 201 . Then, the estimating unit 202 acquires the chemical feeding rate output from the learning model 52 by inputting the input data to the learning model 52, that is, the acquired total precipitation and dehydrator parameters. The chemical dosing rate is a chemical dosing rate at which the moisture content of the cake produced from the current inflowing sewage falls within an appropriate range. The estimation unit 202 outputs the acquired chemical-feeding rate to the output device 4 as the chemical-feeding rate estimation result. When the output device 4 is the control signal output device described above, the coagulant injection device can inject the coagulant into the thickened sludge based on the chemical injection rate.

<Flow of learning model generation processing>
FIG. 8 is a flowchart showing an example of the flow of learning model generation processing executed by the learning model generation device 1. As shown in FIG.

The teacher data acquisition unit 101 acquires a learning start signal from the input device 3 (step S1, hereinafter, description of "step" is omitted), and reads teacher data, that is, a teacher data group 51 from the storage device 5 (S2). . The teacher data acquisition unit 101 outputs the read teacher data group 51 to the learning unit 102 .

The learning unit 102 generates a learning model 52 from the teacher data group 51 acquired from the teacher data acquisition unit 101 (S3). Specifically, the learning unit 102 generates a learning model 52 that receives the total precipitation and dehydrator parameters as inputs and outputs the chemical feeding rate. Then, the learning unit 102 stores the generated learning model 52 in the storage device 5 (S4). With this, the learning model generation processing ends.

(Update of learning model 52)
It is preferable that the learning unit 102 periodically update (automatically update) the learning model 52 . The teacher data group 51 stores teacher data each time the dehydration process is performed so that the moisture content of the cake falls within the appropriate range.

As an example, when the machine learning is executed, the learning unit 102 starts a timer (not shown) and measures the elapsed time from execution of the machine learning. Then, when the elapsed time reaches a predetermined value, the learning unit 102 reads the teacher data group 51 again and executes machine learning (updates the learning model 52).

Note that the method of updating the learning model 52 is the same as the method of generating the learning model 52, so the description will not be repeated here.

<Flow of chemical dosing rate estimation processing>
FIG. 9 is a flowchart showing an example of the flow of the chemical-feeding-rate estimation process executed by the estimation device 2. As shown in FIG.

The input data acquisition unit 201 acquires the estimation start signal, the total precipitation amount, and the dehydrator parameter from the input device 3 (S11). The total precipitation is the total precipitation during the period (target period) in which the current inflowing sewage was flowing through the sewage system, and is the total precipitation in the area where the precipitation flows into the current inflowing sewage, and the dehydrator parameter is the current dehydrator parameter. The input data acquisition unit 201 outputs the acquired total precipitation and dehydrator parameters to the estimation unit 202 .

When the estimation unit 202 acquires the total precipitation amount and the dehydrator parameter from the input data acquisition unit 201, the estimation unit 202 reads out the learning model 52 from the storage device 5 (S12). Then, the estimation unit 202 estimates the chemical feeding rate using the acquired total precipitation amount and dehydrator parameter, and the read learning model 52 (S13). Specifically, the estimating unit 202 acquires the dosing rate output from the learning model 52 by inputting the acquired total precipitation and dehydrator parameters into the learning model 52 . The estimation unit 202 causes the output device 4 to output the chemical injection rate (S14). In other words, the estimation unit 202 causes the output device 4 to output the acquired chemical-feeding rate as the chemical-feeding rate estimation result. With this, the chemical-feeding-rate estimation process ends.

<effect>
As described above, the learning model generation device 1 according to the present embodiment provides the chemical injection rate to the thickened sludge generated from the inflow sewage and the dehydration of the thickened sludge when the moisture content of the cake is within the appropriate range. Machine learning is performed using dehydrator parameters at the time of treatment and the total amount of precipitation during the period when the inflowing sewage was flowing through the sewage system as training data. As a result, the learning model generation device 1 determines the total precipitation amount during the period when the current inflowing sewage was flowing through the sewage system, and the total precipitation in the area where the precipitation flows into the current inflowing sewage Using the current dehydrator parameters as inputs, a learning model 52 is generated that outputs the chemical dosing rate.

