TWI705316B - Boiler operation support device, boiler operation support method, and boiler learning model creation method - Google Patents

Abstract

For factories, comprehensively consider various viewpoints such as economy, safety, and equipment maintenance to predict, adjust or instruct operations. Obtain the actual operating conditions suitable for the actual operation of the factory, and the actual process value obtained after the results of the operation are used, and use the learning model obtained by mechanically learning the relationship between the actual operating conditions and the actual process value to calculate the predicted process value , And calculate the optimal operating condition that the predicted process value meets the predetermined evaluation condition. The actual process value includes: for example, the main control process value formed by the quality index value of the final product produced in the factory and the index related to the environmental regulation value; and for example, the temperature or pressure of the machine in the factory Peripheral process value based on indicators, indicators related to components and concentrations of gases, liquids, or solids discharged from machines in the factory that are not subject to environmental regulatory values, and indicators related to the opening of the operating end of the factory.

Description

Boiler operation support device, boiler operation support method, and boiler learning model creation method

The present invention relates to the operation support technology of factories.

In the operation of power plants, especially large-scale boilers, many input parameters are operated as operating conditions, such as the opening of the damper to adjust the combustion air flow rate in each burner, the angle of the burner nozzle, coal, etc. Process values such as the concentration of NOx and CO, the surface temperature of the heat transfer tube (metal temperature), and the vapor temperature of the solid fuel pulverizer, etc., are obtained as the resultant output (monitoring item). In the combustion adjustment of the boiler, the input parameters must be controlled in such a way that the process values are within the appropriate range. However, the input parameters have a majority of more than 10 items, and relative to the change of the input parameters, each process value is obtained as a result of complex correlation. Therefore, if the process value is improved or deteriorated, the operating system of the input parameter must be very complicated Order.

Therefore, in a large boiler, according to the test operation (combustion As a result of adjustment), an optimized control logic is set in a specific condition, and the input parameters are controlled according to the control logic. However, it may not be able to cope with minute changes in the condition of the equipment or fuel, and may not be optimally operated.

Therefore, it is desirable to use input parameters to simulate the combustion operation of the boiler in advance for optimal operation, and to use the result to perform automatic operation of the boiler. Patent Document 1 discloses a model construction data for correcting a simulation of a plant, and a configuration for performing boiler control based on the result. In addition, the model output (optimization target) system exemplifies the NOx, CO, and H ₂ S concentrations contained in the exhaust gas.

[Prior Technical Literature] [Patent Literature]

[Patent Document 1] JP 2011-210215 A

The boiler system is not only the main control object (values related to the boiler outlet steam temperature, the environmental regulation value of the exhaust gas, etc.), it must be controlled according to various characteristics and comprehensively considering various factors such as economy, safety, and equipment maintenance. Regarding this point, in Patent Document 1, even if the environmental regulation value can be corresponded to the main control object, it does not describe the consideration of other factors, and does not meet the above expectations.

The present invention solves the above-mentioned problems, and aims to provide Boiler, the operation support device of the boiler, the operation support method of the boiler, and the creation of the learning model of the boiler that can predict, adjust or instruct the operation of various viewpoints such as economy, safety, equipment maintenance, etc. method.

In order to achieve the above purpose, it has the composition described in the scope of the patent application. If one example is given, it is an operation support device for a boiler, which is characterized by having: a data acquisition unit that obtains the actual operating conditions suitable for the actual operation of the boiler, and applies the actual operating conditions to operate the aforementioned boiler The actual process value obtained as a result; the model memory unit, which memorizes the learning model obtained by mechanically learning the relationship between the foregoing actual operating conditions and the foregoing actual process value; the operating data memory unit, which memory includes the foregoing actual operating conditions and the foregoing Operating data of actual process values; and an operating condition calculation unit, which calculates the optimal operating conditions under which the predicted process values calculated by applying the foregoing operating data to the memorized learning model meets the predetermined evaluation conditions, and the foregoing actual process values include: Regarding both the main control process value of the main control object of the boiler and the peripheral process value related to the peripheral information, the main control process value is the value of the quality index of the final product produced by the boiler and the relevant environmental regulation value Any one or a combination of indicators, the aforementioned peripheral process value is an indicator related to the temperature or pressure of the equipment in the boiler, and the gas, liquid or solid discharged from the equipment in the boiler does not become an environmental regulatory value Index of the composition and concentration of the object Any one or any combination of the indicators related to the opening of the operating end of the aforementioned boiler.

According to the present invention, it is possible to predict, adjust or instruct the operation of the boiler by comprehensively considering various viewpoints such as economy, safety, and equipment maintenance. The problems, constitution, and effects other than the above are clearly clarified from the description of the following embodiments.

Hereinafter, a more suitable embodiment of the present invention will be explained in detail with reference to the attached drawings. However, the present invention is not limited by this embodiment, and if there are a plurality of embodiments, it also includes those constructed by combining each embodiment. The following is an example of a boiler installed in a thermal power plant as an example of a factory.

