JP2013182471A - Load evaluation device for plant operation - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a load evaluation device and a load evaluation program for a plant operation for quantitatively evaluating a load of the plant operation.SOLUTION: A load evaluation device for a plant operation includes: acquisition means for acquiring history data indicating a history about an alarm in a plant control monitoring system from the plant control monitoring system for giving a control command to a plant while inputting measurement data from the plant; calculation means for calculating alarm generation interval distribution and alarm handling time distribution in the plant on the basis of an occurrence time of the alarm and a recovery time which are included in the history data; evaluation means for calculating a load of the plant operation from the alarm generation interval distribution and a processing time distribution for the alarm by obtaining the processing time distribution from the alarm generation interval distribution and the alarm handling time distribution on the basis of a queuing theory.

Description

  The present invention relates to a plant operation load evaluation apparatus and a load evaluation program for evaluating a load of a plant operation by an operator.

  Generally, various plants are automatically controlled by a plant control monitoring system. Such a plant control monitoring system includes a distributed control system (DCS), a programmable logic controller (PLC), and the like.

  Various events can occur depending on the plant operating conditions and disturbance factors, so in many plants it is difficult to fully automate (no operator at all), and the operator provides various information provided by the plant control monitoring system. While monitoring, necessary operations such as setting value change and process change are performed. That is, the plant control monitoring system notifies an operator of a corresponding alarm when measurement data from the plant deviates from a predetermined normal range.

  Japanese Patent Laying-Open No. 2003-058967 (Patent Document 1) discloses an improvement on a method for setting a condition for generating an alarm. More specifically, Patent Document 1 discloses a function of filtering and displaying an alarm message according to an alarm level (importance rank) in a distributed control system that displays the alarm message in an alarm display window.

  In the plant operation as described above, it is preferable that notification of unnecessary alarms is reduced as much as possible from the viewpoint of the operator. Although not in the field of plant control, in the field of communication network maintenance, for example, as disclosed in Japanese Patent Application Laid-Open No. 04-355543 (Patent Document 2), in an alarm system that notifies the occurrence and recovery of a failure in a communication network, There has been proposed a method for preventing an unnecessary alarm by canceling a duplicate alarm of the same communication device that causes a failure alarm again immediately after the alarm is recovered.

  By the way, “queuing theory” was established for the purpose of mathematically analyzing the congestion of telephone lines. In the information communication field, network traffic analysis and network system design are performed by applying this queuing theory. For example, Japanese translations of PCT publication No. 2004-528765 (patent document 3) describes the scale and performance of a base station of a network for mobile communication devices in voice / data traffic using M / M / N type queuing theory. A method for evaluation is disclosed. More specifically, Japanese Patent Application Laid-Open No. 2004-228620 newly describes the evaluation of voice traffic performance using the well-known M / M / N type queuing theory. It is disclosed that performance in voice / data traffic can be evaluated with high accuracy by applying the same queuing theory.

JP 2003-058967 A Japanese Patent Laid-Open No. 04-355543 Special table 2004-528765 gazette

  It can be easily imagined that the plant operation load increases as the number of alarms to be handled by the operator increases in the plant operated through the plant control monitoring system as described above. However, it has been difficult to quantitatively evaluate the load of the plant operation such as how high the load of the plant operation is.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a plant operation load evaluation apparatus and a load evaluation program for quantitatively evaluating the plant operation load. is there.

  A load evaluation device for plant operation according to an aspect of the present invention is a history data indicating a history of alarms in a plant control monitoring system from a plant control monitoring system that receives measurement data from the plant and gives a control command to the plant. Obtaining means for obtaining the alarm occurrence interval distribution and alarm response time distribution in the plant based on the alarm occurrence time and return time included in the history data, alarm occurrence interval distribution and alarm response time distribution And an evaluation means for calculating a treatment time distribution for the alarm from the alarm and calculating a load of the plant operation from the alarm generation interval distribution and the treatment time distribution.

