TWI542828B - Running simulator - Google Patents

Description

Running simulator

The invention relates to a running simulator.

Conventionally, in a large-scale apparatus such as a circulating fluidized bed (CFB) boiler, an operation simulator is used for the purpose of improving the operator's operation skill and the like. In the above operation simulator, it is important to reproduce the response (program value) of the device with respect to the setting value input by the operator in the same manner as the real machine. For example, Patent Document 1 describes the use of a mathematical model (physical model) by calculation. Find the running simulator of the program value.

(previous technical literature) (Patent Literature)

Patent Document 1: Japanese Laid-Open Patent Publication No. 2006-194550

However, in the operation simulator using the mathematical model described in Patent Document 1, it is necessary to create and verify a complicated mathematical model, which requires a lot of time and effort. Here, it is also considered to retrieve past data to find a program. The value of the operation simulator, but at this time, a large-capacity memory device for storing 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 outputting a program value while reducing the capacity.

In order to solve the above problems, the operation simulator of the present invention derives a program value corresponding to a set value input by an operator, and simulates the behavior of the device. The method includes a model memory unit that is stored in advance and corresponds to a set value. A statistical model relating to the probability of occurrence of the program value; the calculation unit derives the program value based on the set value and the statistical model; and the display unit displays the information related to the program value derived by the calculation unit.

In the operation simulator, the program value can be derived from the set value input by the operator using a statistical model. In this way, since the program value is derived based on the probability of occurrence, the program value can be easily and appropriately derived. Further, since the statistical model is a simple model indicating the probability of occurrence of the program value corresponding to the set value, it is not necessary to have a large-capacity memory device or the like, and the capacity can be reduced.

Further, the statistical model is preferably produced based on the measured data of the program values corresponding to the set values. At this time, a statistical model can be produced from the measured data.

Further, the operation simulator further includes an alarm determination unit that determines whether or not the alarm is generated. When the program value derived by the calculation unit enters the predetermined alarm range, the alarm determination unit determines that an alarm is required, and when the alarm determination unit determines that an alarm is required, It is preferable that the display unit displays information related to the alarm. When the program value of the specified alarm range is derived, the information related to the alarm is displayed, thereby Ability to perform practical training.

Further, as a configuration suitable for obtaining the above-described effects, specifically, the statistical model is a Bayesian network model. At this time, the Bayesian network model can be used to calculate the program value.

Moreover, the Bayesian network model has a conditional probability table, the conditional probability table has a conditional node related to the set value and a result node related to the program value, and the value of the conditional node is set according to the degree thereof, and the value of the result node is A plurality of values are set according to the degree, and at least a plurality of values of the condition node, a plurality of values of the result node, and a probability variable are associated with each other, and the operation unit determines the condition node value corresponding to the set value output by the operator, and The value of the multiplicity of the plurality of result nodes associated with the value of the conditional node that has been determined to be the largest is the preferred value of the probability variable as the program value. At this time, the conditional probability table of the set value, the program value, and the probability variable is held, and the conditional probability table is searched to be able to derive the program value.

Further, the above device is preferably a circulating fluidized bed boiler. In general, in circulating fluidized bed (CFB) boilers, program value calculations are more complex and the application of mathematical models is particularly difficult. Accordingly, in the present invention in which the behavior of the circulating fluidized bed boiler is simulated, it is particularly effective to reduce the capacity and easily derive the program value.

According to the present invention, there is provided an operation simulator capable of easily exporting program values while reducing capacity.

10‧‧‧Running simulator

11‧‧‧Model Memory

12‧‧‧Set value input section

13‧‧‧Program value calculation unit (calculation unit)

14‧‧‧Display Department

15‧‧‧ Data Export Department

16‧‧‧Alarm Judgment Department

Fig. 1 is a block diagram showing an operation simulator according to an embodiment of the present invention.

Figure 2 is a diagram illustrating a Bayesian network model.

Figure 3 is a diagram illustrating a conditional probability table.

Fig. 4 is a view for explaining a Bayesian network model used in the operation simulator shown in Fig. 1.

Fig. 5 is a table showing the conditional probability table of the Bayesian network model shown in Fig. 4.