1, the schematic configuration of the operation support device 100 of the boiler 1 will be described. FIG. 1 is an overall configuration diagram of an operation support device 100 using a predictive model.

As shown in Figure 1, the boiler 1 includes: M sensors 1, 2, ..., M, and N operating terminals 1, 2, ..., N.

The operation control device 120 is connected to each of the N operation terminals 1, 2, ..., N, and input parameters (corresponding to actual input parameters) constituting operating conditions are set for each operation terminal 1, 2, ..., N. The input parameters include, for example, at least one of the opening degree of the air choke, the air flow rate, the fuel flow rate, and the exhaust gas recirculation flow rate.

FIG. 2 is a schematic configuration diagram of the boiler 1. The boiler 1 of this embodiment uses pulverized coal as a pulverized coal fuel (solid fuel) as a solid fuel combustor. The pulverized coal is burned by the burner of the furnace 11, and the combustion A coal-fired boiler that generates steam by exchanging the generated heat with water or steam. Among them, the fuel is not limited to coal, and may be other fuels such as biomass that can be burned in a boiler. It is also possible to mix and use multiple fuels.

The boiler 1 has: a furnace 11, a combustion device 12, and a flue 13. The stove 11 is, for example, in the hollow shape of a quadrangular tube and is installed along the vertical direction. The wall surface of the furnace 11 is composed of an evaporation tube (heat transfer tube) and a fin connected to the evaporation tube. The water or steam circulating in the evaporation tube exchanges heat with the combustion gas in the furnace to suppress the temperature rise of the furnace wall . Specifically, on the side wall surface of the furnace 11, for example, a plurality of evaporation tubes are arranged along the vertical direction and arranged side by side in the horizontal direction. The heat sink system blocks the evaporation tube and the evaporation tube. The furnace 11 is provided with an inclined surface 62 on the furnace bottom, and a furnace bottom evaporation tube 70 is provided on the inclined surface 62 to form a bottom surface.

The combustion device 12 is provided on the vertical lower side of the furnace wall constituting the furnace 11. In this embodiment, the combustion device 12 has a plurality of burners (for example, 21, 22, 23, 24, 25) installed on the furnace wall. For example, the burners 21, 22, 23, 24, 25 are arranged in plural along the circumferential direction of the furnace 11 at equal intervals. However, the shape of the furnace, the arrangement of the burners, or the number of burners in one layer and the number of layers are not limited to those of this embodiment.

The burners 21, 22, 23, 24, 25 are connected to pulverizers (pulverizers/mills) 31, 32, 33, 34, 35 through pulverized coal supply pipes 26, 27, 28, 29, 30 . Coal is transported by a transport system not shown. If it is fed into the pulverizers 31, 32, 33, 34, 35, where it is crushed into a predetermined fine powder size, it can be combined with transport air (primary air). The pulverized coal supply pipes 26, 27, 28, 29, and 30 supply pulverized coal (pulverized coal) to the burners 21, 22, 23, 24, and 25.

In addition, the stove 11 is provided with a wind box 36 at the installation position of the burners 21, 22, 23, 24, and 25. One end of the air duct 37b is connected to the wind box 36, and the other end is secured at the connecting point 37d. It is connected to an air duct 37a that supplies air.

In addition, a flue 13 is connected vertically above the furnace 11, and a plurality of heat exchangers (41, 42, 43, 44, 45, 46, 47) for generating steam are arranged in the flue 13. Therefore, the burners 21, 22, 23, 24, and 25 inject a mixture of pulverized coal fuel and combustion air in the furnace 11 to form a flame, generate combustion gas, and flow to the flue 13. Next, the combustion gas heats the water or steam flowing on the furnace wall and the heat exchangers (41 to 47) to generate superheated steam, and the generated superheated steam is supplied to drive a steam turbine not shown in the figure to rotate. A generator (not shown) connected to the rotating shaft of the steam turbine is driven to generate electricity. In addition, the flue 13 is connected to the exhaust gas passage 48, and is provided with a denitrification device 50 for purifying the combustion gas, air sent to the air duct 37a by the blower 38, and exhaust gas sent to the exhaust gas passage 48 The air heater 49, the coal dust processing device 51, the suction blower 52, etc., which exchange heat therebetween, are provided with a chimney 53 at the downstream end. Among them, the denitration device 50 may not be provided if it can meet the emission gas standard.

The stove 11 of this embodiment is a so-called two-stage combustion type stove, which is made up of air for conveying pulverized coal (primary air) and combustion air (secondary air) fed into the stove 11 from a wind box 36 After the excess fuel is burned, the combustion air (after air) is re-introduced to make the fuel lean burn. Therefore, the stove 11 is equipped with a burn-out tuyere 39, one end of the air duct 37c is connected to the burn-out tuyere 39, and the other end is connected to the air duct 37a for supplying air at the connection point 37d. Among them, if the two-stage combustion method is not adopted, the burn-out tuyere 39 may not be provided.

The air sent to the air duct 37a by the blower 38 is divided into two parts: the air heater 49 is heated by heat exchange with the combustion gas, and is guided to 2 of the wind box 36 through the air duct 37b at the connection point 37d. Secondary air; and through the air duct 37c and led to the burn-out air outlet 39.

