JP2014092798A - Driving simulator - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a driving simulator capable of simply deriving a process value while reducing data volume.SOLUTION: A driving simulator 10 derives a process value in accordance with a set value input by an operator, and the behavior of a device is simulated. The driving simulation 10 includes: a model storage unit 11 in which a statistical model relating to the probability of occurrence of the process value corresponding to the set value is prestored; a process value calculating unit 13 in which a process value is derived on the basis of the set value and the statistical model; and a display unit 14 on which information on the process value derived by the process value calculating unit 13 is displayed.

Description

  The present invention relates to a driving simulator.

  Conventionally, for example, in a large-scale apparatus such as a circulating fluidized bed (CFB) boiler, an operation simulator is used for the purpose of improving an operator's operation skill. In such a driving simulator, it is important to reproduce the response (process value) of the apparatus to the set value input by the operator in the same manner as the actual machine. For example, Patent Document 1 discloses a mathematical model (physical model). An operation simulator is described in which a process value is obtained by calculation using the.

JP 2006-194550 A

  However, in a driving simulator using a mathematical model as described in Patent Document 1, it is necessary to create and verify a complicated mathematical model, which may require a lot of work. Here, an operation simulator that searches from past data and obtains a process value is also conceivable, but in this case, a large-capacity storage device for accumulating a large amount of past data may be required.

  Therefore, an object of the present invention is to provide an operation simulator capable of easily deriving a process value while reducing the capacity.

  In order to solve the above problems, a driving simulator according to the present invention is a driving simulator for deriving a process value according to a set value input by an operator and simulating the behavior of the device, and a process corresponding to the set value. A model storage unit in which a statistical model relating to the occurrence probability of the value is stored in advance, an operation unit for deriving a process value based on the set value and the statistical model, and a display unit for displaying information on the process value derived by the operation unit It is characterized by comprising.

  In this driving simulator, a process value can be derived from a set value input by an operator using a statistical model. In this way, since the process value derivation based on the occurrence probability is performed, the process value can be derived easily and appropriately. Since the statistical model is a simple model that represents the probability of occurrence of a process value corresponding to a set value, a large-capacity storage device or the like is not necessary, and the capacity can be reduced.

  Moreover, it is preferable that the statistical model is created based on actual measurement data of process values corresponding to the set values. In this case, a statistical model can be created from actually measured data.

  In addition, an alarm determination unit that determines whether or not an alarm is necessary, the alarm determination unit determines that an alarm is necessary when the process value derived by the calculation unit is within a predetermined alarm range, The display unit preferably displays information regarding the alarm when the alarm determination unit determines that the alarm is necessary. When the process value of the predetermined alarm range is derived, the information regarding the alarm is displayed, thereby making it possible to perform driving training in accordance with the actual operation.

  Further, as a configuration that favorably exhibits the above-described effects, specifically, the statistical model is preferably a Bayesian network model. In this case, the process value can be calculated using a Bayesian network model.

  Further, the Bayesian network model has a conditional probability table, and the conditional probability table has a condition node related to the set value and a result node related to the process value, and the condition node has a plurality of values depending on the degree. In addition, a plurality of values of the result node are provided according to the degree, at least a plurality of values of the condition node, a plurality of values of the result node, and a random variable are associated with each other, and the arithmetic unit inputs to the operator The value of the condition node corresponding to the determined setting value is determined, and the value having the largest random variable is derived as the process value among the plurality of values of the plurality of result nodes associated with the determined value of the condition node. It is preferable. In this case, it is possible to derive a process value by holding a conditional probability table of set values, process values, and random variables and searching from the conditional probability table.

  Moreover, it is preferable that the said apparatus is a circulating fluidized bed boiler. In general, in a circulating fluidized bed (CFB) boiler, process value calculation is complicated, and application of a mathematical model is particularly difficult. Therefore, in the present invention for simulating the behavior of a circulating fluidized bed boiler, the above-described effect of simply deriving a process value while reducing the capacity is particularly effective.

  According to the present invention, it is possible to provide an operation simulator capable of easily deriving a process value while reducing the capacity.

