JP2024056609A - Self-heating prediction method, learning completion model and self-heating control device - Google Patents

Abstract

To provide a self-heating prediction method, a learning completion model and a self-heating control device that can accurately predict a self-heating property of carbide discharged from a storage tank.SOLUTION: A self-heating prediction method is executed by a computer to predict a self-heating property of carbide cured in a storage tank. The method predicts the self-heating property based on an output value of a learning completion model into which at least one of the following is input: a gas flow rate of oxygen-containing gas supplied to the storage tank; gas temperature of oxygen-containing gas; temperature inside the storage tank; and temperature of the carbide put into the storage tank.SELECTED DRAWING: Figure 3

Description

The present invention relates to a self-heating prediction method for predicting the self-heating properties of charcoal cured in a storage tank, a trained model, and a self-heating control device for controlling the self-heating properties of charcoal cured in a storage tank.

Organic waste such as sewage sludge is carbonized, for example by heating, and used as fuel. The carbonized organic waste obtained in this way has highly active surface functional groups on its particle surface immediately after carbonization, and is known to have self-heating properties if left as is. Therefore, to ensure safety during storage, a process (so-called aging process) is carried out to reduce the self-heating properties (see, for example, Patent Document 1).

Patent Document 1 discloses a powder/granular material storage device that measures the temperature of powder/granular material stored in a storage tank, and controls the exothermic reaction of the powder/granular material by controlling one or more of the speed at which the powder/granular material is transported by a transport device, the amount of gas supplied to the transport device, the amount of heating supplied to the transport device, and the amount of cooling supplied to the transport device based on this temperature information.

JP 2007-246245 A

The powder storage device described in Patent Document 1 controls the exothermic reaction of the powder stored in the storage tank based on temperature information of the powder stored in the storage tank, and is not a technology that predicts the self-heating properties of the carbonized material discharged from the storage tank.

Therefore, there is a demand for a self-heating prediction method, a trained model, and a self-heating control device that can accurately predict the self-heating properties of charcoal discharged from a storage tank.

One aspect of the present invention is a self-heating prediction method executed by a computer to predict the self-heating property of a charcoal cured in a storage tank, which predicts the self-heating property of the charcoal based on the output value of a trained model to which at least one of the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal charged into the storage tank is input.

One aspect of the present invention is a trained model that functions on a computer, and is composed of a decision tree consisting of a plurality of branching points arranged in a tree structure, and at least one of the gas flow rate of the oxygen-containing gas supplied to a storage tank for curing the charcoal, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal put into the storage tank is input, and the self-heating score is output by adding up the evaluation values at each of the branching points.

The inventors have found that the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal charged into the storage tank are correlated with the self-heating property of the charcoal discharged from the storage tank. Therefore, in this embodiment, the self-heating property of the charcoal is predicted based on the output value of a trained model to which at least one of the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal charged into the storage tank is input.

This trained model is composed of a decision tree consisting of multiple branch points arranged in a tree structure, and a performance evaluation of the trained model showed a high correlation with the actual measured value of self-heating. Therefore, in this embodiment, the self-heating property of the charcoal discharged from the storage tank can be predicted with high accuracy.
This predicted self-heating property of charcoal can be used to determine whether the charcoal can be used as fuel or to control the storage tank.

Another aspect is that at least the gas flow rate, the gas temperature, and the internal temperature are input to the trained model.

According to this embodiment, a high correlation was obtained in the order of the gas temperature of the oxygen-containing gas, the gas flow rate of the oxygen-containing gas, and the temperature of the carbonized material introduced into the storage tank, making it possible to accurately predict the self-heating properties of the carbonized material discharged from the storage tank.

Another aspect is that the oxygen-containing gas is supplied to the storage tank from multiple points on the side wall.

By supplying oxygen-containing gas from multiple locations on the side wall as in this embodiment, it is possible to obtain multiple measurements of the gas temperature and gas flow rate, improving the accuracy of predicting the self-heating properties of the carbide.

One aspect of the present invention is a self-heating property control device that controls the self-heating property of charcoal cured in a storage tank, and is provided with a gas flow rate measurement unit that measures the gas flow rate of an oxygen-containing gas supplied to the storage tank, a gas temperature measurement unit that measures the gas temperature of the oxygen-containing gas supplied to the storage tank, an internal tank temperature measurement unit that measures the internal temperature of the charcoal stored in the storage tank, a pre-introduction charcoal temperature measurement unit that measures the pre-introduction temperature of the charcoal before it is introduced into the storage tank, a prediction unit that predicts the self-heating property based on at least one of the gas flow rate, the gas temperature, the internal temperature, and the pre-introduction temperature, and a control unit that controls at least one of the gas flow rate, the gas temperature, the residence time of the charcoal in the storage tank, and the pre-introduction temperature based on the prediction result of the prediction unit.

As described above, the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal charged into the storage tank are correlated with the self-heating property of the charcoal discharged from the storage tank, so the prediction unit of this embodiment can accurately predict the self-heating property of the charcoal. Based on this predicted self-heating property, the control unit of this embodiment controls at least one of the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature, the residence time of the charcoal in the storage tank, and the temperature before the charging of the charcoal. As a result, when the self-heating property of the charcoal is high, the gas flow rate is increased, the gas temperature is raised, the residence time of the charcoal is increased, or the temperature before the charging of the charcoal is raised to promote the low-temperature oxidation treatment, and when the self-heating property of the charcoal is low, the gas flow rate is decreased, the gas temperature is decreased, the residence time of the charcoal is decreased, or the temperature before the charging of the charcoal is decreased to suppress the low-temperature oxidation treatment.

Another aspect is that the oxygen-containing gas supply ports are provided at multiple locations on the side wall of the storage tank.

By supplying oxygen-containing gas from multiple locations on the side wall as in this embodiment, it is possible to uniformly proceed with the low-temperature oxidation treatment of the carbide, and the self-heating tendency of the carbide can be reliably suppressed.

FIG. 1 is a configuration diagram of an organic sludge recycling system. FIG. 2 is an explanatory diagram of a carbide treatment device and a block diagram of a self-heating control device. FIG. 2 is a flow diagram of a self-heating prediction method and a self-heating control method. FIG. 2 is a correlation diagram between the gas temperature and the self-heating property of an oxygen-containing gas. FIG. 4 is a correlation diagram between the gas flow rate of an oxygen-containing gas and self-heating property. FIG. 11 is a correlation diagram between the temperature of the carbonized material put into the storage tank and the self-heating property. FIG. 11 is a correlation diagram between the internal temperature of a storage tank and self-heating property. FIG. 1 shows t-values from linear regression analysis. This is a correlation diagram between predicted values and actual measured values of self-heating properties using a trained model based on a decision tree. FIG. 1 is a diagram showing regression coefficients, significance probability, and correlation coefficients from linear regression analysis. This is a correlation diagram between predicted values and actual measured values of self-heating properties using a trained model based on a decision tree.

The self-heating prediction method, trained model, and self-heating control device according to the embodiment of the present invention will be described with reference to Figures 1 to 9. However, the present invention is not limited to the following embodiment, and various modifications are possible without departing from the spirit of the invention.

[Explanation of overall configuration]
The carbide treatment device 100 will be described with reference to Figures 1 and 2. As shown in Figure 1, the carbide treatment device 100 is used for carbide treatment in an organic sludge recycle system 200 that uses organic sludge such as sewage sludge as recycled fuel F. In this embodiment, a case will be described in which dried sludge L, which is organic sludge that has been dried and granulated into pellets in advance, is supplied to the organic sludge recycle system 200. The organic sludge recycle system 200 obtains recycled fuel F by carbonizing the dried sludge L.

The carbonized material processing device 100 performs a process to reduce the self-heating property of the carbonized material M obtained from the dried sludge L (so-called aging process (low-temperature oxidation process), hereinafter referred to as "carbonized material processing").