Returning to FIG. 1, the operation support device 100 will be described. The operation support device 100 mainly includes: a data acquisition unit 110, an operation data storage unit 130, a cumulative value calculation unit 131, a data extraction conversion unit 133, an RTC (Real-Time Clock) 140, a factory specification storage unit 211, and a process The value candidate storage unit 212, the operation condition calculation unit 220, the model creation unit 231, the evaluation condition review unit 232, the evaluation condition storage unit 233, the model storage unit 241, the operation instruction unit 250, and the input/output unit 260.

The data acquisition unit 110 acquires the actual process values measured by the sensors 1, 2,..., M during actual operation, and the settings of the operation control device 120 at each of the operation terminals 1, 2,..., N Actual input parameters. In addition, the data acquisition unit 110 adds the time information from the RTC 140 to each of the actual input parameters and actual process values, and outputs them to the operation data storage unit 130.

The actual process values include, for example, the nitrogen oxide concentration, the carbon monoxide concentration, the hydrogen sulfide concentration, and the metal temperature contained in the gas discharged from the thermal power plant. In addition, the actual input parameters include operating terminal information such as valve/choke opening.

In the present embodiment, at least one or more actual input parameters applicable to the actual operation of the boiler 1 are collectively referred to as operating conditions. On the other hand, in the operation support device 100, the process value calculated by performing the operation simulation of the boiler 1 using the virtual set operating conditions (temporary input parameters) is called the predicted process value.

The cumulative value calculation unit 131 calculates the cumulative value of at least one actual process value acquired by the data acquisition unit 110 and stores the cumulative value in the operation data storage unit 130.

The data extraction and conversion unit 133 is interposed between the model creation unit 231, the operation condition calculation unit 220, and the operation data storage unit 130. The operation data extracted by the operation data storage unit 130 is converted, if necessary, such as noise removal, etc. In each of the model creation unit 231 and the operating condition calculation unit 220, acceptance is performed.

The factory specification storage unit 211 stores factory specification data indicating the specifications of the boiler 1 input by the input/output unit 260.

Figure 3 is a diagram showing an example of factory specification data. In the factory specification data, the structural specifications and performance specifications of each of the factories A, B, and C are specified. As an example of structural specifications, there is "stove size". In addition, in terms of "performance specifications", there are "gas temperature" and "vapor temperature".

The process value candidate storage unit 212 stores data representing the process value candidates input by the input/output unit 260.

When describing the process value candidates, the description will be based on the type of process value used in this embodiment. The process values used in this embodiment include: the main control process value of the main control object of the boiler 1, and the peripheral process value of the peripheral information. The main control process value uses the measured value. Therefore, if the sensor that measures the main control process value fails, the boiler 1 is based on the principle of stopping operation. However, among the NO _X sensors that measure the NO _X concentration as one of the main control process values, sensors other than those installed at the chimney entrance of the boiler 1 can continue to operate even in the event of a failure.

The main control process value is any one or a combination of the following. (1) The quality index value of the final product produced in the factory (2) Indicators related to environmental regulation values

In addition, the peripheral process value is any one or a combination of the following. (3) Indicators related to the temperature or pressure of the machines in the factory (4) Indicators on the components and concentrations of gases, liquids, or solids discharged from equipment in factories that are not subject to environmental regulatory values (5) Indicators related to the opening of the operating side of the factory

In this embodiment, boiler 1 is used as the factory, so the boiler outlet steam temperature is used as the main control process value (1) the quality index of the final product, and the NO _x value is used as (2) the environmental regulation value And indicators of external regulatory values. In addition, in the peripheral process value, the surface temperature of the heat transfer tube and the boiler pressure difference are used as (3) indicators of the temperature or pressure of the equipment in the factory, and the oxygen concentration in the combustion air or exhaust gas is used as ( 4) Use the spray valve opening as an index for the components and concentrations of the gas, liquid or solid discharged from the equipment in the factory that are not subject to environmental regulation values as (5) the index for the opening of the operating valve . Among them, in addition, the amount of spray can also be used as (3) the temperature or pressure of the equipment in the factory, and the carbon monoxide concentration is used as (4) the amount of gas, liquid or solid discharged from the equipment in the factory. An index of the component and concentration that becomes the target of the environmental regulation value.

The selection and purpose of peripheral parameters are as follows. By using the metal (heat transfer tube) temperature as the temperature, operation support for the purpose of the combustion characteristics (balance), safety, durability, and maintenance of the boiler 1 can be performed. In addition, by using the boiler pressure difference as the pressure, operation support in consideration of the safety of the operation of the boiler 1 can be performed. By using the oxygen concentration of combustion air or exhaust gas as the gas component concentration, operation support can be performed in consideration of the combustion characteristics (balance) and efficiency of the boiler 1. In addition, by using the spray valve opening degree as the valve opening degree, the function of the valve of the boiler 1 (the normal use range of the valve opening degree), the response of the coal characteristics (fouling of the furnace, etc.), and the heat absorption distribution can be performed. (Balance, each heat transfer surface) operation support.