Fig. 6 is a flow chart showing the processing of the operation simulator shown in Fig. 1.

Fig. 7 is a flow chart showing the process of creating the conditional probability table shown in Fig. 5.

Fig. 8 is a view showing an example of actual measurement data at the time of startup of the CFB boiler.

Figure 9 is a table showing the accuracy of the program value derivation.

Figure 10 is a graph comparing the measured values of the program values with the derived values.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS 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 corresponding elements are designated by the same reference numerals, and the description thereof will not be repeated.

First, referring to Fig. 1, each function of the operation simulator will be described. Fig. 1 is a view showing functional blocks of a running simulator according to an embodiment of the present invention. As shown in Fig. 1, the operation simulator 10 of the present embodiment is a computer device used as a running training device for a boiler equipment, that is, a CFB boiler. The operation simulator 10 derives the program value based on the set value input by the operator, thereby simulating the behavior of the device.

The operation simulator 10 includes a model storage unit 11, a set value input unit 12, a program value calculation unit (calculation unit) 13, a display unit 14, a data output unit 15, and an alarm determination unit 16. In addition, the operation simulator 10 physically includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and an auxiliary memory device, and the like. Play the role of each of the above functional blocks.

The model memory unit 11 stores, as a model of the behavior of the presentation device, a statistical model relating to the probability of occurrence of the program value corresponding to the set value. As a statistical model, in the present embodiment, a Bayesian network model is used, and the Bayesian network model has a conditional probability table. Therefore, the basic situation of the Bayesian network and the conditional probability table is first described below.

The Bayesian network is a data structure used to memorize the probabilistic causal relationship between a plurality of probability variables. By using the Bayesian network model, complex knowledge can be effectively expressed, and various probabilistic inferences can be made based on the knowledge. The Bayesian network model includes a node representing a probability variable and a link representing a causal relationship between the probability variables. In this embodiment, a statistical model composed of a structure of a Bayesian network is defined as a Bayesian network model. Bright.

The above Bayesian network is maintained as a Conditional Probability Table (CPT) for each node. The conditional probability refers to the probability that another node B takes a certain value under the condition that node A takes a certain value, and is shown as P(B|A). Moreover, the probability that node A and node B appear simultaneously is referred to as "simultaneous probability" and is shown as P(A∩B). The conditional probability indicates the probability of occurrence of the node B when the node A is regarded as a whole, and therefore the following equation (1) is expressed using the simultaneous probability.

That is, when the conditional probability table is a combination of a certain value of the node A of the parent node of the node B, the conditional probability that the node B has a certain value is used. Further, in the present embodiment, a node corresponding to the node A in the above equation is used as a condition node, and a node corresponding to the node B is used 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, the nodes representing the probability variables are described in three types (probability variables A, B, and X), the probability variable A indicates whether "the amount of coal added is more than the average", and the probability variable B indicates "into" Whether the gas volume is more than average, and the probability variable X means "whether the furnace temperature is higher than the average".

As shown in Fig. 3, the causal relationship of the above three kinds of nodes is obtained as a conditional probability table. The conditional probability table is a data that memorizes strong knowledge in the causal relationship between nodes. For example, P(X=lower|A= More, B = more) = 0.6258 indicates the knowledge that, as the conditional node I, the coal addition amount and the intake air amount are more than average, the probability of the furnace temperature decrease as a result of the node O is 0.6258. Also, P (A = less) indicates the following knowledge, that is, the probability that the amount of coal added is less than the average is 0.7211.

In the above example, the amount of coal added and the amount of intake air correspond to the condition node I, and the operator is the operator of the setting item, and the value becomes the set value. On the other hand, if the furnace temperature corresponds to the result node O, the value becomes a program value. Here, when the operator sets the amount of coal to be added to be large and the amount of intake air to be small, the probability that the furnace temperature is lower than the average as the program value is 0.02, and the probability of higher than the average is 0.97. Therefore, it is inferred that the probability that the furnace temperature is higher than the average at this time is the highest.