Also, the estimation device 2 according to the present embodiment uses the learning model 52 to estimate the chemical feeding rate by inputting the total precipitation corresponding to the current inflowing sewage and the current dehydrator parameters.

Sewage into which precipitation has flowed is easier to dehydrate than sewage into which precipitation has not flowed. Therefore, if the chemical dosing rate is the same, the moisture content of cake produced from sewage into which precipitation has flowed (specifically, thickened sludge produced from the sewage) is lower than the moisture content of the cooked cake.

Here, according to the estimation device 2, using the learning model 52 generated by the learning model generation device 1, the total precipitation corresponding to the current inflowing sewage and the current dehydrator parameters are input, and the chemical feeding rate is estimated. As a result, it is possible to estimate the chemical dosing rate at which the moisture content of the cake produced from the current influent falls within the appropriate range.

As a result, the temperature in the incinerator 98 for incinerating the cake can be kept constant, the operation of the incinerator 98 can be stabilized, and troubles such as failure can be prevented.

As an example, the appropriate range of the moisture content of the cake is based on the moisture content of the cake generated from the influent sewage in which the total precipitation during the period of flowing through the sewer is 0, that is, the influent sewage that was flowing through the sewer during fine weather. Determined. As a result, the moisture content of the produced cake is within an appropriate range regardless of the presence or absence of precipitation. Based on this, the temperature in the incinerator 98 is set to the temperature for incinerating the moisture content of the cake produced from the inflowing sewage that was flowing through the sewage system during fine weather, thereby reducing the frequency of subsequent temperature adjustment. can be lowered.

The total amount of precipitation included in the teacher data and the input data is, for example, the amount of precipitation in each area where precipitation flows into the sewage system, and the amount of time required for the sewage to reach the sewage treatment plant 90 from the area. It is the total amount of precipitation in the past from the inflow of sewage.

Since the total amount of precipitation when it rains in each area is used as teacher data or input data, it is possible to realize highly accurate estimation considering the time lag of arrival from each area to the sewage treatment plant 90.

[Embodiment 2]
Other embodiments of the invention are described below. For convenience of description, members having the same functions as those of the members described in the above embodiments are denoted by the same reference numerals, and description thereof will not be repeated.

The chemical injection rate estimation system according to the present embodiment includes the chemical injection rate to thickened sludge generated from inflowing sewage, dehydrator parameters during dehydration treatment of the thickened sludge, and the period during which the inflowing sewage was flowing through the sewage system. In addition to the total amount of precipitation, machine learning is performed using teacher data including the moisture content of the cake produced by the dehydration treatment. As a result, the chemical dosing rate estimating system according to the present embodiment inputs the total precipitation during the period when the current inflowing sewage was flowing through the sewage system, the current dehydrator parameters, and the planned chemical dosing rate. Generate a learning model that outputs moisture content.

That is, the chemical dosing rate estimation system according to the present embodiment inputs the total precipitation during the period when the current inflowing sewage was flowing through the sewage system, the current dehydrator parameters, and the planned chemical dosing rate. Estimate rate. And the chemical-feeding-rate estimation system which concerns on this embodiment corrects the planned chemical-feeding rate, and outputs it as an estimated value of a chemical-feeding rate, when the estimated moisture content is not within an appropriate range.

<Main configuration of the chemical dosing rate estimation system>
FIG. 10 is a functional block diagram showing an example of the main configuration of the chemical injection rate estimation system 200 according to this embodiment. Hereinafter, with respect to each member of the chemical-feeding rate estimation system 200, only differences from the same-named members of the chemical-feeding rate estimation system 100 will be described.

The chemical-feeding rate estimation system 200 differs from the chemical-feeding rate estimation system 100 in that it includes a learning model generation device 1a, an estimation device 2a, and a storage device 5a.

(Principal Configuration of Learning Model Generation Device 1a)
The control unit 10a of the learning model generation device 1a differs from the learning model generation device 1 in that it includes a learning unit 102a.

The learning unit 102a performs machine learning based on the teacher data group 51a acquired from the teacher data acquisition unit 101, and generates a learning model 52a. The learning unit 102a stores the generated learning model 52a in the storage device 5a. Note that the learning unit 102a generates the learning model 52a using a known machine learning method such as a neural network (NN), for example.