In addition to the carbonized material treatment device 100, the organic sludge recycle system 200 mainly comprises a carbonization furnace 10, a cooler 20, and a stock tank 27.

The dried sludge L is granulated into cylindrical pellets, for example, by extrusion granulation. This is to homogenize the particle properties, such as particle shape and bulk density, of the dried sludge L or the charcoal M and recycled fuel F obtained from the dried sludge L, thereby improving the handling properties of the charcoal M and the like in the organic sludge recycling system 200 and the charcoal processing device 100, and the uniformity of the charcoal processing. The dried sludge L may be granulated into other shapes, such as spheres, or may not be granulated.

The dried sludge L is supplied to the carbonization furnace 10 and carbonized, thereby obtaining a cylindrical pellet-shaped carbonized material M. The carbonized material M discharged from the carbonization furnace 10 is humidified and cooled in the cooler 20.
The carbonized material M is then stored in the carbonized material processing device 100 for a predetermined time. The carbonized material processing device 100 performs carbonization processing by passing a processing gas G (an example of an oxygen-containing gas) containing oxygen through the stored carbonized material M for a predetermined time (see also FIG. 2). The carbonized material M that has been carbonized becomes recycled fuel F that can be safely stored in a tank or the like, and is discharged from the carbonized material processing device 100. The recycled fuel F discharged from the carbonized material processing device 100 is stored in a stock tank 27 in preparation for shipment to the market. Hereinafter, the downstream side of the transport path of the carbonized material M or recycled fuel F from the carbonization furnace 10 to the stock tank 27 will be simply referred to as the downstream side, and the opposite side will be referred to as the upstream side.

[Explanation of each part]
The carbonization furnace 10 is an apparatus for heating the dried sludge L in a low-oxygen atmosphere (hereinafter, sometimes referred to as "carbonization treatment") to obtain a carbonized material M. The carbonization furnace 10 is configured as a rotary kiln. The carbonization furnace 10 may be of a fluidized bed type or a screw type in addition to a rotary kiln type. The carbonization furnace 10 carbonizes the dried sludge L at a temperature of about 250°C to 600°C.

In this embodiment, in the carbonization furnace 10, combustible gas is generated from the dried sludge L during the carbonization process. The combustible gas is supplied to, for example, a secondary combustion furnace 12 and burned, and then discharged to the outside as combustion exhaust Ef via exhaust gas treatment equipment 13 such as an exhaust heat recovery machine or a scrubber. After being discharged from the carbonization furnace 10, the carbonized material M is fed into the cooler 20 via a chute 11 or the like.

The cooler 20 is a device that cools the carbide M. In this embodiment, the cooler 20 is a device that is equipped with a nozzle 20E that supplies cooling water CW to a screw conveying device that moves the carbide M using a screw installed inside a casing. In this embodiment, the cooler 20 conveys the carbide M in one direction using a screw, and sprays and supplies cooling water CW to the carbide M being conveyed.

The carbide M inside the cooler 20 is cooled by the latent heat of vaporization of the cooling water CW. In this embodiment, the carbide M is rapidly cooled to less than 60°C. In this embodiment, the carbide M is cooled and humidified by spraying the cooling water CW. In this embodiment, the carbide M is humidified to a moisture content of 5% to 20%, and preferably 8% to 15%, on a dry basis (weight ratio to the weight of completely dried carbide). In addition, a jacket through which a refrigerant circulates inside may be provided on the outside of the cooler 20, and a method of indirectly cooling the carbide M inside the cooler 20 may also be used.

In the present embodiment, a carrier gas CG, which is an inert gas such as nitrogen, flows through the cooler 20 in a direction opposite to the flow of the carbide M. In the cooler 20, in addition to the carrier gas CG and the water vapor of the cooling water CW, flammable gases such as carbon monoxide gas and odorous gases are generated. Therefore, the exhaust gas from the cooler 20 is introduced into the secondary combustion furnace 12 through the exhaust pipe 14. The cooled carbide M is supplied from the cooler 20 to the flight conveyor 22. In this embodiment, the carbide M is supplied to the flight conveyor 22 through a rotary valve 21 that separates the cooler 20 from the inlet (not shown) of the flight conveyor 22.

The carbide M discharged from the cooler 20 is transported to the cushion tank 23 by the flight conveyor 22, and is supplied from the cushion tank 23 to the carbide processing device 100. In this embodiment, the carbide M transported to the cushion tank 23 is supplied to the carbide processing device 100, for example, via a rotary valve (illustrated) or piping. Note that instead of using the flight conveyor 22, the carbide M discharged from the cooler 20 may be supplied to the carbide processing device 100 via a chute or the like. Also, instead of transporting by the flight conveyor 22, transport may be performed using air transport, a belt conveyor, a bucket conveyor, or the like.

The carbide processing device 100 is a device that processes the carbide M into a carbide to obtain recycled fuel F. The carbide processing device 100 will be described later. The recycled fuel F obtained by the carbide processing device 100 is transported from the flight conveyor 25 to the cushion tank 26. At this time, an air cooling device 25a such as a heat exchanger to which a refrigerant is supplied from a chiller is arranged upstream of the flight conveyor 25, and the recycled fuel F may be cooled using air A cooled by this air cooling device 25a. The recycled fuel F is loaded from the cushion tank 26 into the stock tank 27 and stored until shipment. The cushion tank 26 may be omitted and the recycled fuel F may be directly loaded into the stock tank 27 via a chute or the like, or the recycled fuel F may be transported using air transport, a belt conveyor, a bucket conveyor, or the like instead of being transported by the flight conveyor 25. The cushion tank 26 and the stock tank 27 may be omitted and the transported recycled fuel F may be directly supplied to the utilization facility.

The carbide processing device 100 will be described in detail. As shown in FIG. 2, the carbide processing device 100 includes a storage tank 3 for storing carbide M and performing carbide processing, and an aeration section 4 for aerating (supplying) processing gas G to the deposit B of carbide M stored in the storage tank 3. The carbide processing device 100 also includes an in-tank temperature measuring section T for measuring the internal temperature of the storage tank 3, a gas temperature measuring section St for measuring the gas temperature of the processing gas G supplied to the storage tank 3, a gas flow rate measuring section Sv for measuring the gas flow rate of the processing gas G supplied to the storage tank 3, and a pre-input carbide temperature measuring section Sc for measuring the temperature (pre-input temperature) of the carbide M before being input into the storage tank 3. The organic sludge recycling system 200 may also include a pre-shipment char temperature measuring unit Sf that measures the temperature (pre-shipment temperature) of the recycled fuel F produced by curing the char M in the storage tank 3 before shipping (see FIG. 1). The gas temperature measuring unit St, the pre-input char temperature measuring unit Sc, and the pre-shipment char temperature measuring unit Sf are configured with known temperature sensors, and the gas flow rate measuring unit Sv is configured with a known flow rate sensor, so detailed description will be omitted. In this embodiment, the pre-shipment char temperature measuring unit Sf is provided in the stock tank 27, but it may be provided in the cushion tank 26, and the installation location of the pre-shipment char temperature measuring unit Sf is not particularly limited as long as it measures the temperature of the recycled fuel F before shipping. Here, "curing" is synonymous with a process (so-called aging process) that reduces the self-heating property of the recycled fuel F before shipping in order to ensure safety during storage.

As shown in FIG. 1, the processing gas G is a gas containing at least a first processing gas G1 containing oxygen such as air A. In this embodiment, the ventilation section 4 supplies the processing gas G, which is a mixture of the first processing gas G1 and the second processing gas G2. The ventilation section 4 includes a first fan 41 for supplying the first processing gas G1 and a second fan 42 for supplying the second processing gas G2. The internal temperature of the storage tank 3 refers to the measured temperature of the deposit B. The second fan 42 for supplying the second processing gas G2 may be omitted and a constant amount of the second processing gas G2 may be supplied, or the supply amount may be adjusted by a valve or the like instead of the second fan 42.