It is a figure which shows the functional block of the driving simulator which concerns on one Embodiment of this invention. It is a figure which illustrates a Bayesian network model. It is a figure which illustrates a conditional probability table | surface. It is a figure explaining the Bayesian network model used with the driving simulator shown in FIG. It is a table | surface which shows the conditional probability table | surface of the Bayesian network model shown in FIG. It is a flowchart which shows the process of the driving simulator shown in FIG. It is a flowchart which shows the preparation process of the conditional probability table | surface shown in FIG. It is a figure which shows an example of the measurement data at the time of CFB boiler starting. It is a table | surface which shows process value derivation precision. It is the graph which compared the actual value of the process value, and the derived value.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

  First, each function of the driving simulator will be described with reference to FIG. FIG. 1 is a diagram showing functional blocks of a driving simulator according to an embodiment of the present invention. As shown in FIG. 1, an operation simulator 10 according to the present embodiment is a computer device used as an operation training apparatus for a CFB boiler that is a boiler plant. The driving simulator 10 simulates the behavior of the device by deriving a process value in accordance with the set value input by the operator.

  The driving simulator 10 includes a model storage unit 11, a set value input unit 12, a process value calculation unit (calculation unit) 13, a display unit 14, a data output unit 15, and an alarm determination unit 16. Composed. The driving simulator 10 physically includes hardware such as a CPU (Central Processing Unit), a RAM (Random Access Memory) and a ROM (Read Only Memory), which are main storage devices, and an auxiliary storage device. As these operate, the functions of the above functional blocks are exhibited.

  The model storage unit 11 stores in advance a statistical model relating to the occurrence probability of the process value corresponding to the set value as a model expressing the behavior of the apparatus. In the present embodiment, a Bayesian network model is used as the statistical model, and the Bayesian network model has a conditional probability table. First, an overview of the Bayesian network and the conditional probability table will be described below.

  A Bayesian network is a data structure for storing a probabilistic causal relationship between a plurality of random variables. By using a Bayesian network, complex knowledge can be expressed efficiently, and various probabilistic inferences can be made based on the knowledge. A Bayesian network includes nodes representing random variables and links representing causal relationships between random variables. A statistical model based on the structure of a Bayesian network is defined and described as a Bayesian network model in this embodiment.

Such a Bayesian network holds what is called a conditional probability table (CPT) for each node. The conditional probability indicates a probability that another node B takes a certain value under a condition that the node A takes a certain value, and is expressed as P (B | A). Further, the probability that the node A and the node B occur simultaneously is called “simultaneous probability” and is expressed as P (A∩B). Since the conditional probability represents the probability that the node B will occur when the node A is considered as a whole, it can be expressed by the following equation (1) using the joint probability.

  That is, the conditional probability table is a table showing the conditional probabilities that the node B has a certain value when the set of the node A that is the parent node of the node B has a certain value combination. In the present embodiment, a node corresponding to node A in the above formula is described as a condition node, and a node corresponding to node B is described as a result node.

  FIG. 2 is a diagram illustrating a Bayesian network model, and FIG. 3 is a diagram illustrating a conditional probability table. As shown in FIG. 2, in this example, three types of nodes representing random variables are described (probability variables A, B, X), and the random variable A is “whether the amount of coal supply is greater than the average. ", The random variable B represents" whether the air supply amount is higher than the average ", and the random variable X represents" whether the furnace temperature is higher than the average ".

  The causal relationship between the three types of nodes is given as a conditional probability table as shown in FIG. The conditional probability table is data that stores knowledge of strength in the causal relationship between nodes. For example, when P (X = low | A = many, B = many) = 0.6258, when the coal supply amount and the air supply amount that are the condition nodes I are larger than the average, the furnace temperature that is the result node O is This represents the knowledge that the probability of lowering is 0.6258. P (A = less) represents knowledge that the probability that the amount of coal supply is less than the average is 0.7211.

  In the example described above, the coal supply amount and the air supply amount correspond to the condition node I, and are the operation ends as setting items by the operator, and these values are set values. On the other hand, the furnace temperature corresponds to the result node O, and the value becomes the process value. Here, when the operator sets the coal supply amount = large and the air supply amount = small, the probability that the furnace temperature as the process value is lower than the average is 0.02, and the probability that the furnace temperature is higher is 0.97. Therefore, in this case, it is inferred that the probability that the furnace temperature is higher than the average is the highest.

  If a Bayesian network model (a network structure of a Bayesian network) is given and there is a large amount of observation data for a set of values of each random variable, the values of the elements of the conditional probability table can be determined based on the observation data. This is called conditional probability table learning. For example, if the number of observation data of 1000 pieces is less than the average of 721, P (coal supply amount = small) is 721/1000. In addition, if 500 furnace furnace temperatures are higher than the average among them, P (furnace temperature = high | coating amount = small) is 500/721.