  Preferably, the evaluation unit calculates a value indicating an operation load by an operator in the plant based on an average alarm occurrence interval that is a parameter of the alarm occurrence interval distribution and an average treatment time that is a parameter of the treatment time distribution.

  Preferably, the alarm occurrence interval distribution calculating means includes an aggregating means for aggregating a plurality of alarms in a dependent relationship among the alarms included in the history data.

  More preferably, the aggregating unit aggregates alarms generated due to the same measurement data, into a single alarm among a plurality of alarms generated within a predetermined period.

  Preferably, the means for calculating the alarm occurrence interval distribution includes means for calculating an occurrence interval from the occurrence of each alarm to the occurrence of a subsequent alarm for the alarms included in the history data, and arithmetic for the calculated occurrence intervals. Means for calculating an average alarm occurrence interval that is a parameter of the alarm occurrence interval distribution by any one of an average, logarithmic average, and median method.

  Preferably, the alarm response time distribution calculating means calculates, from alarms corresponding to the occurrence of the alarm occurrence, alarms included in the history data that are within a predetermined time period from the occurrence of the alarm to the return. Includes means to exclude.

  Preferably, the alarm response time distribution calculating means includes means for calculating the time from the occurrence of each alarm to recovery as an alarm response time for each alarm included in the history data, and interval arithmetic for each calculated corresponding time. Means for calculating an average alarm response time which is a parameter of the alarm response time distribution by any one of an average, logarithmic average, and median method.

  A load evaluation program for plant operation according to another aspect of the present invention includes a history of alarms in a plant control monitoring system from a plant control monitoring system that receives measurement data from the plant and gives a control command to the plant. A step of obtaining historical data indicating the alarm, a step of calculating an alarm occurrence interval distribution and an alarm response time distribution in the plant based on an alarm occurrence time and a return time included in the history data, an alarm occurrence interval distribution and an alarm response A treatment time distribution for the alarm is obtained from the time distribution, and a step of calculating a load of the plant operation from the alarm generation interval distribution and the treatment time distribution is executed.

  According to the present invention, the load of plant operation can be quantitatively evaluated.

It is a schematic diagram which shows an example of the plant used as the object of load evaluation of plant operation in embodiment of this invention. It is a conceptual diagram for demonstrating quantitative evaluation of the load of plant operation in embodiment of this invention. It is a schematic diagram which shows the structure of the plant control monitoring system according to embodiment of this invention. It is a schematic diagram which shows the structure of the load evaluation apparatus according to embodiment of this invention. It is a block diagram which shows the control structure which the load evaluation apparatus according to embodiment of this invention provides. It is a flowchart which shows the whole process procedure performed in the load evaluation apparatus according to embodiment of this invention. It is a figure for demonstrating the aggregation process in the load evaluation method of the plant operation according to embodiment of this invention. It is a figure for demonstrating the concrete data processing of the pre-process in the load evaluation method of the plant operation according to embodiment of this invention. It is a figure for demonstrating the effect of the aggregation process in the load evaluation method of the plant operation according to embodiment of this invention. It is a figure for demonstrating the instantaneous return alarm elimination process in the load evaluation method of the plant operation according to embodiment of this invention. It is a figure for demonstrating the effect of the instant return alarm elimination process in the load evaluation method of the plant operation according to embodiment of this invention. It is a flowchart which shows the more detailed procedure of the pre-processing (step S2) shown in FIG. 13 is a flowchart showing a more detailed procedure of a master table generation process (step S21) in FIG. 13 is a flowchart showing a more detailed procedure of processing table generation processing (step S22) in FIG. 13 is a flowchart showing a more detailed procedure of generation processing (step S23) of alarm generation interval data and alarm response time data in FIG. It is a figure which shows the calculation example of the average alarm generation | occurrence | production interval and the average alarm corresponding | compatible time in the load evaluation method of the plant operation according to embodiment of this invention. It is a figure which shows an example of the alarm generation interval distribution computed using the load evaluation method of the plant operation according to embodiment of this invention. It is a figure which shows an example of the alarm corresponding | compatible time distribution calculated using the load evaluation method of the plant operation according to embodiment of this invention.