The carbide processing device 100 performs a storage process for storing the carbide M in a storage tank 3, an aeration process for passing an oxygen-containing processing gas G through the deposit B of the carbide M, and a discharge process for retaining the carbide M for a predetermined period of time and then discharging it, thereby achieving carbide processing of the carbide M and obtaining recycled fuel F.

As shown in FIG. 2, the storage tank 3 is a metal container that stores the carbide M in layers by piling it up vertically (from bottom to top). The storage tank 3 is also a supply container that supplies the carbide M stored in layers to the next process while maintaining the layered state. The storage tank 3 has a supply section 31 that serves as an inlet for introducing the carbide M into the internal space of the container body 30 of the storage tank 3, a discharge section 32 that discharges the recycled fuel F from the internal space of the storage tank 3, and an emergency discharge port 35 that can discharge the carbide M in an emergency.

The container body 30 of the storage tank 3 in this embodiment is a tank with a circular cross-sectional shape that narrows at the bottom, and the narrowing angle (cone angle) of the bottom is set so that the residence time of the carbonized material M (recycled fuel F) discharged from the discharge section 32 in the tank is uniform. Note that a baffle plate in the shape of a crest (a so-called cone baffle) may be provided near the boundary between the upper and lower parts of the storage tank 3 and near the center of the storage tank 3 in the radial direction to make the residence time of the carbonized material M (recycled fuel F) in the tank uniform.

The storage tank 3 has a supply unit 31 at the upper end of the container body 30. The supply unit 31 includes a supply pipe connected to the internal space of the storage tank 3, and may include a rotary valve provided on the supply pipe. When the supply unit 31 includes a rotary valve, the carbide M can be introduced into the internal space of the storage tank 3 in a state in which it is isolated from the upstream atmosphere. In this embodiment, the carbide M is continuously supplied to the supply unit 31 at a constant supply rate.

The storage tank 3 has a discharge section 32 at the lower end of the container body 30. The discharge section 32 has a discharge pipe 32b that is connected to the internal space of the storage tank 3, and a rotary valve 32a that is provided on the discharge pipe 32b as a discharge device for the carbide M (recycled fuel F).

The discharge pipe 32b is a cylindrical pipe that is provided downward from the lower end of the container body 30. The storage tank 3 can discharge the recycled fuel F from the internal space of the storage tank 3 to the bottom of the storage tank 3 while being isolated from the downstream atmosphere by the rotary valve 32a. The recycled fuel F is continuously discharged from this discharge section 32.

The charcoal M is continuously fed into the storage tank 3 at a constant feed rate, and the recycled fuel F is continuously discharged from the storage tank 3. The average residence time of the charcoal M (recycled fuel F) in the storage tank 3 is controlled to be, for example, 2 to 4 days (48 hours to 96 hours). The residence time is set to a length necessary and sufficient for obtaining the recycled fuel F from the charcoal M.
If the residence time is too short, it is not possible to sufficiently reduce the self-heating property of the carbide M, and it is not possible to ensure the safety of the recycled fuel F. If the residence time is too long, the production efficiency of the recycled fuel F decreases, making it uneconomical, which is not preferable.

The storage tank 3 has a gas supply port 34 (an example of a supply port) on the side wall, which introduces the processing gas G into the internal space and supplies the processing gas G containing oxygen to the deposit B. The gas supply port 34 has a plurality of first supply ports 34a (four in this embodiment) connected to the upper side wall of the storage tank 3, a plurality of second supply ports 34b (four in this embodiment) connected to the side wall near the center of the storage tank 3, and a plurality of third supply ports 34c (four in this embodiment) connected to the lower side wall of the storage tank 3. In other words, these gas supply ports 34 for the processing gas G are provided at a plurality of locations (three in this embodiment) along the vertical direction of the side wall of the storage tank 3. In addition, each of the supply ports 34a, 34b, and 34c is provided at a plurality of locations (four in this embodiment) along the circumferential direction of the side wall of the storage tank 3. As a result, the processing gas G supplied from the gas supply port 34 is uniformly passed through the particle layer of the deposit B from the bottom to the top of the storage tank 3. The processing gas G comes into solid-gas contact with the particle surfaces of the carbide M while passing through the particle layer of the deposit B, and oxidizes the surface functional groups on the particle surfaces. A third supply port 34c connected to the side wall at the bottom of the storage tank 3 may be provided on the bottom wall of the storage tank 3.

The deposit B in the storage tank 3 forms a plurality of layer regions made of particle layers, and is composed of deposit B1, deposit B2, and deposit B3. Hereinafter, when deposit B1, deposit B2, and deposit B3 are collectively described as the particle layers of deposit B, they will be simply referred to as "layer regions."
The processing gas G that has passed through the particle layer of the deposit B is exhausted to the outside as exhaust gas E from an exhaust pipe 33 provided at the upper end of the storage tank 3. The exhaust gas E is introduced into a secondary combustion furnace 12 (see FIG. 1) or the like and purified, and then discharged into the atmosphere.

Temperature sensors T1 to T6, which are sensor probes of the tank temperature measurement section T, are attached to the storage tank 3. In this embodiment, the temperature sensors T1 to T6 are rod-shaped sensor probes with a temperature detection element at the tip, and use resistance temperature detectors. Note that the temperature sensors T1 to T6 may also use thermocouples or other sensors.

The temperature sensors T1 to T6 are installed on the side wall of the storage tank 3. The temperature sensors T1 to T6 are attached along the radial direction of the container body 30, with a rod-shaped sensor probe penetrating the wall of the container body 30 from the outside of the side of the storage tank 3, and the tip of the sensor being positioned inside the tank of the storage tank 3. Note that the temperature sensors T1 to T6 are not limited to being attached along the radial direction of the container body 30, but may be attached at an incline vertically upward or downward from the radial direction of the container body 30, or at an incline from the radial direction to the circumferential direction of the container body 30.

The temperature sensors T1 and T2 are provided near the first supply port 34a, with the tip of the temperature sensor T1 located near the center of the deposit B1 and the tip of the temperature sensor T2 located near the outer periphery of the deposit B1. The temperature sensors T3 and T4 are provided near the second supply port 34b, with the tip of the temperature sensor T3 located near the center of the deposit B2 and the tip of the temperature sensor T4 located near the outer periphery of the deposit B2. The temperature sensors T5 and T6 are provided near the third supply port 34c, with the tip of the temperature sensor T5 located near the center of the deposit B3 and the tip of the temperature sensor T6 located near the outer periphery of the deposit B3. Note that, although the temperature sensors T1 and T2, the temperature sensors T3 and T4, or the temperature sensors T5 and T6 in this embodiment are each inserted to a different depth, they may each be inserted to the same depth, and the number and arrangement of the temperature sensors T1 to T6 are not particularly limited.

The oxygen concentration measuring unit So is attached to the exhaust pipe 33 of the storage tank 3. The oxygen concentration measuring unit So is attached so that the gas in the exhaust pipe 33 can be sucked in and the oxygen concentration can be measured by the oxygen concentration meter. The exhaust pipe 33 is provided with a resistor 33a that acts as a resistance to the flow of the exhaust gas E in order to maintain the internal space of the storage tank 3 at a positive pressure. The resistor 33a can be, for example, a valve device with an adjustable opening such as a butterfly valve. The resistance to the flow of the exhaust gas E can be changed by changing the opening of the resistor 33a. If the resistance to the flow of the exhaust gas E is increased, the pressure in the internal space of the storage tank 3 changes to the positive pressure side. By maintaining the pressure in the internal space of the storage tank 3 on the positive pressure side, it is possible to prevent the flow (intrusion) of oxygen-containing gas into the internal space of the storage tank 3 from any source other than the gas supply port 34. In other words, by preventing the intrusion into the internal space of the storage tank 3, it is possible to prevent the carbide M from locally heating up and catching fire due to the intrusion.