The Bayesian network model (the network structure of the Bayesian network) is provided, and when a large number of observation data is available for each group of probability variable values, the value of the conditional condition table can be determined based on this. This is called learning with a conditional probability table. For example, if 1000 of the 1000 observations are less than the average, the number of coal added is 721, and P (the amount of coal added = less) becomes 721/1000. Further, if the furnace temperature is further higher than the average of 500, P (furnace temperature = higher | coal addition amount = less) becomes 500 / 721.

Further, in the above example, the condition node I is set to two, and the result node O related to the program value is one. However, the present invention is not limited thereto, and the condition node I and the result node O may be one or two. More than one. Further, in the above example, the value of the conditional node I is set to 2 values (less or more than the average), and the value of the result node O is set to 2 (lower or higher), but it is also possible to set It is a plural of 1 value or more.

In the model storage unit 11 of the present embodiment, the conditional probability table H (see FIG. 5) created by the actual measurement data according to the program value corresponding to the set value is stored as will be described later. The set value input unit 12 receives the value (set value) of the setting item (operating end) set by the operator. For example, when the operation simulator 10 is used for training, the operator refers to a trainer (operator).

The program value calculation unit 13 derives a program value based on the set value and the statistical model. The program value calculation unit 13 reads the statistical model from the model storage unit 11, inputs the set value to the statistical model, and performs a simulation operation on the program value corresponding to the set value.

The program value calculation unit 13 of the present embodiment derives the program value using the conditional probability table H (see FIG. 5), which will be described later, and specifically, the condition node I corresponding to the set value received by the set value input unit 12. The value is determined, and the value of the probability variable having the largest value among the plurality of values of the result node O associated with the value of the conditional node I is derived as the program value.

The display unit 14 is for the operator to visually check the operation status, and displays information related to the program value derived by the program value calculation unit 13 on the display screen. Further, the display unit 14 may further display information necessary for confirming other operating conditions such as set values. When the alarm determination unit 16 determines that an alarm is required, the display unit 14 further displays information related to the alarm.

The data output unit 15 outputs a set value and a program value to an external or built-in memory device (not shown), and outputs it to the alarm determination unit 16. The data of the set value and the program value outputted by the data output unit 15 is used as the operation mode. The operational status record of the simulator is stored in the memory device and used for the alarm determination by the alarm determination unit 16.

The alarm determination unit 16 determines whether or not the alarm is issued. When the program value output from the data output unit 15 enters the predetermined alarm range, the alarm determination unit 16 determines that an alarm is required. The predetermined alarm range may be set, for example, according to the actual machine (CFB boiler) of the simulation object, or may be appropriately set by the operator.

Next, the processing of the operation simulator 10 will be described with reference to Fig. 6 . Fig. 6 is a flow chart showing the processing of the operation simulator according to the embodiment of the present invention. In addition, as an example, the simulation calculation of the program value at the time of startup of a CFB boiler is illustrated.

Fig. 4 is a view for explaining a Bayesian network model used in the operation simulator shown in Fig. 1. Fig. 5 is a table showing the conditional probability table of the Bayesian network model shown in Fig. 4. In the fifth drawing, a part of the conditional probability table H used for the operation simulator 10 is omitted.

As shown in Fig. 4, the Bayesian network model used in the operation simulator 10 has a heavy oil flow rate, a total amount of coal added, and a boiler MCR as condition nodes I related to the set value. Further, as a result of the correlation with the program value, the node O has a boiler inflow flow rate, a furnace bottom representative temperature, a main steam temperature, a main steam pressure, and a generated electric power. The arrow of the link L indicating the causal relationship between the conditional node I and the resulting node O is directed from all conditional nodes I to the respective result nodes O.

As shown in Fig. 5, the conditional probability table H therein sets the conditional node I, that is, the heavy oil flow rate, the total amount of coal added, and the boiler MCR to a certain value. Under the condition, the probability node (probability variable) is used to represent the result node O, that is, the boiler inlet flow rate, the furnace bottom representative temperature, the main steam temperature, the main steam pressure, and the generated electricity. The values obtained by each condition node I and each result node O are not continuous values, but are discretized values, and are provided with plural numbers (here, 10).