  In the above example, there are two condition nodes I and one result node O related to the process value. However, the present invention is not limited to this. The number of condition nodes I and result nodes O may be one or more. Good. In the above example, the value of the condition node I is binary (either less or more than the average) and the value of the result node O is binary (either low or high). A plurality of three or more values may be used.

  Returning to FIG. 1, the model storage unit 11 of the present embodiment stores a conditional probability table H (see FIG. 5) created based on measured data of process values corresponding to set values, as will be described later. ing. The set value input unit 12 receives a value (set value) of a setting item (operation end) set by the operator. For example, when the driving simulator 10 is used for training, the operator refers to a trainee (operator).

  The process value calculation unit 13 derives a process value based on the set value and the statistical model. Here, the process value calculation unit 13 reads a statistical model from the model storage unit 11, inputs a set value to the statistical model, and simulates a process value that is a simulated response value corresponding to the set value.

  As will be described later, the process value calculation unit 13 of the present embodiment derives a process value using the conditional probability table H (see FIG. 5). Specifically, the process value calculation unit 13 receives the set value input unit 12. A value of the condition node I corresponding to the set value is determined, and a value having the largest random variable among a plurality of values of the result node O associated with the value of the condition node I is derived as a process value.

  The display unit 14 is for the operator to visually confirm the driving situation, and displays information on the process value derived by the process value calculation unit 13 on a monitor screen or the like. In addition, this display part 14 may further display information required for confirmation of other driving | running conditions, such as a setting value. Moreover, the display part 14 displays further the information regarding this alarm, when it determines with the alarm determination part 16 requiring an alarm.

  The data output unit 15 outputs the set value and the process value to an external or built-in storage device (not shown) and also outputs to the alarm determination unit 16. The set value and process value data output by the data output unit 15 is stored in the storage device as an operation status record of the operation simulator and used for alarm determination by the alarm determination unit 16.

  The alarm determination unit 16 determines whether or not an alarm is necessary. The alarm determination unit 16 determines that an alarm is necessary when the process value output from the data output unit 15 is within a predetermined alarm range. The predetermined alarm range here may be set in accordance with, for example, an actual machine (CFB boiler) to be simulated, or may be set as appropriate by an operator.

  Next, processing of the driving simulator 10 will be described with reference to FIG. FIG. 6 is a flowchart showing processing of the driving simulator according to the embodiment of the present invention. In the following, as an example, a simulation calculation of a process value at the time of starting a CFB boiler is illustrated.

  4 is a diagram illustrating a Bayesian network model used in the driving simulator shown in FIG. 1, and FIG. 5 is a table showing a conditional probability table of the Bayesian network model shown in FIG. In FIG. 5, a part of the conditional probability table H used for the driving simulator 10 is omitted.

  As shown in FIG. 4, the Bayesian network model used in the operation simulator 10 here has a heavy oil flow rate, a total coal supply amount, and a boiler MCR as a condition node I related to the set value. Further, as a result node O related to the process value, there are a boiler feed water flow rate, a furnace bottom representative temperature, a main steam temperature, a main steam pressure, and generated power. In the link L representing the causal relationship between the condition node I and the result node O, the arrows are directed from all the condition nodes I to each result node O.

  As shown in FIG. 5, the conditional probability table H here shows the boiler feedwater flow rate that is the result node O under the condition that the heavy oil flow rate that is the condition node I, the total coal supply amount, and the boiler MCR are set to certain values. The values taken by the furnace bottom representative temperature, the main steam temperature, the main steam pressure, and the generated power are represented by probabilities (random variables). The values that each condition node I and each result node O take are not continuous values but discretized values, and a plurality (here, 10) are provided.

  In such a driving simulator 10, first, the set value input by the operator is received by the set value input unit 12 (S1). Subsequently, the value (process value) of each result node O is derived by the process value calculation unit 13 based on the set value and the conditional probability table H of the Bayesian network model (S2).

  Subsequently, the process value derived by the process value calculation unit 13 is displayed on the monitor screen by the display unit 14 (S3). Then, whether or not the process value derived by the process value calculation unit 13 is within a predetermined alarm range is determined by the alarm determination unit 16 (S4), and if it is determined that the process value is within the predetermined alarm range, The information about the alarm is displayed on the monitor screen by the display unit 14 (S5). On the other hand, when the alarm determination unit 16 determines in S4 that the process value is not within the predetermined alarm range, or when the process of S5 is completed, the process proceeds to S6 described later.