The peripheral process value system includes: required peripheral process value, and any peripheral process value.

The required peripheral process value is in principle the peripheral process value selected as the operating data when the model is created. However, in the specifications of the boiler 1, cases where there are no measured values or target equipment are excluded.

The arbitrary peripheral process value is the peripheral process value arbitrarily selected when the model is created. Any peripheral process value is selected from the process values that are the target of the alarm if the actual process value in the boiler 1 indicates an abnormal value. Although the model creation of the boiler 1 is not essential, the process value to be the alarm target of the boiler 1 is selected as an arbitrary peripheral process value, and the model is created, thereby suppressing fluctuations in the optimal stage before the alarm is issued. As a result, the boiler 1 can be expected to suppress the operation support of the alarm.

Figure 4 is a diagram showing candidate data for process values. In the process value candidate data, the types of process value candidates including the main control process value and peripheral process values, the data acquisition method of each process value candidate, and the attributes of each process value candidate (entry in the remarks column) are linked to each other. Be prescribed. In the "data acquisition method", it is stipulated to obtain the measured value at a certain point in time or set it as a cumulative value. The attribute entered in the "Remarks" column describes the process value candidates that must be selected or arbitrarily selected at the time of model creation, and their reasons.

The model creation unit 231 mechanically learns the relationship between the operating conditions of the boiler 1 and the process value to create a learning model, and the learning model is memorized in the model memory unit 241.

The model creation unit 231 refers to the factory specification data and the process value candidate data, and arbitrarily selects peripheral process values according to the factory specifications or user requirements from the operation data obtained by the data extraction conversion unit 133.

FIG. 5 is a diagram showing an example of a learning model created by the model creating unit 231. The learning model is set according to each actual process value. In this embodiment, the main control process values that become the main control target of the boiler 1 include "boiler outlet steam temperature" and "exhaust gas environmental regulation value", and the model creation unit 231 creates corresponding main control processes Value learning model 1, learning model 2. In addition, in terms of the peripheral process values constituting the peripheral information of the boiler 1, there are "temperature in the boiler 1 (metal temperature)", "pressure in the boiler 1 (boiler pressure difference)", "gas component concentration (combustion air) Or oxygen concentration of exhaust gas)” and “valve opening (spray valve opening)”, the model creation unit 231 creates learning model 3, learning model 4, learning model 5, and learning model 6 corresponding to the peripheral process values.

The operating condition calculation unit 220 includes a simulation unit 221 and an optimization unit 225.

The simulation unit 221 is a learning model memorized in the model storage unit 241, applies the operating data acquired by the data extraction and conversion unit 133 to perform simulation, and calculates the predicted process value.

The optimization unit 225 uses the predicted process value calculated by the simulation unit 221 to calculate the recommended operating conditions. If the difference between the calculated recommended operating conditions and the actual operating conditions included in the actual operating data exceeds a predetermined standard, the operating instruction unit 250 is presented with additional acquisition of operating data for model learning.

The evaluation condition review unit 232 updates the second evaluation condition (optimum range) of the peripheral process value based on the instruction of the operator input from the input/output unit 260 and/or the operating data of the boiler 1. The details will be described later.

The operation instruction unit 250 outputs the operation condition to the operation control device 120 when the operation condition is acquired by the optimization unit 225 or the input/output unit 260.

The operation control device 120 sets the operation conditions acquired by the operation instruction unit 250 to the operation terminals 1, 2, ..., N. In addition, the operation instruction unit 250 may also output the operation conditions acquired by the optimization unit 225 to the input/output unit 260. Next, the operating conditions acquired by the optimization unit 225 may be displayed on the display device constituting the input/output unit 260.

The input/output unit 260 is composed of an input device (equivalent to an input unit) such as a mouse, a keyboard, and a touch panel, and a display device made of an LCD or the like. The input device and the display device can also be integrally formed. The input/output unit 260 displays instructions from the operation instruction unit 250 and waits for instructions from the operator.

FIG. 6 is a diagram showing the hardware configuration of the operation support device 100. The operation support device 100 includes: CPU (Central Processing Unit) 301, RAM (Random Access Memory) 302, ROM (Read Only Memory) 303, HDD (Hard Disk) Drive (hard disk drive) 304, input I/F305, and output I/F306 are constructed using these computers connected to each other through a bus 307. Among them, the hardware configuration of the operation support device 100 is not limited to the above, and may be constituted by a combination of a control circuit and a memory device. In addition, the operation support device 100 is configured by a computer (hardware) executing an operation support program that realizes each function of the operation support device 100.

FIG. 7 is a flowchart showing the flow of the learning model creation process performed by the operation support device 100.

The model creation unit 231 reads the operation data acquired by the data extraction and conversion unit 133 (S101). The model creation unit 231 reads the factory specification data conforming to the boiler 1 from the factory specification storage unit 211 (S102).

In addition, the model creation part 231 reads the process value candidates from the process value candidate memory part 212 (S103), and selects the customized process value (S104).