1, the processing gas G is a gas containing at least a first processing gas G1 containing oxygen such as air A. The processing gas G in this embodiment is a gas obtained by mixing the first processing gas G1 and a second processing gas G2 containing an inert gas (e.g., nitrogen).
Therefore, the oxygen concentration of the processing gas G is determined by the mixing ratio of the first processing gas G1 and the second processing gas G2. The oxygen concentration of the processing gas G decreases as the mixing ratio of the second processing gas G2 increases.

The first processing gas G1 is produced by sucking in air A (outside air) with the first fan 41, heating the air A with the heater 43 to keep it at a predetermined temperature, and then merging the second processing gas G2 with the second fan 42 or the like to produce the processing gas G. The processing gas G is humidified with the humidifier 45 and blown (supplied) into the multiple air supply pipes 41a connected to the gas supply port 34, thereby supplying the gas to the storage tank 3. In this embodiment, the multiple air supply pipes 41a are connected to the first supply port 34a, the second supply port 34b, and the third supply port 34c, which are branched after the first processing gas G1 and the second processing gas G2 are merged. In addition, the air supply pipe 41a connected to the first supply port 34a is provided with a valve V1, the air supply pipe 41a connected to the second supply port 34b is provided with a valve V2, and the air supply pipe 41a connected to the third supply port 34c is provided with a valve V3. By adjusting the opening of these valves V1, V2, and V3, the flow rate of the processing gas G supplied to each of the deposits B1, B2, and B3 can be changed.

The first processing gas G1 is a carrier gas that supplies oxygen to the carbide M. The oxidation of the carbide M progresses according to the flow rate of the first processing gas G1, i.e., according to the amount of oxygen supplied. The carbide M generates heat in response to this oxidation, while the carbide M is cooled by the flow of the first processing gas G1. When the flow rate of the first processing gas G1 is increased, i.e., when the oxygen supply rate is increased, the oxidation rate (processing rate) of the carbide M becomes faster. As a result, the amount of heat generated per unit time by the carbide M increases, while the amount of heat removed per unit time by the carbide M increases, and the temperature of the deposit B, i.e., the internal temperature of the storage tank 3, decreases.

The second processing gas G2 is a gas that is inert to the carbide M. The second processing gas G2 is also a refrigerant that cools the carbide M. By supplying the second processing gas G2, the oxygen concentration of the processing gas G is reduced, thereby slowing down the reaction rate of the oxidation reaction of the carbide M. This makes it possible to prevent the carbide M from locally generating heat within the storage tank 3. In addition to cooling by the first processing gas G1, the supply of the second processing gas G2 can also cool the carbide M (deposit B).

2, the gas supply pipe 41a connected to the gas supply port 34 is provided with a gas temperature measuring section St and a gas flow rate measuring section Sv. In addition, the supply section 31, which is an inlet for feeding the carbide M into the storage tank 3, is provided with a pre-feed carbide temperature measuring section Sc. In this embodiment, the gas temperature measuring sections St1 to St3 and the gas flow rate measuring sections Sv1 to Sv3 are provided at the first supply port 34a, the second supply port 34b, and the third supply port 34c, respectively, and can measure the gas temperature and gas flow rate of the processing gas G supplied to each of the deposits B1, B2, and B3. The pre-feed carbide temperature measuring section Sc may be provided, for example, at the outlet portion of the cooler 20 or on the flight conveyor 22, and is not particularly limited as long as it is a location downstream of the cooler 20 where the temperature of the carbide M before being fed into the storage tank 3 can be measured. The gas temperature measuring unit St may be provided at one location in the flow path before the air supply pipe 41a branches, rather than at each of the first supply port 34a, the second supply port 34b, and the third supply port 34c.

[Description of the Self-Heating Control Device]
Next, a method for predicting the self-heating property of the carbide M (recycled fuel F) by the control device X (corresponding to the self-heating property control device) will be described with reference to FIGS. 2 and 3. The control device X is a central control mechanism that controls the overall operation of the carbide processing device 100. The control device X can be configured, for example, by a software program for implementing various processes, a CPU or the like that executes the software program, and various hardware controlled by the CPU or the like. The control device X in this embodiment is a computer that includes a CPU and an input/output circuit or the like. Programs, data, and control parameters required for the operation of the control device X are stored in the storage unit 5. Note that the destination where these programs and data are stored is not particularly limited. These programs and data may be stored in a storage device such as a disk or flash memory that is provided separately for exclusive use. They may also be an external server connected to be able to communicate with each other.

The control device X includes a memory unit 5, a communication unit 6, a model generation unit 7, a prediction unit 8, and a control unit 9. The self-heating prediction method executed by the control device X predicts the self-heating property of the recycled fuel F (an example of a carbide) based on the output value of a trained model 50 to which at least one of the gas flow rate of the oxygen-containing gas supplied to the storage tank 3, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, and the temperature of the carbide M put into the storage tank 3 is input.

The memory unit 5 is composed of a non-transient storage medium such as an HDD or SSD, or a temporary storage medium such as a RAM, and stores programs and applications executed by the processor. The memory unit 5 also stores various data acquired by operation of the carbide processing device 100 in chronological order, such as the measurement value of the pre-injection carbide temperature measurement unit Sc, the measurement value of the gas temperature measurement unit St, the measurement value of the gas flow rate measurement unit Sv, the measurement value of the tank temperature measurement unit T, and the measurement value of the oxygen concentration measurement unit So. Furthermore, the memory unit 5 stores a trained model 50 that is operated by a computer.

The trained model 50 can output a score of the self-heating property of the recycled fuel F using as input values the gas flow rate of the oxygen-containing gas measured by the gas flow rate measuring unit Sv, the gas temperature of the oxygen-containing gas measured by the gas temperature measuring unit St, the internal temperature of the storage tank 3 measured by the tank temperature measuring unit T, and the temperature of the charcoal M to be added to the storage tank 3 measured by the charcoal temperature measuring unit Sc before addition. The input value may be time-series data acquired by the operation of the charcoal processing device 100, or may be a feature obtained by processing (e.g., averaging) this time-series data. The score of the self-heating property of the recycled fuel F may be the oxygen consumption characteristic of the recycled fuel F discharged from the storage tank 3 described later, or may be the difference between the oxygen consumption characteristic of the charcoal M to be added to the storage tank 3 and the oxygen consumption characteristic of the recycled fuel F discharged from the storage tank 3 (hereinafter referred to as the "oxygen consumption reduction amount"), and is not particularly limited as long as it is numerical data that can evaluate the self-heating property of the recycled fuel F.

This trained model 50 is a model that functions by a computer, and is obtained by machine learning learning with teacher data. The machine learning is configured with a decision tree consisting of multiple branching points arranged in a tree structure. This machine learning is structured in such a way that feature values are evaluated at each branching point of the decision tree, and an evaluation value according to the evaluation result is assigned to each branching point. The evaluation values are then summed along the branches of the decision tree to obtain a score for the self-heating property of the recycled fuel F. The model may be an ensemble model such as XGBoost, Random Forest, LightGBM, CatBoost, or AdaBoost, in which multiple decision trees are associated with each other.

The trained model 50 may be generated by other machine learning methods such as linear regression analysis. The trained model 50 generated by linear regression analysis can input multiple explanatory variables (such as the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, and the temperature of the carbide M put into the storage tank 3) and output the score of the self-heating property of the recycled fuel F as the objective variable.