(Details of teacher data group 51a)
FIG. 11 is a diagram showing a specific example of the teacher data group 51a. Only differences between the teacher data group 51a and the teacher data group 51 will be described here.

The teacher data included in the teacher data group 51a differs from the teacher data included in the teacher data group 51 in that it includes the moisture content of the cake as essential data. Also, the teacher data group 51a differs from the teacher data group 51 in that it may include teacher data including cake moisture content outside the appropriate range.

In the teacher data included in the teacher data group 51a, the total precipitation, the dehydrator parameter and the chemical dosing rate are the input data, and the cake water content is the correct data. When the learning model 52a is input with the total precipitation during the period when the current inflowing sewage was flowing through the sewer, the current dehydrator parameters, and the planned chemical dosing rate as input data, the cake moisture content is obtained. is output.

(Principal configuration of estimation device 2a)
The estimating device 2a differs from the estimating device 2 in that it includes an input data acquiring unit 201a, an estimating unit 202a, and a chemical injection rate correcting unit 203 (determining unit).

The input data acquisition unit 201a acquires the input data input to the input device 3, that is, the total precipitation during the period when the current inflow sewage was flowing through the sewage system, the current dehydrator parameters, and the planned chemical feeding rate. to get The input data acquisition unit 201a outputs the acquired input data to the estimation unit 202a.

The estimation unit 202a estimates the moisture content of cake produced from the current influent (specifically, thickened sludge produced from the current influent) based on the input data and the learning model 52a.

The estimating unit 202a reads out the learning model 52a from the storage device 5a upon acquiring the input data from the input data acquiring unit 201a. Then, the estimating unit 202a acquires the cake moisture content output from the learning model 52a by inputting input data to the learning model 52a, that is, the acquired total precipitation amount, dehydrator parameter, and dosing rate. The estimating unit 202 a outputs the acquired cake water content to the chemical addition rate correcting unit 203 .

Based on the difference between the moisture content of the cake estimated by the estimating unit 202a and the appropriate moisture content 53 (target value of moisture content) stored in the storage device 5a, the chemical feeding rate correcting unit 203 uses the estimating device as input data. Modify the dosing rate entered in 2a. In other words, the chemical-feeding-rate correction unit 203 determines the chemical-feeding rate to thickened sludge based on the difference.

The appropriate moisture content 53 is the target moisture content of the cake, and is determined, for example, based on the moisture content of the cake produced in fine weather. The proper moisture content 53 may be a predetermined value (eg, 75%) or a predetermined numerical range (eg, 70-80%). In addition, below, it demonstrates that the proper moisture content 53 which concerns on this embodiment is a predetermined numerical range.

The chemical addition rate correction unit 203 determines whether or not the cake moisture content estimated by the estimation unit 202 a is within the range of the appropriate moisture content 53 . When it is determined that it is within the range of the appropriate water content 53, the chemical feeding rate correction unit 203 outputs the chemical feeding rate input to the estimating device 2a as the input data to the output device 4 as the chemical charging rate estimation result. .

On the other hand, when it is determined that it is not within the range of the appropriate moisture content 53, the chemical feeding rate correction unit 203 corrects the chemical feeding rate input to the estimation device 2a as input data. As an example, the chemical-feeding rate correction unit 203 changes the chemical-feeding rate based on the input chemical-feeding rate. Then, the chemical feeding rate correction unit 203 inputs the changed chemical feeding rate, the total precipitation amount and the dehydrator parameter input as input data to the learning model 52a, and the output cake moisture content (hereinafter referred to as " (referred to as "Modified Cake Moisture Content"). Then, the chemical addition rate correction unit 203 determines whether or not the corrected cake moisture content is within the range of the appropriate moisture content 53 . The chemical feeding rate correction unit 203 repeats these processes until the corrected cake moisture content falls within the range of the proper moisture content 53 . Then, the chemical-feeding-rate correcting unit 203 changes the chemical-feeding rate (hereinafter referred to as "corrected chemical-feeding rate") in which the corrected cake moisture content is within the range of the appropriate moisture content 53, to the chemical-feeding-rate estimation result. is output to the output device 4.