  In S6, the target of various processes using the driving simulator 10 has been reached (for example, if the driving simulator 10 is used for driving training of the operator, the driving training target has been reached), or the operator It is determined whether an operation for ending the process of the driving simulator 10 (for example, selection of an end menu) has been performed (S6). If YES in S6, the process by the driving simulator 10 is terminated. If NO in S6, the process proceeds to S1 (input of set value) again.

  Next, the creation process (learning) of the conditional probability table H will be described.

  FIG. 7 is a flowchart showing processing for creating a conditional probability table, and FIG. 8 is a diagram showing an example of actually measured data when the CFB boiler is activated. In actual measurement data D shown in FIG. 8, the values of heavy oil flow rate, total coal supply amount, boiler MCR, boiler feed water flow rate, furnace bottom representative temperature, main steam temperature, main steam pressure, and generated power are measured for about 15 hours. ing.

  As shown in FIGS. 7 and 8, in this embodiment, the conditional probability table H is created based on the actual measurement data D that is the trend data of the process value corresponding to the set value. Here, for example, as shown below, the conditional probability table H is learned from the actual measurement data D when the CFB boiler is activated.

  That is, for example, after setting / specifying each condition node I and each result node O by the operator, the measurement value D is received by the set value input unit 12 and the measurement data D is accumulated (S11, S12). Subsequently, each value of the actual measurement data D is discretized (S13). Specifically, for each value of the measured data D, the range of ± 3σ is discretized into 10 groups.

  After each value of the measured data D is discretized, a conditional probability table H (see FIG. 5) is created based on the discretized data (S14). Specifically, the values of each condition node I and each result node O at a certain time are acquired, and the acquisition is performed for a plurality of times. Then, the conditional probability table H is created by deriving the probabilities of the values of the result nodes O when the values of the conditional nodes I have the same condition. Finally, the conditional probability table H is output, and the creation process of the conditional probability table H is terminated (S15).

  Here, an example of the process value derivation (the above S2) using the conditional probability table H will be described in detail with reference to FIG.

  First, the value of each condition node I in the conditional probability table H is determined from the set value received by the set value input unit 12. Here, as shown by the thick frame in the figure, heavy oil flow rate: 7.5 [kl / h], coal feed amount: 3.0 [t / h], boiler MCR: 45.0 [%] Assume that the value of the condition node I is determined.

  Subsequently, the value having the highest probability among the plurality of values of each result node O associated with the value of each condition node I is derived as the process value of each result node O. Here, as shown in a thick frame in the figure, the boiler feed water flow rate is 110 [t / h] with a probability of 56%, the furnace bottom representative temperature is 585 [° C.] with a probability of 51%, and the main steam temperature is The process value is derived for 510 [° C.] with a probability of 62%, the main steam pressure is 9.75 [MPaG] with a probability of 58%, and the generated power is 0.5 [MW] with a probability of 72%. Will be.

  Next, functions and effects of the driving simulator 10 according to the present embodiment will be described.

  In the driving simulator 10 of the present embodiment, as described above, the process value can be derived from the set value input by the operator using the statistical model (Bayesian network model). Since the process value derivation based on the occurrence probability is performed as described above, the process value can be derived easily and appropriately. Since the statistical model is a simple model that represents the probability of occurrence of a process value corresponding to a set value, a large-capacity storage device or the like is not necessary, and the capacity can be reduced. In addition, in this embodiment as a driving training apparatus, it is not necessary to use the same specification product as a real machine for a simulation, a large-scale training facility becomes unnecessary, and it becomes possible to provide a cheap and simple driving training apparatus.

  FIG. 9 is a table showing process value derivation accuracy. The table in the figure shows how much the derived value of the process value using the conditional probability table H matches the actual measurement value for each result node O shown in FIG. As shown in FIG. 9, the derivation accuracy shows an accuracy of 70% or more in any result node O, and it is shown that a highly accurate inference can be made.

  FIG. 10 is a graph comparing the measured value and the derived value of the process value of the main steam temperature. As shown in FIG. 10, it can be confirmed from the graph comparing the actual measurement value and the derived value of the process value that the derived value is approximately approximate to the actual measurement value. I understand that.