  Embodiments of the present invention will be described in detail with reference to the drawings. In addition, about the same or equivalent part in a figure, the same code | symbol is attached | subjected and the description is not repeated.

<A. About plant and plant control monitoring system>
First, a plant and a plant control monitoring system targeted by the load evaluation device for plant operation according to the present embodiment will be described.

  In various plants such as petrochemical plants and chemical plants, operations are generally performed by an operator through a plant control monitoring system. The plant control monitoring system is equipped with an alarm function in addition to the function of automatically controlling the plant. For example, when a measured value acquired from a plant deviates from a preset normal range, an alarm is sounded to notify the plant abnormality.

  FIG. 1 is a schematic diagram illustrating an example of a plant that is a target for load evaluation of a plant operation in the embodiment of the present invention. FIG. 1 shows an example of a plant 1 including a drum 10. The plant 1 is provided with a plant control monitoring system SYS. The plant control monitoring system SYS receives measurement data from the plant 1 and gives a control command to the plant 1. Typically, the plant control monitoring system SYS includes a control device 200 and a display operation device 300.

  More specifically, the plant 1 is provided with a sensor 12 that measures a liquid level such as a substance existing in the drum 10 and a sensor 18 that measures a flow rate of the substance discharged from the drum 10. . Measurement values (measurement data) measured by these sensors 12 and 18 are input to the control device 200.

  Further, the plant 1 is provided with a pump 14 and a flow rate control valve 16 in a pipe through which a substance discharged from the drum 10 flows. The start / stop control of the pump 14 and the opening adjustment of the flow control valve 16 are performed according to a control command from the control device 200.

  The display operation device 300 is arranged in an operation room or the like, notifies the operator of the operation status of the plant, an alarm generated in the plant, and the like, and accepts an operation by the operator. Details of the control device 200 and the display operation device 300 will be described later.

  In the above-mentioned plant, even if the operating conditions such as production volume and production speed are constant, due to unpredictable disturbances such as changes in outside air temperature, changes in raw material composition, changes in catalyst activity in the reactor, scaling, The state of the process changes. Therefore, such a change in the state of the process is detected by an alarm of the plant control monitoring system, and an operator takes measures to return the state of the process appropriately.

  In a continuous plant or the like, an event-driven operation mode is generally adopted in which when an alarm is generated, an operator performs an operation (treatment) for the generated alarm. In such an event-driven type operation mode, if the number of alarms generated is large, the number of operations (treatments) for the alarms increases, and the plant operation load on the operator increases.

<B. Overview>
Next, an outline of a method for quantitatively evaluating the load of the plant operation according to the present embodiment will be described. In the present embodiment, the plant data is acquired from the plant control monitoring system SYS, and the “queue theory” is applied, and the load of the plant operation is determined from the occurrence time and return time of each alarm included in the plant data. evaluate. Note that “return” means a state in which the measured value in the alarm state returns to the normal range. More specifically, history data indicating an alarm history in the plant control monitoring system SYS (an alarm / event log file 150 described later) is acquired, and based on the alarm occurrence time and return time included in the history data, An alarm generation interval distribution and an alarm response time distribution in the plant 1 are calculated.

  In this specification, the “alarm occurrence interval” means an occurrence interval from the occurrence of each alarm to the occurrence of a subsequent alarm, and the average alarm occurrence interval is a method as described below for a plurality of “alarm occurrence intervals”. This means the representative value of the alarm occurrence interval calculated by. “Alarm response time” means the time required from the occurrence of each alarm to recovery, and the average alarm response time is an alarm response calculated by the method described below for multiple “alarm response times”. It means the representative value of time. This alarm response time includes “waiting time” when the operator is taking action for an alarm generated before or separately from the alarm response time.

  When the alarm generation interval distribution and the alarm response time distribution are calculated, the treatment time distribution for the alarm is calculated using these. The “treatment time” means a net time necessary for treatment by the operator for each alarm. Therefore, the load of plant operation can be quantitatively evaluated using the alarm generation interval distribution and the average treatment time.