The trained model 50 may be generated by deep learning. Deep learning is performed by AI (artificial intelligence) including known convolutional neural networks (CNN, DCGAN, etc.). In the convolutional neural network, a deep hierarchical model that mimics human neural circuits is constructed, and the self-heating property of the recycled fuel F is inferred and a score is output based on multiple input data (the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, and the temperature of the carbide M put into the storage tank 3, etc.). This deep learning is configured by known applications provided via an Internet line.

The communication unit 6 is an interface that transmits and receives data to and from the carbide processing device 100 via an Internet line. The communication unit 6 may receive data directly from the carbide processing device 100, or may store data acquired by the carbide processing device 100 in a server (not shown) and receive the stored data from the server.

The model generation unit 7 includes a processor and generates the trained model 50. The processor includes an ASIC, an FPGA, a CPU, or other hardware for executing applications stored in the memory unit 5 (same below). The model generation unit 7 generates the trained model 50 based on past performance data (operational performance data of the carbide treatment device 100).

Examples of actual data used by the model generation unit 7 include the gas flow rate of the oxygen-containing gas measured by the gas flow rate measurement unit Sv, the gas temperature of the oxygen-containing gas measured by the gas temperature measurement unit St, the internal temperature of the storage tank 3 measured by the tank temperature measurement unit T, the temperature of the charcoal M introduced into the storage tank 3 measured by the charcoal temperature before introduction measurement unit Sc, the post-cooling temperature (pre-shipment temperature) of the recycled fuel F measured by the charcoal temperature before shipment measurement unit Sf, and the oxygen consumption characteristics of the recycled fuel F discharged from the storage tank 3. The actual data may also include the oxygen concentration of the processing gas G supplied from the ventilation unit 4 and the oxygen concentration of the exhaust gas E measured by the oxygen concentration measurement unit So.

Here, the oxygen consumption characteristic will be explained. The oxygen consumption characteristic is the mass of oxygen consumed per unit mass of carbide M per unit time. The oxygen consumption characteristic means the degree of activity of carbide M in the storage tank 3 (activity to be oxidized by oxygen). The higher the oxygen consumption characteristic, the more the carbide M reacts with oxygen in an oxygen-containing atmosphere to generate heat, and the greater the risk of fire, etc. Therefore, in order to use carbide M as recycled fuel F, it is necessary to reduce this oxygen consumption characteristic to a predetermined value or less.

The oxygen consumption characteristic can be measured directly by sampling the recycled fuel F discharged from the storage tank 3. In this embodiment, the recycled fuel F discharged from the storage tank 3 is sampled, the recycled fuel F is stored in a sealed container for a predetermined time (for example, one hour), and the oxygen consumption characteristic of the recycled fuel F is measured after the predetermined time has elapsed. The oxygen consumption characteristic can be obtained from the mass of the carbonized material M input into the storage tank 3, the amount of oxygen supplied to the storage tank 3 (the oxygen concentration of the processing gas G supplied from the ventilation section 4), and the amount of oxygen discharged from the storage tank 3 (the oxygen concentration of the exhaust gas E measured by the oxygen concentration measuring section So). In this embodiment, the volume of each layer region is set to be constant, so that the oxygen consumption characteristic for each layer region can be obtained as a value proportional to the difference obtained by subtracting the oxygen concentration contained in the processing gas G discharged from each layer region from the oxygen concentration contained in the processing gas G supplied to each layer region, multiplied by the ventilation speed, which is the ventilation volume of the processing gas G per unit time. In addition, a sampling nozzle may be provided in each layer area, and the gas in the storage tank 3 sucked in through this sampling nozzle may be measured using an oxygen concentration meter.

The prediction unit 8 includes a processor and predicts the self-heating property of the recycled fuel F based on at least one of the gas flow rate of the oxygen-containing gas measured by the gas flow rate measurement unit Sv, the gas temperature of the oxygen-containing gas measured by the gas temperature measurement unit St, the internal temperature of the storage tank 3 measured by the tank temperature measurement unit T, and the temperature of the carbide M to be added to the storage tank 3 measured by the carbide temperature measurement unit Sc before addition. In this embodiment, the prediction unit 8 inputs the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, and the temperature of the carbide M to be added to the storage tank 3, etc., into the trained model 50 generated by the model generation unit 7, and sets the score of the self-heating property of the recycled fuel F output by the trained model 50 as the prediction result.

The control unit 9 is equipped with a processor and controls at least one of the ventilation rate (gas flow rate) of the processing gas G supplied to the storage tank 3, the temperature (gas temperature) of the processing gas G, the residence time of the carbonized material M in the storage tank 3, and the temperature of the carbonized material M before it is introduced, based on the prediction result of the prediction unit 8 (the predicted value of the oxygen consumption characteristics of the recycled fuel F discharged from the storage tank 3, hereinafter simply referred to as the "predicted value").

[Control flow explanation]
The control flow of the control device X will be described with reference to FIG. 3. First, in the control device X, the communication unit 6 acquires the operation performance data of the carbide treatment device 100 over a predetermined period, and the memory unit 5 stores it as time-series operation performance data (#31). As described above, this operation performance data is composed of the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, the temperature of the carbide M inputted into the storage tank 3 measured at the outlet of the cooler 20, the post-cooling temperature (pre-shipment temperature) of the recycled fuel F discharged from the storage tank 3, and the oxygen consumption characteristic of the recycled fuel F discharged from the storage tank 3. Since these operation performance data are measured at different points in the storage tank 3, the model generation unit 7 takes into account the residence time of the carbide M in the storage tank 3 and performs interpolation (time correction) from the time-series operation performance data to obtain operation performance data related to the same carbide M.

Then, the model generation unit 7 generates a trained model 50 based on the interpolated operating performance data (#32). In this embodiment, the trained model 50 is generated by machine learning using the decision tree or linear regression analysis described above. This trained model 50 is stored in the memory unit 5. Note that once the trained model 50 is generated, it may be updated at a predetermined interval based on the operating data of the carbide treatment device 100.

Then, the prediction unit 8 acquires real-time operation data (process data) of the carbide processing device 100 via the communication unit 6 (#33). As described above, this process data is composed of the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, and the temperature of the carbide M put into the storage tank 3. This process data is also composed of time-series data, and the prediction unit 8 takes into account the residence time of the carbide M in the storage tank 3 and interpolates (time-corrects) the time-series process data to obtain process data related to the same carbide M.

Next, the prediction unit 8 inputs the acquired process data into the trained model 50, and the score of the self-heating property of the recycled fuel F output by the trained model 50 based on the process data is set as the prediction result (#34). In this embodiment, since the process data acquired upstream of the discharge unit 32 is used, the control unit 9 can feedback control the operation of the carbide processing device 100 using the prediction result of the self-heating property of the recycled fuel F. For example, the control unit 9 automatically adjusts at least one of the aeration amount (gas flow rate) of the processing gas G supplied to the storage tank 3, the temperature (gas temperature) of the processing gas G, the residence time of the carbide M in the storage tank 3, and the temperature before the introduction of the carbide M based on the prediction result of the prediction unit 8, to control the oxygen consumption rate to be between the upper limit and the lower limit.