The model creation unit 231 selects explanatory variables (input parameters) from the operating data acquired by the data extraction and conversion unit 133 (S105), and creates a learning model (S106). Specifically, the model preparation unit 231 inputs all selected explanatory variables for each learning model according to each process value, that is, learning model 1 to learning model 6, and calculates each process value. The calculated value is equivalent to the predicted process value.

The model preparation unit 231 compares the actual process value contained in the operating data obtained by the data extraction and conversion unit 133 with the predicted process value calculated in step S106. If the error is within the allowable range, the predicted process value can be regarded as It is roughly the same as the actual process value. Among them, the data for a predetermined time is accumulated and the tendency of errors is also considered. There are cases where errors occur only in a short period of time due to temporary reasons. If the error is within the allowable range, it is judged that the learning model is appropriate (S107/Yes), the created learning model is memorized in the model storage unit 241, and the process ends.

The model preparation unit 231 may also set the predicted process value and the actual process value obtained from all the learning models as conditions within the allowable range, as conditions for judging the suitability of the learning model.

FIG. 8 is a diagram showing the magnitude of the first allowable error and the second allowable error. The allowable error for the main control process value (the first allowable error) is smaller than the allowable error for the peripheral process value (the second allowable error). As a result, the learning model corresponding to the main control process value can more follow the boiler 1 and set the appropriateness judgment conditions.

If the model creation unit 231 determines that the learning model is not appropriate (S107/No), it returns to step 106 and uses the operating data obtained by the data extraction and conversion unit 133 and the process value candidates stored in the process value candidate memory unit 212 , Revise/update the learning model by statistical methods represented by neural networks. If the updated learning model is appropriate (S107/Yes), it is memorized in the model memory 241.

FIG. 9 is a flowchart showing the flow of an operation support method using a predictive model by the operation support device 100.

The simulation unit 221 reads the learning model from the model storage unit 241, and the optimization unit 225 reads the evaluation condition (the scoring conversion criterion described later) from the evaluation condition storage unit 233 (S201). In addition, the simulation unit 221 sets the operating data acquired by the data extraction and conversion unit 133 as simulation conditions (S202). The simulation unit 221 applies the actual input parameters included in the operating data to the learning model to perform simulation (S203), and calculates the predicted process value.

The simulation unit 221 outputs the predicted process value to the optimization unit 225, and the optimization unit 225 scores and converts the simulation result for evaluation (S204). The optimization unit 225 converts the predicted process value into a score based on the score conversion criteria shown in FIGS. 10 and 11 to determine whether the simulation result is optimal.

Figures 10 and 11 are diagrams showing an example of a scoring conversion criterion. The scoring system of the vertical axis of FIG. 10 and FIG. 11 is such that the dashed line indicates a positive value on the paper surface direction, and the dashed line indicates a negative value on the paper surface direction. The absolute value of the coefficient of the allowable range is set to a value larger than the absolute value of the target range. That is, the slope of the scoring conversion line of the allowable range is set to be larger than the slope of the scoring conversion line of the target range.

Fig. 10 is a scoring conversion standard defined for the process value for the purpose of minimization, and the upper limit is set, and the target value formed by a value smaller than it is set. The range smaller than the target value is set as the target range, and a coefficient formed by a positive value is assigned. The range from the target value to the upper limit is set as the allowable range, and coefficients formed by negative values are assigned. A range larger than the upper limit value is set as a non-allowable range, and a coefficient formed by a negative value having an absolute value larger than the absolute value of the coefficient of the allowable range is allocated. That is, the slope of the scoring conversion line of the non-allowable range is set to be greater than the slope of the scoring conversion line of the allowable range.

Fig. 11 is a scoring conversion standard defined for the process value for maximization, setting the lower limit value and the target value formed by a value larger than it. The range larger than the target value is set as the target range, and a coefficient formed by a positive value is assigned. The range from the target value to the lower limit is set as the allowable range, and a coefficient formed by a negative value is assigned. A range smaller than the lower limit value is set as an unallowable range, and a coefficient formed by a negative value having an absolute value larger than the absolute value of the coefficient of the allowable range is allocated. That is, the slope of the scoring conversion line of the non-allowable range is set to be greater than the slope of the scoring conversion line of the allowable range.

The optimization unit 225 uses the following formula (1) to set the target value as shown in FIG. 10 and the upper limit value formed by a value larger than that for each predicted process value, and set the target as shown in FIG. 11 The value and the lower limit value formed by the smaller value are calculated using the following formula (2) to calculate the score of each predicted process value (S106). SAi=CAi×(upper limit-predicted process value)...(1) SAi=CAi×(predicted process value-lower limit)...(2) among them, SAi: Scoring of predicted process value Ai of test number i CAi: the coefficient assigned to predict the process value Ai

The evaluation condition review unit 232 changes the evaluation condition by changing the slope and inflection point of the scoring conversion standard. The change of the evaluation conditions is to review the evaluation conditions (both the non-allowable range and the target range of the score conversion standard) according to the deterioration (corrosion) of the boiler 1 or the degree of influence of the surrounding process values, so that it can be carried out according to actual conditions. Operation support with high accuracy of the state is performed. Among them, the first evaluation condition formed by the target range of the main control process value is not updated as soon as it is set in principle. In particular, the non-allowable range of the scoring conversion criteria has not changed. It is based on factory specifications or regulations. However, the target range may change during operation mode setting (NO _X priority, etc.).