  More specifically, based on the queuing theory, the load of the plant operation by the operator in the plant 1 is calculated from the average alarm generation interval that is a parameter of the alarm generation interval distribution and the average treatment time that is the parameter of the treatment time distribution for the alarm. The index value shown is calculated.

  "Alarm occurrence interval distribution" and "Alarm response time distribution" or "Action time distribution" for alarms both use exponential functions, but any of them is Poisson distribution, Erlang distribution, or a distribution derived from them The use of is also considered. On the other hand, instead of the exponential distribution of the alarm interval distribution, the combination of the alarm occurrence distribution per unit time as the Poisson distribution and the alarm response time distribution or the treatment time distribution for the alarm as the exponential distribution is the same as the method of the present invention. It is clear.

  The “queuing theory” is a theory for analyzing, for example, a congestion phenomenon of a system that stochastically fluctuates such that a customer forms a queue for receiving a service using a mathematical model. This queuing theory is widely used as a technique for analyzing congestion and designing systems in fields such as information communication systems, production systems, and traffic systems.

  The inventors of the present application have newly found that the queuing theory can be applied to a plant operating in the event-driven type as described above. That is, for the purpose of evaluating the state of operation that can be handled by one operator in the plant, in the conventional queuing theory, the arrival of customers, the service of the window, and the queuing are respectively generated and treated for the alarm. Inventing a new method of evaluating the load of the plant operation using the M / M / 1 type queuing theory by considering it as a treatment waiting alarm.

  FIG. 2 is a conceptual diagram for explaining quantitative evaluation of the load of the plant operation in the embodiment of the present invention. FIG. 2 shows two-dimensional coordinates with the horizontal axis representing the number of treatments per unit time by the operator and the vertical axis representing the number of alarm occurrences per unit time. In this two-dimensional coordinate, if the number of alarm occurrences is large, the number of treatments is also increased, and it can be said that the operation load is higher in an unstable plant toward the upper left region.

  In the present embodiment, first, an alarm generation interval distribution and an alarm response time distribution are calculated from plant data (history data), and a treatment time for the alarm is calculated based on these. Further, an operation limit may be obtained from the alarm generation interval distribution and the treatment time distribution, and a plant operation load based on the operation limit may be obtained, or conversely, an operator margin ratio derived therefrom may be calculated.

  FIG. 2 shows an example in which the operation load index that quantitatively indicates the load of the plant operation increases as the number of alarm occurrences increases. The operation load index is a value standardized with an operation limit value of “1”. By using such an operation processing index, the plant operation load can be quantitatively evaluated.

  Note that the plant to which the present invention is applied is not limited to a continuous plant, and basically any plant that is operated in an event-driven manner that handles alarms. It is also applicable to.

<C. Application of queuing theory to quantitative evaluation of plant operation load>
Next, a mathematical process for quantitative evaluation of plant operation load using queuing theory will be described.

  First, assuming that the number of alarms generated within a unit time in the plant 1 follows the Poisson distribution, the alarm generation interval from the generation of each alarm to the generation of the subsequent (next) alarm has an exponential distribution. That is, when x is a time interval and λ is an average number of alarm occurrences per unit time (where 1 / λ is equal to the average alarm occurrence interval), the probability density function f of the average number of alarm occurrences λ per unit time f (X) can be expressed by the following equation (1).

From the equation (1), the ratio F (i) of the interval [a _i , b _i ] (where i = 1,..., N (N is the number of classes in the time interval)) in the frequency distribution. ) Can be expressed by the following equation (2).

  Next, in the plant 1, if the treatment time for an alarm that does not include a waiting time also follows an exponential distribution, the probability density function g (t) of the treatment time is expressed as follows, where t is time and 1 / μ is the average treatment time. (3).