  Further, in the present embodiment, as described above, the conditional probability table H of the Bayesian network model is created based on the actual measurement data D, so that it is possible to derive practical process values with higher accuracy in accordance with the actual situation. It becomes. Further, as shown in FIG. 5, in the conditional probability table H, the value of the result node O corresponding to the value of each condition node I is represented by the probability, so the basis for deriving each process value becomes clear. Therefore, by using the Bayesian network model having the conditional probability table H, a practical simulation that is easy for the operator to use can be performed.

  In addition, as described above, the present embodiment further includes the alarm determination unit 16 that determines whether or not an alarm is necessary. When the process value is within a predetermined alarm range, the alarm determination unit 16 When it is determined that an alarm is necessary, and the display unit 14 determines that an alarm is necessary by the alarm determination unit, the display unit 14 displays information related to the alarm. As a result, when the driving simulator 10 is used for driving training, driving training can be performed in accordance with actual driving, which can contribute to improvement of the operator's operating skill.

  Further, as described above, the driving simulator 10 of the present embodiment can be applied to a CFB boiler as a simulation target. In this case, since the process value calculation is generally complicated in a CFB boiler, the above-described effect of easily deriving the process value while reducing the capacity becomes remarkable.

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments. The present invention is modified without departing from the scope described in the claims or applied to others. It may be.

  For example, in the above embodiment, a Bayesian network model is used as a statistical model, and the process value calculation unit 13 derives a process value from the conditional probability table H. However, the present invention is not limited to this, and other statistical models (for example, The process value may be derived using a neural network model, a genetic algorithm model, or the like.

  Moreover, although the said embodiment simulated the behavior of the CFB boiler, it is not limited to this, The present invention can also simulate the behavior of other large-scale devices such as power generation facilities and water treatment facilities. it can. Even in this case, the above-described effects are exhibited.

  Moreover, in the said embodiment, although the conditional probability table | surface H was created based on the measurement data D, it is not limited to this, You may create by calculation and you may create by another method. Moreover, in the said embodiment, although the conditional probability table | surface H was created in the driving simulator 10, you may create with another apparatus. Further, when the conditional probability table H is created by the driving simulator 10, the driving simulator 10 may have a learning unit that learns and creates the conditional probability table H.

  Incidentally, the conditional probability table H of FIG. 5 is a conditional probability table in which the condition node I and the result node O are reversed (that is, boiler feed water flow rate, furnace bottom representative temperature, main steam temperature, main steam pressure, and generated power). A condition probability table in which each value of the heavy oil flow rate, the total coal supply amount, and the boiler MCR is expressed as a probability under the condition that the values of and are set values. In this case, for example, by using an inverse probability, an operation substantially equivalent to the process value derivation using the conditional probability table H can be performed.

DESCRIPTION OF SYMBOLS 10 ... Driving simulator, 11 ... Model memory | storage part, 12 ... Setting value input part, 13 ... Process value calculating part (calculating part), 14 ... Display part, 15 ... Data output part, 16 ... Alarm determination part.

Claims (6)

A driving simulator that derives a process value according to a setting value input by an operator and simulates the behavior of the device,
A model storage unit in which a statistical model related to the occurrence probability of the process value corresponding to the set value is stored in advance;
A calculation unit for deriving the process value based on the set value and the statistical model;
And a display unit that displays information related to the process value derived by the arithmetic unit.   The driving simulator according to claim 1, wherein the statistical model is created based on actual measurement data of the process value corresponding to the set value. It further includes an alarm determination unit that determines whether an alarm is necessary,
The alarm determination unit determines that an alarm is necessary when the process value derived by the calculation unit is within a predetermined alarm range;
The driving simulator according to claim 1, wherein the display unit displays information related to the warning when the warning determination unit determines that the warning is necessary.   The driving simulator according to claim 1, wherein the statistical model is a Bayesian network model. The Bayesian network model has a conditional probability table;
The conditional probability table is
A condition node for the set value and a result node for the process value;
A plurality of condition node values are provided according to the degree, and a plurality of result node values are provided according to the degree.
At least a plurality of values of the condition node, a plurality of values of the result node, and a random variable are associated with each other;
The computing unit is
Determining a value of the condition node corresponding to the set value input by the operator;
The driving simulator according to claim 4, wherein a value having the largest random variable among a plurality of values of the result node associated with the determined value of the condition node is derived as the process value. The operation simulator according to claim 1, wherein the device is a circulating fluidized bed boiler.

Images