The optimization unit 225 judges whether the simulation result satisfies the optimal condition based on the calculated score. The optimization unit 225 may perform one simulation, add up the scores for all the predicted process values obtained by using the model 1 to the model 7, and use the total value as the simulation result. If there is at least one simulation result above the optimal scoring threshold set in advance for judging that the total value of scoring is optimal (S205/Yes), the optimization unit 225 selects the optimal condition (S206) and performs The instruction unit 250 outputs the selected optimal condition and ends the process.

Optimum here can also mean that each predicted process value (NO _X value or steam temperature, etc.) is converted into a score (non-dimensional) with a predetermined conversion factor, and the total value of the score becomes a predetermined value or more. Alternatively, it can also be formed as a simulation with a plurality of cases (simulation conditions), and among these results, when the score is the highest, or among the cases in the upper digits, the operator judges the most appropriate time. In addition, genetic algorithms or particle swarm optimization techniques can also be used to automatically explore and score higher cases, and the results can be used to determine whether they are optimal.

In addition, the optimizing part 225 can also make the main control process value meet the first evaluation condition and meet the second evaluation condition than the peripheral process value, that is, satisfying the main control process value is included in the target range (the first evaluation condition) is better than meeting the surrounding process value. The process value included in its target range (the second evaluation condition) is given priority to calculate the optimal operating conditions.

The so-called "priority" here means that the process value of the main control object must be prevented from exceeding the allowable range, and in addition, the peripheral process value is brought close to the allowable range or target range. Specifically, in the scoring conversion criterion, the main control parameter may maximize the negative slope of the non-allowable range.

In addition, the optimization unit 225 may also calculate an optimal operating condition in which the cumulative value of the predicted process value satisfies the predetermined evaluation condition. Among the peripheral process values, the temperature of the machine can be converted into an index that can accurately evaluate the temperature history and other deterioration by using the accumulated value. At this time, in the cumulative value of the past process value (operation data), the process value (predicted value obtained by the learning model) is added to calculate the predicted cumulative value. The scoring evaluation is performed using the scoring conversion basis of the predicted cumulative value.

If there is no simulation result that satisfies the optimal condition (S205/No), the optimization unit 225 changes the simulation condition to the simulation unit 221, and outputs an instruction to perform the simulation again. In response to this, the simulation unit 221 changes the temporary input parameters included in the simulation conditions and executes the simulation again.

The effect of this embodiment will be explained. With this embodiment, the process value used in the operation support of the factory is not only related to the main control process value of the main control object of the factory, but by using peripheral process values, the economy, safety, and equipment maintenance of the factory can be comprehensively considered Forecast, adjust, or instruct operations based on various viewpoints.

In addition, according to the present embodiment, the operating condition calculation unit 220 includes a simulation unit 221 and an optimization unit 225. The optimization unit 225 evaluates the simulation results to select the optimal operating conditions, or the optimization unit 225 compares the simulation unit 221 It urges another simulation, and can efficiently calculate the optimal operating conditions according to a series of processes.

In addition, with this embodiment, the optimal operating condition is calculated by making the first evaluation condition that the main control process value meets priority over the second evaluation condition that the peripheral process value should meet, and can reflect the importance of the main control process value. Optimal operating conditions.

In addition, with this embodiment, when the learning model is created, process value candidates are prepared in advance, so that the most suitable process value corresponding to the factory specifications can be selected efficiently. In addition, by limiting the objects (process values) to be made into the learning model, the learning model can be easily used.

The above-mentioned embodiment is not intended to limit the present invention, and there are various modifications within the scope not departing from the gist of the present invention. For example, in the operation support device 100 described above, the correction of the learning model is performed when the error exceeds a predetermined range, but in addition to this, the correction may be performed periodically.