  In the conventional queuing theory, the arrival interval of customers (corresponding to the alarm generation interval distribution in this embodiment) and the service time (corresponding to the treatment time distribution in this embodiment) both follow an exponential function. For example, it is known that the time from the arrival of the customer to the completion of the service, that is, the staying time of the customer has an exponential distribution as shown in the following equation (4). Accordingly, the alarm response time including the waiting time corresponding to the staying time of the customer also follows the exponential distribution. That is, the probability density function h (x) of the alarm response time including the waiting time can be expressed by equation (4).

Similar to the above equation (2), the ratio of the interval [a _i , b _i ] (where i = 1,..., N (N is the number of classes in the time interval)) in the frequency distribution. H (i) can be expressed by the following equation (5).

  Here, assuming that one operator is operating the plant 1 through the plant control monitoring system SYS, a “M / M / 1 type” queuing system is formed. Therefore, the ratio of the time during which an action corresponding to the generated alarm corresponding to the service utilization rate in the conventional queuing theory, that is, the operation load factor ρ can be expressed by the following equation (6). Here, the operation load factor ρ is a value from 0 to 1.

  On the other hand, the idle time when no action is taken for the alarm, that is, the operator's margin rate can be expressed by 1−ρ.

  Furthermore, the average number M of standing alarms corresponding to the number of customers in the queue including the customers receiving services in the queue theory can be expressed by the following equation (7). Here, the “standing alarm” means an alarm that is staying (including an alarm that has not been restored), including an alarm that is being treated by the operator.

  The theoretical operation limit that cannot be handled by one operator is the operation load factor ρ = 1. Therefore, the operation load factor ρ may be used as a quantitative value for the plant operation load evaluation as it is.

However, in the vicinity of the operation load factor ρ = 1, depending on the occurrence frequency and content of the alarm, standing alarms may be accumulated and a time zone in which it becomes impossible to respond may occur. Therefore, considering such a situation, an operation load index ξ represented by the equation (8) based on the practical operation limit ρ ₀ is considered. This operation load index ξ is an example of an index value indicating the load of the plant operation.

  The operation load index ξ is a value slightly larger than 0 from 1 and means that if the operation load index ξ exceeds 1, there is a risk that the operator cannot respond to the alarm.

The value of the practical operation limit ρ ₀ may be determined in consideration of the unique situation in each plant.

In this embodiment, according to the description of `` Engineering Equipment & Materials User's Association, Alarm Systems-A Guide to Design, management and Procurement, EEMUA Publication No. 191 (1999) '' which is the standard of alarm systems in plants. Set the value of the operation limit ρ ₀ .

More specifically, the EEMUA guidelines describe that the average number (M) of standing alarms should be 10 or less. Therefore, the practical operation limit ρ ₀ can be determined as 0.909 according to the above equation (7).

  That is, the above equation (8) is an equation for calculating an index value indicating the load of plant operation with the denominator as a constant, and the operation load index ξ based on EEMUA guidelines can be expressed by equation (9). it can.

  As described above, if the alarm occurrence interval and the alarm response time follow an exponential distribution, the operation load can be quantitatively evaluated based on the queue theory.

<D. Plant control monitoring system SYS>
Next, the configuration of the plant control monitoring system SYS shown in FIG. 3 will be described. FIG. 3 is a schematic diagram showing a configuration of a plant control monitoring system SYS according to the embodiment of the present invention.

  Referring to FIG. 3, plant control monitoring system SYS includes a control device 200 and a display operation device 300. The control device 200 acquires measurement data from the plant 1 and gives a control command to the plant 1. The display operation device 300 notifies the operator of the operation status of the plant, an alarm generated in the plant, and the like, and accepts an operation by the operator.

  More specifically, the control device 200 includes a processor 202, a main memory 204, a process input unit 206, a process output unit 208, a nonvolatile memory 210, a field bus interface (I / F) 220, a local memory Network interface (I / F) 222. Each component is connected to each other via an internal bus 224 so that data communication is possible.

  The processor 202 is a main body that executes processing related to the control device 200, and implements processing related to control of the plant 1 by executing a program stored in the nonvolatile memory 210. The processor 202 includes an arbitrary arithmetic device such as a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or an FPGA (Field-Programmable Gate Array).