When the predicted value of the recycled fuel F is greater than the upper limit value or when the predicted reduction in oxygen consumption of the recycled fuel F is smaller than a predetermined reduction threshold (the target reduction in the oxygen consumption characteristics of the carbonized material M cured in the storage tank 3) (No. 35 Yes), the control unit 9 increases the opening of valves V1, V2, V3 and/or the rotation speed of the first fan 41, and increases the flow rate of the processing gas G supplied to each of the deposits B1, B2 and B3, thereby promoting the carbonized material processing in the storage tank 3 while reducing the self-heating tendency of the recycled fuel F (No. 36, carbonized material processing promotion control). Other examples of the carbonized material processing promotion control by the control unit 9 include reducing the amount of humidification water W in the humidifier 45 or increasing the temperature of the humidification water W so that the temperature of the processing gas G increases, reducing the discharge rate or discharge amount of the discharge unit 32 so that the residence time of the carbonized material M increases, and reducing the amount of cooling water CW in the cooler 20 or increasing the temperature of the cooling water CW so that the temperature before the introduction of the carbonized material M increases. In addition, when controlling so that the temperature of the processing gas G increases in order to reduce the self-heating property of the recycled fuel F, if the internal temperature of the storage tank 3 (measurement value of the tank temperature measurement unit T) is equal to or lower than a predetermined value (e.g., 50°C), an upper limit value (e.g., 50°C) may be set for the increase in the temperature of the processing gas G.

On the other hand, in the present embodiment, when the predicted value of the recycled fuel F is smaller than the lower limit value or the predicted oxygen consumption reduction amount of the recycled fuel F is greater than a predetermined reduction threshold value (the target reduction amount of the oxygen consumption characteristic in the carbonized material M cured in the storage tank 3) (No in #35, Yes in #37), the control unit 9 reduces the supply amount of the processing gas G by lowering the opening degree of the valves V1, V2, V3 and/or the rotation speed of the first fan 41 (Carbide processing suppression control in #38). As a result, the processing efficiency can be improved while suppressing the carbonized material processing in the storage tank 3. In other words, the supply amount of the processing gas G is made appropriate without excessive so that the predicted value of the recycled fuel F does not further decrease under the situation where it is smaller than the lower limit value. Other examples of the carbide processing suppression control by the control unit 9 include increasing the amount of humidification water W in the humidifier 45 or decreasing the temperature of the humidification water W so that the temperature of the processing gas G decreases, increasing the discharge speed or discharge amount of the discharge unit 32 so that the residence time of the carbide M decreases, and increasing the amount of cooling water CW in the cooler 20 or decreasing the temperature of the cooling water CW so that the temperature before the introduction of the carbide M decreases. The upper limit value, lower limit value, or reduction threshold value that serves as the judgment threshold is set in advance taking into consideration the specifications of the storage tank 3 in order to set the oxygen consumption characteristic of the recycled fuel F generated by curing the carbide M in the storage tank 3 to a specified value suitable for shipping. The supply amount of the processing gas G may be reduced after other controls (control of the humidifier 45, discharge unit 32, and cooler 20) are performed while maintaining the supply amount of the processing gas G, or multiple controls may be performed in sequence. In this case, when the control value of a certain control becomes the upper limit value or the lower limit value, the control method is not limited, for example, to move to the next control.

As another control form, when the predicted value of the recycled fuel F is smaller than the lower limit value or when the predicted oxygen consumption reduction amount of the recycled fuel F is greater than a predetermined reduction threshold (a target reduction amount of the oxygen consumption characteristic in the carbonized material M cured in the storage tank 3) (No in #35, Yes in #37), it is preferable that the control unit 9 maintains the opening degree of the valves V1, V2, and V3 and/or the rotation speed of the first fan 41 and does not change the supply amount of the processing gas G (#38, maintain the status quo). Similarly, it is preferable not to execute other controls (control of the humidifier 45, the exhaust unit 32, and the cooler 20). In this case, the prediction accuracy of the self-heating property of the recycled fuel F can be improved by inputting and learning the operating conditions that reduce the oxygen consumption characteristic into the learned model 50.

Using Figures 4 to 9, Example 1 showing the performance of the control device X will be described. The model generation unit 7 generated the trained model 50 using the operating data (parameter n = 3746) of the carbide processing device 100 acquired between December 2019 and July 2020. The residence time of the deposit B1 in the storage tank 3 was set to 37 hours, the residence time of the deposit B2 to 20 hours, and the residence time of the deposit B3 to 15 hours. In other words, from the measurement time of the oxygen consumption characteristics of the recycled fuel F, the measurement values of the temperature sensors T1 to T2, the gas temperature measurement unit St1, and the gas flow measurement unit Sv1 from 72 hours ago, the measurement values of the temperature sensors T3 to T4, the gas temperature measurement unit St2, and the gas flow measurement unit Sv2 from 35 hours ago, and the measurement values of the temperature sensors T5 to T6, the gas temperature measurement unit St3, and the gas flow measurement unit Sv3 from 15 hours ago were extracted and used as input data for the same carbide M. The residence time of the deposit B is calculated based on the amount of carbide M filled up to the installation positions of each temperature sensor T1-T2, T3-T4, T5-T6 (the amount of carbide M remaining in the storage tank 3 between the bottom and the height of the temperature sensor) and the discharge speed of the carbide M at the discharge section 32.

4 to 7 show the correlation between the input data input to the trained model 50 and the oxygen consumption characteristics of the recycled fuel F. As shown in FIG. 4, the correlation coefficient R 2 indicating the variance from the approximation equation between the average value of the gas temperature of the oxygen-containing gas measured by the gas temperature measuring unit St (average of the measured values of the gas temperature measuring units St1 to St3) and the oxygen consumption characteristics of the recycled fuel F was 0.5 or more. As shown in FIG. 5, the correlation coefficient R ² indicating the variance from the approximation equation between the average value of the gas flow rate of the oxygen-containing gas measured by the gas flow rate measuring unit Sv (average of the measured values of the gas flow rate measuring units Sv1 to Sv3) and the oxygen consumption characteristics of the recycled fuel F was 0.5 or more. As shown in FIG. 6, the correlation coefficient R ² indicating the variance from the approximation equation between the temperature of the carbide M to be introduced into the storage tank 3 measured by the carbide temperature measuring unit Sc before introduction provided at the outlet of the cooler ²⁰ and the oxygen consumption characteristics of the recycled fuel F was 0.5 or more. 7, the correlation coefficient R2 indicating the variance from the approximation equation between the internal temperature of the storage tank 3 (measurement values of the temperature sensors T2 to T6) measured by the internal temperature measuring unit T (temperature sensor T4 on the middle wall surface of the storage tank 3) and the oxygen consumption characteristics of the recycled fuel F was 0.5 or more. Although not shown, the correlation coefficients ^R2 between the measurement values of the temperature sensor T2 on the upper wall surface of the storage tank 3, the temperature sensor T3 inside the middle stage, the temperature sensor T5 inside the lower stage, and the temperature sensor T6 on the lower wall surface of the storage tank 3 and the oxygen consumption characteristics of the recycled fuel F were 0.79, 0.57, 0.80, and ^0.86 , respectively, all of which were 0.5 or more.

From these results, it can be seen that it is useful to use the gas flow rate of the oxygen-containing gas measured by the gas flow rate measuring unit Sv, the gas temperature of the oxygen-containing gas measured by the gas temperature measuring unit St, the internal temperature of the storage tank 3 measured by the tank temperature measuring unit T, and the temperature of the carbide M to be added to the storage tank 3 measured by the carbide temperature before addition measuring unit Sc as input values for the trained model 50.

Figure 8 shows the weights (t values) of each index in the trained model 50 generated by the model generation unit 7 using the linear regression method. If the absolute value of this t value is 2 or more, it means that it is a significant index, and the larger the absolute value, the higher the correlation. It can be understood that the correlation between the self-heating property of the recycled fuel F (carbide) can be obtained in the following order of data input to the trained model 50: the gas temperature of the oxygen-containing gas measured by the gas temperature measurement unit St, the gas flow rate of the oxygen-containing gas measured by the gas flow rate measurement unit Sv, the temperature of the carbide M to be introduced into the storage tank 3 measured by the pre-introduction carbide temperature measurement unit Sc, and the internal temperature of the storage tank 3 measured by the tank temperature measurement unit T. It can be seen that among the internal temperatures of the storage tank 3 measured by the tank temperature measuring unit T, the lower internal temperature of the storage tank 3 (internal temperature of deposit B3) measured by temperature sensor T5, the lower wall temperature of the storage tank 3 (surface temperature of deposit B3) measured by temperature sensor T6, the middle wall temperature of the storage tank 3 (surface temperature of deposit B2) measured by temperature sensor T4, the upper wall temperature of the storage tank 3 (surface temperature of deposit B1) measured by temperature sensor T2, and the middle internal temperature of the storage tank 3 (internal temperature of deposit B2) measured by temperature sensor T3 can be correlated with the self-heating property of the recycled fuel F (carbide).