1‧‧‧Boiler 11‧‧‧Stove 12‧‧‧Combustion device 13‧‧‧ Flue 21, 22, 23, 24, 25‧‧‧Burner 26, 27, 28, 29, 30‧‧‧Pulverized coal supply pipe 31, 32, 33, 34, 35‧‧‧Crusher (pulverizer/mill) 36‧‧‧Blowers 37a‧‧‧Air duct 37b‧‧‧Air duct 37c‧‧‧Air duct 37d‧‧‧Connecting point 38‧‧‧Air blower 39‧‧‧Out of Burning Outlet 41, 42, 43, 44, 45, 46, 47‧‧‧ Heat exchanger 48‧‧‧Exhaust gas passage 49‧‧‧Air heater 50‧‧‧Denitration device 51‧‧‧Coal dust treatment device 52‧‧‧Induction blower 53‧‧‧Chimney 62‧‧‧inclined surface 70‧‧‧Bottom evaporation tube 100‧‧‧Operation support device 110‧‧‧Data Acquisition Department 120‧‧‧Operation control device 130‧‧‧Operating data memory 131‧‧‧Cumulative value calculation unit 133‧‧‧Data Extraction and Conversion Department 140‧‧‧RTC (Real-Time Clock, real-time clock) 211‧‧‧Factory Specification Memory Department 212‧‧‧Process value candidate memory 220‧‧‧Operating condition calculation unit 221‧‧‧Analog Department 225‧‧‧Optimization Department 231‧‧‧Model Making Department 232‧‧‧ Evaluation Condition Review Department 233‧‧‧Evaluation Condition Memory 241‧‧‧Model Memory 250‧‧‧Operation instruction department 260‧‧‧I/O (input) 301‧‧‧CPU 302‧‧‧RAM 303‧‧‧ROM 304‧‧‧HDD 305‧‧‧Input I/F 306‧‧‧Output I/F 307‧‧‧Bus

Figure 1 is the overall configuration diagram of the operation support device using the predictive model

Figure 2 Schematic diagram of the boiler

Figure 3 is a diagram showing an example of factory specification data

Figure 4 is a diagram showing an example of candidate data for process values

Figure 5 is a diagram showing an example of a learning model created by the model creation section

Figure 6 is a diagram showing the hardware configuration of the operation support device

Fig. 7 is a flowchart showing the flow of the learning model creation process by the operation support device

Fig. 8 is a diagram showing the magnitude of the first allowable error and the second allowable error

9 is a flowchart showing the flow of the operation support method using the predictive model by the operation support device

Figure 10 is a diagram showing an example of scoring conversion criteria

Figure 11 is a diagram showing an example of a scoring conversion criterion

1: boiler

100: Operation support device

110: Data Acquisition Department

120: Operation control device

131: Cumulative value calculation section

133: Data Extraction and Conversion Department

140: RTC (Real-Time Clock, real-time clock)

211: Factory Specification Memory Department

212: Process value candidate memory

220: Operating condition calculation unit

221; Simulation Department

225: Optimum Department

231: Model Making Department

232: Evaluation Condition Review Department

233: Evaluation Condition Memory Department

241: Model Memory Department

250: Operation instruction department

260: I/O section (input section)

Claims (9)