<Flow of learning model generation processing>
FIG. 12 is a flow chart showing an example of the flow of learning model generation processing executed by the learning model generation device 1a.

The teacher data acquiring unit 101 acquires a learning start signal from the input device 3 (step S21), and then reads teacher data, that is, a teacher data group 51a from the storage device 5a (S22). The teacher data acquisition unit 101 outputs the read teacher data group 51a to the learning unit 102a.

The learning unit 102a generates a learning model 52a from the teacher data group 51a acquired from the teacher data acquisition unit 101 (S23). Specifically, the learning unit 102a generates a learning model 52a that outputs the cake moisture content by inputting the total precipitation amount, the dehydrator parameter, and the chemical feeding rate. Then, the learning unit 102a stores the generated learning model 52a in the storage device 5a (S24). With this, the learning model generation processing ends.

Regarding the update of the learning model 52a, the teacher data is stored in the teacher data group 51a each time the dehydration process is performed, regardless of whether the moisture content of the generated cake is within the appropriate range. , different from updating the learning model 52 . Other points are the same as the updating of the learning model 52, so the description will not be repeated here.

<Flow of chemical injection rate estimation processing>
FIG. 13 is a flow chart showing the flow of the chemical injection rate estimation process executed by the estimation device 2a.

The input data acquisition unit 201a acquires the estimated start signal, the total precipitation, the dehydrator parameter, and the chemical feeding rate from the input device 3 (S31). The total precipitation amount is the total precipitation amount during the period when the current influent sewage was flowing through the sewage system. Also, the dehydrator parameter is the current dehydrator parameter. Also, the chemical dosing rate is the scheduled chemical dosing rate of the flocculant to be injected into the thickened sludge generated from the current influent. The input data acquisition unit 201a outputs the acquired total precipitation amount, dehydrator parameter, and chemical feeding rate to the estimation unit 202 .

When the estimation unit 202a acquires the total precipitation, the dehydrator parameter, and the chemical dosing rate from the input data acquisition unit 201a, the estimation unit 202a reads the learning model 52a from the storage device 5a (S32). Then, the estimating unit 202 estimates the moisture content of the cake using the acquired total rainfall amount, dehydrator parameter and chemical feeding rate, and the read learning model 52a (S33). Specifically, the estimating unit 202a acquires the cake moisture content output from the learning model 52a by inputting the acquired total precipitation amount, dehydrator parameter, and chemical feeding rate into the learning model 52a. The estimating unit 202 a outputs the acquired cake water content to the chemical addition rate correcting unit 203 .

After acquiring the moisture content of the cake from the estimating unit 202a, the dosing rate correction unit 203 reads the appropriate moisture content 53 from the storage device 5a and determines whether the acquired moisture content of the cake is within the appropriate range (S34). .

When it is determined that the obtained cake moisture content is not within the appropriate range (NO in S34), the chemical addition rate correcting unit 203 adjusts the chemical addition rate input as the input data so that the cake moisture content is within the appropriate range. Correct (S35). And chemical-feeding-rate correction part 203 outputs the chemical-feeding rate after correction to output device 4 as an estimation result of a chemical-feeding rate. Then, the chemical-feeding-rate estimation process ends.

On the other hand, when it is determined that the obtained cake moisture content is within the appropriate range (YES in S34), the chemical charging rate correcting unit 203 outputs the chemical charging rate input as the input data as the chemical charging rate estimation result. Output to the device 4. Then, the chemical-feeding-rate estimation process ends.

<effect>
As described above, the learning model generation device 1a according to the present embodiment includes the chemical dosing rate to the thickened sludge generated from the inflowing sewage, the dehydrator parameter at the time of the dehydration treatment of the thickened sludge, and the dehydration treatment. Machine learning is performed using the moisture content of the cake obtained and the total amount of precipitation corresponding to the inflowing sewage as training data. As a result, the learning model generation device 1a generates a learning model 52a that outputs the moisture content of the cake with the total precipitation amount corresponding to the current inflowing sewage, the current dehydrator parameters, and the planned chemical dosing rate as inputs. .