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

Next, the program value derived by the program value calculation unit 13 is displayed on the display screen by the display unit 14 (S3). When the program value derived by the program value calculation unit 13 determines whether or not the alarm determination unit 16 is within the predetermined alarm range (S4) and determines that it is within the predetermined alarm range, the display unit 14 displays the information related to the alarm on the display screen. (S5). On the other hand, when the alarm determination unit 16 determines in the above S4 that the program value is not within the predetermined alarm range, or when the processing of the above S5 is completed, the process proceeds to S6, which will be described later.

In S6, it can be determined whether or not the target of the various processes using the operation simulator 10 is reached (for example, whether the operation simulator 10 is used for the operation training of the operator, whether the target of the running training is reached), or whether the operator has performed The operation of ending the processing of the operation simulator 10 (for example, selecting the end list) (S6). When the determination in S6 is YES, the processing by the operation simulator 10 is completed, and when it is NO in the above S6, the processing of the above S1 is again performed (input setting value).

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

Fig. 7 is a flow chart showing the process of creating the conditional probability table, and Fig. 8 is a view showing an example of the measured data at the time of startup of the CFB boiler. In the measured data D shown in Fig. 8, it takes about 15 hours to measure the heavy oil flow, the total amount of coal added, the boiler MCR, the boiler inlet flow rate, the furnace bottom representative temperature, the main steam temperature, the main steam pressure, and the power generation. The value of electricity.

As shown in Figs. 7 and 8, in the present embodiment, the conditional probability table H is created based on the actual data D, which is the dynamic data of the program value corresponding to the set value. Here, for example, as shown below, the conditional probability table H is learned from the measured data D when the CFB boiler is started.

In other words, for example, after the operator sets/specifies each condition node I and each result node O, the measured value input unit 12 receives the measured data D and accumulates the measured data D (S11, S12). Next, each value of the measured data D is discretized (S13). Specifically, regarding each value of each item of the measured data D, the range of ±3 σ is divided into 10 groups and discretized.

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 with respect to a plurality of times. Further, when the value of each condition node I is the same condition, the probability that the value of each result node O can be obtained is derived, and the conditional probability table H is created. Finally, the output comes with The conditional probability table H, the creation process of the conditional probability table H is completed (S15).

Here, an example of the derived program value (S2) using the conditional probability table H will be described in detail using 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 indicated by the thick frame in the figure, the value of each node I is determined as the heavy oil flow rate: 7.5 [k1/h], the coal addition amount: 3.0 [t/h], and the boiler MCR: 45.0 [%].

Next, 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 program value of each result node O. Among them, as shown in the thick box in the figure, as follows, they are respectively derived as program values, that is, the probability that the boiler inlet flow rate is 56% of 110 [t/h], and the probability that the bottom of the furnace represents the temperature is 51%. [°C], the probability of the main steam temperature is 510 [° C] of 62%, the probability of the main steam pressure is 9.75 [MPaG] of 58%, and the probability of generating electric power is 0.5 [MW] of 72%.

Next, the action and effect of the operation simulator 10 of the present embodiment will be described.

In the operation simulator 10 of the present embodiment, as described above, the program value can be derived from the set value input by the operator using a statistical model (Bayesian network model). Thereby, in order to derive the program value according to the probability of occurrence, the program value can be easily and appropriately derived. Further, since the statistical model is a simple model indicating the probability of occurrence of the program value corresponding to the set value, the capacity can be reduced without requiring a large-capacity memory device or the like. In addition, in the present embodiment as the running training device, it is not necessary to use the same simulation as the real machine. Specifications, without the need for large-scale training facilities, can also provide an inexpensive and simple running training device.

Figure 9 is a table showing the accuracy of the program value derivation. The table in the figure shows how much the degree of agreement between the derived value of the program value of the conditional probability table H and the real side value is used 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 of the result nodes O, thereby showing a case where a highly accurate inference can be performed.

Figure 10 is a graph comparing the measured values of the programmed values of the main steam temperature with the derived values. As shown in Fig. 10, it can be confirmed from the graph comparing the measured value of the program value with the derived value that the derived value is substantially close to the actual measured value, and it can be seen that the inference with high accuracy can be performed.

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 a practical value that is more practical and more accurate. 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 a probability, and therefore the basis for deriving each program value is clear. Thereby, it is possible to make the simulation easy and practical for the operator by using the Bayesian network model with the conditional probability table H.