Figure 9 shows a comparison result between the predicted value (predicted oxygen consumption characteristics of recycled fuel F (carbonized material)) of the trained model 50 generated by the model generation unit 7 using a learning algorithm of a binary decision tree composed of CART (Classification And Regression Trees) and the operating history data of the carbonized material processing device 100 (actually obtained oxygen consumption characteristics of recycled fuel F (carbonized material)). Into this trained model 50, the gas flow rate of the oxygen-containing gas measured by the gas flow rate measuring unit Sv (average of the measured values of the gas flow rate measuring units Sv1 to Sv3), the gas temperature of the oxygen-containing gas measured by the gas temperature measuring unit St (average of the measured values of the gas temperature measuring units St1 to St3), the internal temperature of the storage tank 3 measured by the in-tank temperature measuring unit T (measurement values of the temperature sensors T2 to T6), and the temperature of the carbide M to be introduced into the storage tank 3 measured by the carbide temperature measuring unit Sc before introduction are input, and the evaluation values at each branch point are summed up to output a score (oxygen consumption characteristic) of the self-heating property of the recycled fuel F (carbide). As shown in FIG. 9, the correlation coefficient R ² between the predicted value and the actual value of the trained model 50 is 0.9 or more, so that it can be understood that the trained model 50 in this embodiment has extremely high accuracy.

Next, Example 2 showing the performance of the control device X will be described with reference to Figures 10 and 11. In Example 2, the purpose is to consider a combination of superior indicators among the input values of the trained model 50. As in Example 1, the model generation unit 7 selected input values to be used in the linear regression method to generate the trained model 50. In detail, a total of six trained models 50 were created: trained model I, in which only the gas temperature of the oxygen-containing gas was used as an input value; trained model II, in which the gas flow rate of the oxygen-containing gas was added to the input value of trained model I; trained model III, in which the temperature of the carbide M to be introduced into the storage tank 3 measured by the carbide temperature measurement unit Sc before introduction was added to the input value of trained model II; trained model IV, in which the upper wall temperature of the storage tank 3 measured by the temperature sensor T2 (surface temperature of deposit B1) was added to the input value of trained model III; trained model V, in which the lower internal temperature of the storage tank 3 measured by the temperature sensor T5 (internal temperature of deposit B3) was added to the input value of trained model IV; and trained model VI, in which the middle wall temperature of the storage tank 3 measured by the temperature sensor T4 (surface temperature of deposit B2) was added to the input value of trained model V.

FIG. 10 shows the regression coefficients, significance probability, and correlation coefficients R 2 between the predicted values and actual values of the trained models I to VI, which are obtained by performing multiple regression analysis on the trained models I to VI generated by the model generation unit ⁷ using the linear regression method. If the significance probability is lower than the significance level (for example, 5%), it can be said that each input value is related to the self-heating property of the recycled fuel F (carbide). In this embodiment 2, since the significance probability is below the significance level in all trained models, it can be understood that there is a significant correlation between each input value and the self-heating property in all trained models I to VI. In addition, the correlation coefficient R ² of the trained model I is 0.806, while the correlation coefficient R ² of the trained model II is 0.895, the correlation coefficient R ² of the trained model III is 0.903, and the correlation coefficient R ² of the trained models IV to VI is 0.904. It can be seen that the correlation coefficient R ² increases as the input value increases, and the accuracy of the trained model 50 increases. On the other hand, since the correlation coefficient ^R2 of the trained model II has a value close to 0.9, it was found that a highly accurate trained model 50 can be created using only two indicators, the gas temperature of the oxygen-containing gas and the gas flow rate of the oxygen-containing gas, which are input values of the trained model II. Furthermore, since the correlation coefficient ^R2 of the trained model III has a value exceeding 0.9, it was found that the three indicators, the gas temperature of the oxygen-containing gas, the gas flow rate of the oxygen-containing gas, and the input temperature of the recycled fuel F (carbide), which are input values of the trained model III, work particularly advantageously.

FIG. 11 shows a comparison result between the predicted value (predicted oxygen consumption characteristic of recycled fuel F (carbide)) of the trained model 50 generated by the model generation unit 7 using the learning algorithm of a binary decision tree composed of CART (Classification And Regression Trees) as in Example 1 using the index used as the input value of the trained model VI, and the operating performance data of the carbide processing device 100 (actually obtained oxygen consumption characteristic of recycled fuel F (carbide)). As shown in FIG. 11, since the correlation coefficient ^R2 is 0.9 or more, it can be understood that the trained model 50 in this embodiment has extremely high accuracy. Therefore, the self-heating property of the recycled fuel F (carbide) predicted by the trained model 50 can be used to determine whether the recycled fuel F can be used as fuel or to control the storage tank 3.

In this embodiment, the supply ports 34a, 34b, and 34c are provided at multiple locations along the vertical and circumferential directions of the side wall of the storage tank 3, so that oxygen-containing gas is supplied from multiple locations on the side wall, allowing the low-temperature oxidation process of the carbide M to proceed uniformly, and reliably suppressing the self-heating tendency of the recycled fuel F.

[Description of Modifications]
In the above-described embodiment, the learned model 50 generated by the model generation unit 7 is input with the following features as features: the gas flow rate of the oxygen-containing gas measured by the gas flow rate measurement unit Sv (average of the measured values of the gas flow rate measurement units Sv1 to Sv3), the gas temperature of the oxygen-containing gas measured by the gas temperature measurement unit St (average of the measured values of the gas temperature measurement units St1 to St3), the internal temperature of the storage tank 3 measured by the in-tank temperature measurement unit T (measurement values of the temperature sensors T2 to T6), and the temperature of the carbide M to be introduced into the storage tank 3 measured by the pre-introduction carbide temperature measurement unit Sc. The score of the self-heating property of the recycled fuel F generated by the learned model 50 is set as the prediction result of the prediction unit 8. As input data of the learned model 50 used by the prediction unit 8 for prediction, the oxygen concentration of the processing gas G supplied from the ventilation unit 4, the oxygen concentration of the exhaust gas E measured by the oxygen concentration measurement unit So, or the temperature of the recycled fuel F before shipment (temperature before shipment) measured by the pre-shipment carbide temperature measurement unit Sf may be used.

The gas flow rate of the oxygen-containing gas input to the trained model 50 may be the measurement value of each of the gas flow rate measurement units Sv1 to Sv3 in addition to or instead of the average measurement value of the gas flow rate measurement units Sv1 to Sv3, or the variance of the measurement value of the gas flow rate measurement units Sv1 to Sv3, the average or variance of any two of the gas flow rate measurement units Sv1 to Sv3, or the sum of the measurement values of the gas flow rate measurement units Sv1 to Sv3 or the integrated value over a certain period of time may be used as the feature value. The gas temperature of the oxygen-containing gas input to the trained model 50 may be the measurement value of each of the gas temperature measurement units St1 to St3 in addition to or instead of the average measurement value of the gas temperature measurement units St1 to St3, or the variance of the measurement value of the gas temperature measurement units St1 to St3, the average or variance of any two of the gas temperature measurement units St1 to St3, or the measurement value of one temperature sensor provided in the flow path before the air supply pipe 41a branches off may be used as the feature value. Furthermore, the internal temperature of the storage tank 3 input to the trained model 50 may be the average or variance of the temperature sensors T1 to T6 as a feature in addition to or instead of the measured values of the temperature sensors T1 to T6, or may be the measured value of at least one of the temperature sensors T1 to T6.