In addition, the estimating device 2a according to the present embodiment uses the learning model 52a to input the total precipitation amount corresponding to the current inflowing sewage, the current dehydrator parameters, and the planned chemical dosing rate, and the moisture content of the cake to estimate Then, when the estimated moisture content is outside the numerical range indicated by the appropriate moisture content 53, the estimating device 2a corrects the input chemical dosing rate. The proper moisture content 53 is determined based on the moisture content of the cake produced in fine weather.

According to the estimation device 2a, by using the learning model 52a generated by the learning model generation device 1a, the dehydration treatment of the thickened sludge generated from the current influent is performed with the current dehydration parameters and the planned chemical dosing rate. It is possible to estimate the moisture content of the cake that would have been produced.

Further, according to the estimating device 2a, the chemical feeding rate can be automatically corrected (determined) when the inputted chemical feeding rate is not suitable for producing a cake with a moisture content within the appropriate range. . As a result, the moisture content of the cake can be kept substantially constant regardless of the presence or absence of precipitation. As a result, the temperature in the incinerator 98 can be kept constant, the operation of the incinerator 98 can be stabilized, troubles such as failures can be prevented, and the frequency of temperature adjustment can be reduced.

Moreover, according to the chemical-feeding-rate estimation system 200 which concerns on this embodiment, the teacher data group 51a may contain the teacher data whose cake moisture content is out of the appropriate range. As a result, all teaching data can be included in the teaching data group 51a without judging whether or not the moisture content of the cake is within the appropriate range before storage. As a result, the number of teacher data can be increased, and more accurate estimation can be achieved.

[Modification]
In Embodiments 1 to 3 of the present disclosure, the thickened sludge that has passed through the gravity thickening tank 95 and the mechanical thickener 96 is dehydrated by the dehydrator 97. Sludge that does not pass through the mechanical thickener 96 may also be used.

Further, as described above, the dehydrator included in the dehydrator 97 according to each embodiment of the present disclosure is not limited to the screw press dehydrator. When the dehydrator included in the dehydrator 97 is a dehydrator other than the screw press dehydrator, the dehydrator parameters may be set according to the dehydrator. More specifically, among various parameters corresponding to the dehydrator, a parameter that affects the moisture content of the cake may be used.

For example, if the dehydrator is a centrifugal dehydrator, the dehydrator parameters are, for example, hydraulic pressure, feed sludge flow rate, and differential speed.

Hydraulic force is the pressure applied when the hydraulic motor rotates the outer and inner drums of the centrifugal dehydrator. As the hydraulic pressure increases, the centrifugal effect increases and the moisture content of the cake decreases. On the other hand, decreasing the hydraulic pressure increases the moisture content of the cake.

The feed sludge flow rate is the flow rate of thickened sludge that is fed into the centrifugal dehydrator. If the feed sludge flow rate is reduced, the moisture content of the cake will decrease. On the other hand, increasing the feed sludge flow rate increases the moisture content of the cake.

The differential speed is the difference in rotational speed between the outer drum provided in the centrifugal dehydrator and the inner drum provided with the screw. Increasing the differential speed increases the moisture content of the cake. On the other hand, when the differential speed is lowered, the moisture content of the cake is lowered.

Further, for example, when the dehydrator is a belt press dehydrator, the dehydrator parameters are, for example, the amount of sludge supplied, the tension pressure of the filter cloth, and the running speed of the filter cloth.

The amount of sludge supplied is the amount of thickened sludge fed into the belt press dehydrator. Increasing the amount of sludge supplied increases the moisture content of the cake. On the other hand, when the amount of sludge supplied is decreased, the moisture content of the cake decreases.

The filter cloth tensioning pressure is the pressure applied to the thickened sludge passing between two stretched filter cloths (belts). The moisture content of the cake decreases when the tension pressure of the filter cloth is increased. On the other hand, when the tension pressure of the filter cloth is lowered, the moisture content of the cake increases.

The filter cloth running speed is the moving speed of the belt due to the rotation of the rolls. The moisture content of the cake increases as the running speed of the filter cloth increases. On the other hand, when the running speed of the filter cloth is lowered, the moisture content of the cake is lowered.