A boiler operation support device, characterized by: having: a data acquisition unit that acquires the actual operating conditions suitable for the actual operation of the boiler, and the actual process values obtained by applying the actual operating conditions to the results of the aforementioned boiler operation ; Model memory unit, which memorizes the learning model obtained by mechanically learning the relationship between the aforementioned actual operating conditions and the aforementioned actual process values; operating data memory unit, which stores operating data including the aforementioned actual operating conditions and the aforementioned actual process values ; And the operating condition calculation unit, which calculates the optimal operating conditions for the predicted process value calculated by applying the foregoing operating data to the memorized learning model to meet the predetermined evaluation conditions, and the foregoing actual process value includes: the main control of the boiler Both the main control process value of the object and the peripheral process value related to peripheral information. The main control process value is any one of the quality index value of the final product generated by the boiler and the index related to environmental regulation value or In combination, the aforementioned peripheral process value is an index related to the temperature or pressure of the equipment in the aforementioned boiler, and the component and concentration of the gas, liquid, or solid discharged from the aforementioned equipment in the boiler that are not subject to environmental regulatory values Any one or any combination of indicators and indicators related to the opening of the operating end of the aforementioned boiler, In the aforementioned main control process value, the index related to the quality of the aforementioned final product is the boiler outlet steam temperature, and the index related to the aforementioned environmental regulatory value or external regulatory value is the NOx value. Among the aforementioned peripheral process values, it is related to the boiler internal The indicator of the temperature or pressure of the equipment is the surface temperature of the heat transfer tube or the pressure difference of the boiler, and the gas, liquid or solid discharged from the equipment in the boiler is not subject to environmental regulatory values or external regulatory values The index of composition and concentration is the concentration of oxygen in combustion air or exhaust gas, and the index of the opening of the aforementioned operating end is the opening of the spray valve. For example, the operation support device for boilers in the scope of the patent application, wherein the aforementioned learning model is a predictive model that obtains predicted process values if hypothetical operating conditions are input, and the aforementioned operating condition calculation unit includes: a simulation unit, which combines the aforementioned At least one or more hypothetical operating conditions are input to the prediction model to calculate the predicted process value; and an optimizing part, which applies a pre-set scoring conversion criterion to the predicted process value to calculate the score of the virtual operating condition, according to For this scoring, the aforementioned optimal operating condition is selected from among the virtual operating conditions satisfying the aforementioned predetermined evaluation condition. For example, the boiler operation support device of the first item of the scope of patent application, in which the aforementioned operating condition calculation unit is able to satisfy the first evaluation condition that the aforementioned main control process value should satisfy, than whether it can satisfy the second evaluation condition that the aforementioned peripheral process value should satisfy. The evaluation conditions are given priority to calculate the aforementioned optimal operating conditions. For example, the boiler operation support device of the third item of the scope of patent application, which additionally has: an input unit that accepts the input of instructions from an operator; and an evaluation condition review unit that is based on the operation received by the aforementioned input unit The instructions of the personnel or the operating data of the aforementioned boilers update the aforementioned second evaluation condition regarding the aforementioned peripheral process values. For example, the boiler operation support device of the first item of the scope of patent application is additionally equipped with: a cumulative value calculation unit, which calculates the cumulative value of at least one or more of the aforementioned actual process values obtained by the aforementioned data acquisition unit, and calculates the aforementioned operating conditions The system calculates the above-mentioned optimal operating condition that the cumulative value of the above-mentioned predicted process value satisfies the above-mentioned predetermined evaluation condition. A boiler operation support method, which is characterized by: including: obtaining actual operating conditions suitable for the actual operation of the boiler, and applying the actual operating conditions to the actual process value obtained from the result of the aforementioned boiler operation; Memorize the steps of the learning model obtained by mechanically learning the relationship between the foregoing actual operating conditions and the foregoing actual process values; memorize the steps of operating data including the foregoing actual operating conditions and the foregoing actual process values; and calculate the memorized learning model The step of applying the aforementioned operating data to calculate the predicted process value to meet the predetermined evaluation condition and the optimal operating condition. The aforementioned actual process value includes: the main control process value of the main control object of the boiler and the peripheral process value related to peripheral information On both sides, the aforementioned main control process value refers to any one or a combination of the quality index value of the final product produced by the aforementioned boiler and the index related to environmental regulation value, and the aforementioned peripheral process value refers to the equipment in the aforementioned boiler Among the indicators of temperature or pressure, indicators related to the components and concentrations of the gas, liquid, or solid discharged from the equipment in the boiler that are not subject to environmental regulation values, and indicators related to the opening of the operating end of the boiler In any one or any combination of the above-mentioned main control process values, the index related to the quality of the final product is the boiler outlet steam temperature, and the index related to the aforementioned environmental regulation value or external regulation value is the NOx value. Among the process values, the index related to the temperature or pressure of the equipment in the boiler is the heat transfer tube The surface temperature or boiler pressure difference of the boiler, and the index of the composition and concentration of the gas, liquid or solid discharged from the equipment in the boiler that is not subject to the environmental regulatory value or the environmental regulatory value is the combustion air or exhaust gas The index of the oxygen concentration in the above-mentioned operating end opening is the opening of the spray valve. A method for creating a learning model of a boiler, which is characterized by: obtaining the boiler including the actual operating conditions suitable for the actual operation of the boiler, and the actual process value obtained by applying the actual operating conditions to the result of operating the boiler The step of reading the boiler specification data that specifies the specifications of the aforementioned boiler; the step of reading the process value candidates used to make the learning model of the aforementioned boiler; based on the aforementioned boiler specification data, the aforementioned process value candidates The step of selecting the process value used to create the aforementioned learning model; the step of selecting the descriptive variable of the selected process value from the aforementioned operating data; and mechanically learning the relationship between the aforementioned descriptive variable and the aforementioned operating data. In the step of learning the model, the actual process value includes: the main control process value of the main control object of the boiler and the peripheral process value of the peripheral information. The main control process value is the final result generated by the boiler. Any one or a combination of the quality index value of the result and the index related to the environmental regulation value Together, the aforementioned peripheral process values are indicators related to the temperature or pressure of the equipment in the boiler, and indicators related to the components and concentrations of the gas, liquid, or solid discharged from the equipment in the boiler that are not subject to environmental regulatory values , And any one or any combination of indicators related to the opening degree of the operating end of the aforementioned boiler, in the aforementioned main control process value, the indicator related to the quality of the aforementioned final product is the boiler outlet steam temperature, which is related to the aforementioned environmental regulatory value Or the index of the external regulatory value is the NOx value. Among the aforementioned peripheral process values, the index related to the temperature or pressure of the equipment in the boiler is the surface temperature of the heat transfer tube or the boiler pressure difference. The index of the component and concentration of the discharged gas, liquid or solid that are not subject to environmental regulatory values or external regulatory values is the oxygen concentration in the combustion air or exhaust gas, and the index related to the opening of the aforementioned operating end is the spray valve Opening. For example, the method for creating the learning model of the boiler in the scope of patent application, in which, in the learning model created above, the first of the allowable range of the error between the predicted process value and the actual process value of the main control process value is set 1 The allowable error is the second allowable error that is less than the allowable range of the error between the predicted process value and the actual process value set for the aforementioned peripheral process value difference. For example, the method for creating a boiler learning model in the scope of the patent application, wherein the aforementioned peripheral process values are: when the aforementioned learning model is created, the necessary peripheral process values that must be selected objects, and any peripheral process values that become arbitrarily selected objects The aforementioned required peripheral process value is a peripheral process value related to the safety of the aforementioned boiler, and the aforementioned arbitrary peripheral process value is a peripheral process value that becomes an alarm target of the aforementioned boiler.

Images