In the above-described embodiment, the actual data used by the model generation unit 7 to generate the trained model 50 are the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, the temperature of the carbide M input into the storage tank 3, the temperature after cooling of the recycled fuel F discharged from the storage tank 3, and the oxygen consumption characteristics of the recycled fuel F discharged from the storage tank 3. These actual data may include at least one of the gas flow rate of the oxygen-containing gas, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank 3, the temperature of the carbide M input into the storage tank 3, and the temperature after cooling of the recycled fuel F discharged from the storage tank 3, and the oxygen consumption characteristics of the recycled fuel F discharged from the storage tank 3. In addition, the oxygen concentration of the processing gas G supplied from the ventilation unit 4 and the oxygen concentration of the exhaust gas E measured by the oxygen concentration measurement unit So may be used as the actual data. Furthermore, the model generation unit 7 may generate the trained model 50 using the average or variance of each actual data as a feature.

In this way, it is possible to provide a self-heating prediction method, a trained model 50, and a control device X that can accurately predict the self-heating property of the recycled fuel F (carbide) discharged from the storage tank 3.

[Another embodiment]
(1) In the above embodiment, the case where nitrogen is used as the second treatment gas G2 is exemplified. However, the second treatment gas G2 may be an inert gas such as exhaust gas E or combustion exhaust Ef generated (discharged) in the organic sludge recycle system 200, or carbon dioxide, or may be introduced from other industrial equipment, such as exhaust gas from a waste incinerator or exhaust gas from a thermal power plant. In addition to nitrogen, an inert gas such as exhaust gas E or combustion exhaust Ef generated (discharged) in the organic sludge recycle system 200 or carbon dioxide may be introduced as necessary. The second treatment gas G2 may be mainly composed of an inert gas such as nitrogen or carbon dioxide (for example, 92 volume percent or more). The second treatment gas G2 may have a lower oxygen concentration than air, and preferably has less than 8 volume percent oxygen. In addition, the second treatment gas G2 may be omitted, and only the first treatment gas G1 may be used for the carbide treatment.

(2) In the above embodiment, a rotary valve 32a is used as a discharge device for discharging the recycled fuel F from the storage tank 3, and an example is given of a case in which the recycled fuel F is discharged in a state in which it is isolated from the downstream atmosphere. However, the discharge device may be a device that can be used according to the shape of the recycled fuel F, such as pellets, and may be a device or configuration that can discharge the recycled fuel F from the internal space of the storage tank 3 in a state in which the downstream atmosphere does not flow into the storage tank 3 without restriction.

For example, instead of the rotary valve 32a, a double damper, which is an exhaust device that can exhaust the storage tank 3 while physically isolating it from the downstream atmosphere, may be used. Alternatively, a screw-type exhaust machine or a table feeder-type exhaust device may be used, and the processing gas G and the second processing gas G2 may be passed through the exhaust port in a countercurrent manner to maintain a state in which the downstream atmosphere does not flow freely into the storage tank 3.

(3) In the above embodiment, a rotary valve 32a is provided on the discharge pipe 32b that is provided downward from the end of the lower container, which is the lower end of the storage tank 3, and recycled fuel F is discharged below the storage tank 3. However, the discharge device is not limited to a configuration that discharges recycled fuel F below the storage tank 3. For example, a discharge port of a screw-type discharger or a table feeder-type discharge device may be provided on the side of the storage tank 3, and recycled fuel F may be discharged from the side of the storage tank 3.

(4) In the above embodiment, the gas supply port 34 for the processing gas G is provided at multiple locations (three locations in this embodiment) along the vertical direction of the side wall of the storage tank 3, and each of the supply ports 34a, 34b, and 34c is provided at multiple locations (four locations in this embodiment) along the circumferential direction of the side wall of the storage tank 3. However, the gas supply port 34 for the processing gas G may be provided only at the lower part of the side wall of the storage tank 3, or at two or four or more locations along the vertical direction of the side wall, and each of the supply ports 34a, 34b, and 34c may be provided only at one location on the side wall of the storage tank 3, or at two, three, five or more locations along the circumferential direction of the side wall. In addition, the gas supply port 34 for the processing gas G may be extended to near the center of the storage tank 3, and the processing gas G may be blown into near the center of the storage tank 3.

(5) The trained model 50 may be generated by machine learning other than decision trees or linear regression analysis, or by deep learning other than convolutional neural networks. For example, known learning methods such as support vector machines and logistic regression can be used.

The configurations disclosed in the above embodiment (including other embodiments, the same applies below) can be applied in combination with configurations disclosed in other embodiments, so long as no contradiction arises. Furthermore, the embodiments disclosed in this specification are merely examples, and the present invention is not limited to these embodiments. They can be modified as appropriate within the scope of the purpose of the present invention.

The present invention can be applied to a self-heating prediction method for predicting the self-heating properties of charcoal cured in a storage tank, a trained model, and a self-heating control device for controlling the self-heating properties of charcoal cured in a storage tank.

3: Storage tank 8: Prediction unit 9: Control unit 34: Gas supply port (supply port)
50: Trained model F: Recycled fuel (carbide)
G: Processing gas M: Carbonized material Sc: Carbonized material temperature measurement unit before charging St: Gas temperature measurement unit Sv: Gas flow rate measurement unit T: Thermometer side unit inside tank X: Control device (self-heating control device)

Claims (7)

A self-heating prediction method implemented by a computer for predicting the self-heating property of a charcoal cured in a reservoir, comprising the steps of:
A self-heating prediction method that predicts the self-heating property based on the output value of a trained model to which at least one of the gas flow rate of the oxygen-containing gas supplied to the storage tank, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal put into the storage tank is input. The self-heating prediction method according to claim 1, wherein at least the gas flow rate, the gas temperature, and the internal temperature are input to the trained model. The self-heating prediction method according to claim 1 or 2, wherein the oxygen-containing gas is supplied to the storage tank from multiple locations on the side wall. A trained model that operates on a computer,
It is composed of a decision tree consisting of multiple branching points arranged in a tree structure.
A trained model that receives as input at least one of the gas flow rate of an oxygen-containing gas supplied to a storage tank for curing charcoal, the gas temperature of the oxygen-containing gas, the internal temperature of the storage tank, and the temperature of the charcoal put into the storage tank, and outputs a score of the self-heating property of the charcoal by adding up the evaluation values at each of the branching points. The trained model according to claim 4, in which at least the gas flow rate, the gas temperature, and the internal temperature are input. A self-heating property control device for controlling the self-heating property of a carbonized material cured in a storage tank,
a gas flow rate measuring unit that measures a gas flow rate of the oxygen-containing gas supplied to the storage tank;
a gas temperature measuring unit that measures a gas temperature of the oxygen-containing gas supplied to the storage tank;
an in-tank temperature measuring unit that measures an internal temperature of the carbonized material stored in the storage tank;
a pre-injection carbonized material temperature measuring unit that measures a pre-injection temperature of the carbonized material before it is introduced into the storage tank;
a prediction unit that predicts the self-heating property based on at least one of the gas flow rate, the gas temperature, the internal temperature, and the pre-insertion temperature;
A control unit that controls at least one of the gas flow rate, the gas temperature, the residence time of the carbonized material in the storage tank, and the pre-injection temperature based on the prediction result of the prediction unit. The self-heating control device according to claim 6 , wherein the oxygen-containing gas supply ports are provided at a plurality of locations on the side wall of the storage tank.

Images