In each embodiment of the present disclosure, the appropriate range of moisture content is the influent sewage with a total precipitation of 0 during the period when it was flowing through the sewer, that is, the water content of the cake produced from the influent sewage flowing through the sewer during fine weather. Although it was determined based on the rate, the appropriate range is not limited to this example.

In the dosing rate estimation systems 100 and 200, dehydrator parameters may not be included in the teacher data and input data.

The teacher data according to each embodiment of the present disclosure may include information other than the total amount of precipitation as information on precipitation. For example, the teacher data includes location data indicating areas where precipitation flows into the sewage system 99 (for example, areas R1, R2, and R3 shown in FIGS. 4 and 5), time lags shown in FIG. May include regional precipitation data.

The estimating device 2a according to the second embodiment may be configured to output the estimated cake moisture content to the output device 4 . In the case of this example, the chemical addition rate estimation system 200 can also be expressed as a water content rate estimation system that estimates the cake water content rate. In this example, the control unit 20a of the estimation device 2a does not need to include the chemical injection rate correction unit 203. Also, in this example, if the moisture content of the cake output from the output device 4 is out of the appropriate range, the chemical addition rate corrected by the operator so that the moisture content of the cake to be generated is within the appropriate range, A flocculant may be injected into the thickened sludge.

In Embodiment 1, the learning model generation device 1 and the estimation device 2 may be configured as an integrated device. Similarly, in the second embodiment, the learning model generation device 1a and the estimation device 2a may be integrated.

In the first to third embodiments, the estimating device may receive the learning model generated by the learning model generating device from the learning model generating device and store it in its own storage.

The present invention is not limited to the above-described embodiments, but can be modified in various ways within the scope of the claims, and can be obtained by appropriately combining technical means disclosed in different embodiments. is also included in the technical scope of the present invention. Furthermore, new technical features can be formed by combining the technical means disclosed in each embodiment.

1 learning model generating device 2 estimating device 102, 102a learning unit 202, 202a estimating unit 203 dosing rate correcting unit (determining unit)
51, 51a Teacher data 52, 52a Learning model 53 Appropriate moisture content (target value of moisture content)

Claims (5)

When the moisture content of the dehydrated sludge generated by injecting a flocculant into the sludge generated at a sewage treatment plant into which sewage flows from the sewage system is dehydrated and incinerated or melted after generation falls within a predetermined range, By performing machine learning using the injection rate of the coagulant injected into the sludge and the precipitation information including the amount of precipitation in the area where precipitation flows into the sewage system for a predetermined period before the injection as teacher data, in the target period A learning model generation device including a learning unit that generates a learning model that outputs the injection rate using the rain information as an input. A flocculant that is produced by injecting a flocculant into sludge generated in a sewage treatment plant into which sewage from a sewage system flows and dehydrating the sludge, and that is injected into the sludge in order to produce dehydrated sludge that is incinerated or melted after production. An estimating device for estimating the injection rate of
Injection rate of the coagulant injected into the sludge when the water content of the dewatered sludge falls within a predetermined range, and precipitation, including the amount of precipitation in the area where precipitation flows into the sewage system during a predetermined period before the injection. An estimating device comprising an estimating unit that estimates the injection rate using the rainfall information for the target period as an input using a learning model generated by performing machine learning using the information as teacher data. The predetermined range of the water content is determined based on the water content of the dewatered sludge generated when the precipitation in the area where precipitation flows into the sewage system is 0 in the predetermined period before the injection of the flocculant. 3. The estimating apparatus of claim 2, wherein The amount of precipitation included in the precipitation information as the training data is
The amount of precipitation in each of the areas where precipitation flows into the sewage system, which is the amount of precipitation at the point in time retroactively from the injection of the coagulant by the time required for the sewage to reach the sewage treatment plant from the area. 2. The learning model generation device according to claim 1 , which is the sum of . The amount of precipitation included in the precipitation information as the training data and the precipitation information input to the estimating unit is
The amount of precipitation in each of the areas where precipitation flows into the sewage system, which is the amount of precipitation at the point in time retroactively from the injection of the coagulant by the time required for the sewage to reach the sewage treatment plant from the area. 4. An estimating device according to claim 2 or 3 , wherein the sum of .

Images