Further, in the above-described embodiment, the alarm determination unit 16 that determines whether or not to alert is provided, and when the program value enters the predetermined alarm range, the alarm determination unit 16 determines that an alarm is required, and when the alarm determination unit determines that an alarm is required, The display unit 14 displays information related to the alarm. Thereby, when the operation simulator 10 is used for running training, it is possible to conform to the actual situation. Running training and can help improve the operator's operational skills.

Further, as described above, the operation simulator 10 of the present embodiment can be applied to a CFB boiler as a simulation target. At this time, since the calculation of the program value is usually complicated for the CFB boiler, the above-described effects of simply releasing the program value while reducing the capacity are remarkable.

The preferred embodiments of the present invention have been described above, but the present invention is not limited to the above-described embodiments, and may be modified within the scope of the technical idea described in the claims, or may be applied to others.

For example, in the above-described embodiment, the Bayesian network model is used as the statistical model, and the program value calculation unit 13 derives the program value by the conditional probability table H. However, the present invention is not limited thereto, and another statistical model may be used (for example, Neural network models, genetic algorithm models, etc.) to derive program values.

Further, in the above embodiment, the behavior of the CFB boiler has been simulated. However, the present invention is not limited thereto, and the present invention can simulate the behavior of other large-scale devices such as a power generation facility or a water treatment facility. At this time, the above effects can also be obtained.

Further, in the above-described embodiment, the conditional probability table H is created based on the actual measurement data D. However, the present invention is not limited thereto, and may be produced by calculation or by other methods. Further, in the above embodiment, the conditional probability table H is created in the operation simulator 10, but it may be produced by another device. Further, when the conditional probability table H is created in the operation simulator 10, the operation simulator 10 may have a learning unit that learns to create the conditional probability table H.

In addition, the conditional probability table H of Fig. 5 can also be used as a conditional section. Probability of the conditional probability table (ie, the boiler inlet flow rate, the furnace bottom representative temperature, the main steam temperature, the main steam pressure, and the generated power value) is set to the set value under the condition that the point I and the result node O are twisted. Indicates the conditional probability table for the heavy oil flow rate, the total amount of coal added, and the values of the boiler MCR. At this time, for example, the calculation is performed using the inverse probability, and it is possible to perform an operation substantially equivalent to the derivation of the program value using the above-described conditional probability table H.

10‧‧‧Running simulator

11‧‧‧Model Memory

12‧‧‧Set value input section

13‧‧‧Program Value Computing Department

14‧‧‧Display Department

15‧‧‧ Data Export Department

16‧‧‧Alarm Judgment Department

Claims (6)

An operation simulator is configured to derive a program value corresponding to a set value input by an operator, and simulate a behavior of the device; and the method includes: a model memory unit that stores in advance a program value corresponding to the set value. A statistical model relating to occurrence probability; the calculation unit derives the program value based on the set value and the statistical model; and the display unit displays information related to the program value derived by the calculation unit. The operation simulator of claim 1, wherein the statistical model is created based on actual measurement data corresponding to the program value of the set value. The operation simulator according to claim 1 or 2, further comprising: an alarm determination unit that determines whether or not the alarm is generated; wherein the alarm determination unit determines that the program value obtained by the calculation unit enters a predetermined alarm range The display unit displays information related to the alarm when the alarm determination unit determines that an alarm is required. The operation simulator of claim 1 or 2, wherein the foregoing statistical model is a Bayesian network model. The operation simulator of claim 4, wherein the Bayesian network model has a conditional probability table; the conditional probability table: having a condition node associated with the set value, and the program value a result node of the correlation, wherein the value of the condition node is a plurality of values according to the degree, and the value of the result node is a plurality of values according to the degree, at least a plurality of values of the condition node, a plurality of values of the result node, and The probability variables are associated with each other; the calculation unit determines a value of the condition node corresponding to the set value input by the operator, and associates the result node associated with the determined value of the condition node The value of the plurality of values in which the aforementioned probability variable is the largest is derived as the aforementioned program value. The operation simulator of claim 1 or 2, wherein the device is a circulating fluidized bed boiler.