  The main memory 204 includes a volatile memory such as a DRAM (Dynamic Random Access Memory), and functions as a work memory when the processor 202 executes a program.

  The process input unit 206 acquires measurement values (measurement data) measured by sensors or the like arranged in the plant 1. The process input unit 206 outputs the acquired measurement data to the processor 202 or the like.

  The process output unit 208 outputs a control command to an actuator or the like disposed in the plant 1 according to an internal command from the processor 202 or the like.

  The non-volatile memory 210 stores in a non-volatile manner a program executed by the processor 202 and data generated when the processor 202 executes a process. Typically, the nonvolatile memory 210 is a flash memory or the like.

  In the present embodiment, the nonvolatile memory 210 holds a plant control program 212, an alarm set value 214, and log data 216. The plant control program 212 includes a group of instructions for realizing processing related to control and monitoring of the plant 1 by being executed by the processor 202. The alarm set value 214 includes a parameter for setting a normal range of a measured value (process value) acquired from the plant 1. That is, the alarm set value 214 includes information for setting a normal range for each measurement value acquired from the plant 1. The log data 216 stores the contents when a preset event occurs in the control monitoring of the plant 1 by the control device 200. Typically, in the log data 216, when the acquired measurement value exceeds the normal range set by the alarm setting value 214, the target measurement value, the time of occurrence, the time of return, and the like are added. .

  The fieldbus interface 220 communicates with remote I / O (input / output devices) distributed in the plant 1.

  The local network interface 222 exchanges data with the display operation device 300, outputs the processing result of the processor 202 to the display operation device 300, and receives information input by the operator through the display operation device 300.

  On the other hand, the display operation device 300 includes a processor 302, a main memory 304, a display 306, an input unit 308, an audio output unit 310, a hard disk 311, and a local network interface (I / F) 312. Each component is connected to each other via an internal bus 314 so that data communication is possible.

  The processor 302 is a main body that executes processing related to the display / operation device 300, and implements processing necessary for monitoring control of the plant 1 by the operator by executing a program installed in advance. The processor 302 is configured by an arbitrary arithmetic device such as a CPU, MPU, or FPGA.

  The main memory 304 includes a volatile memory such as a DRAM, and functions as a work memory when the processor 302 executes a program.

  The display 306 is a device that visually notifies the operator of various types of information. The display 306 typically includes a CRT (Cathode Ray Tube), a liquid crystal display, a plasma display, an organic EL (Electro Luminescence) display, and the like.

  The input unit 308 is a device that receives an operation from an operator. The input unit 308 typically includes a keyboard, a mouse, a touch panel, and the like.

  The audio output unit 310 is an apparatus that audibly notifies various information to the operator. The voice output unit 310 typically includes a speaker, an amplifier, and the like, and includes a voice synthesizer and the like when outputting an alarm message as a voice.

  The hard disk 311 stores an alarm / event log file 150. The alarm / event log file 150 is synchronized with the log data 216 in the control device 200 via the local network system interface 312, and the contents are written in the alarm / event log file 150.

<E. Load evaluation device>
Next, the configuration of load evaluation apparatus 100 that quantitatively evaluates the load of the plant operation according to the present embodiment will be described. The load evaluation method for plant operation according to the present embodiment is basically executed by load evaluation apparatus 100. Various types of information processing apparatuses can be used as the load evaluation apparatus 100. In the present embodiment, an example will be described in which the load evaluation apparatus 100 is realized using a personal computer having a general-purpose architecture. Instead of a method realized by a general-purpose personal computer, part or all of the processing may be realized by using dedicated hardware, or this function may be implemented in the plant control monitoring system SYS. Good.

  FIG. 4 is a schematic diagram showing a configuration of the load evaluation device 100 according to the embodiment of the present invention. Referring to FIG. 4, load evaluation apparatus 100 includes a processor 102, a main memory 104, a display 106, an input device 108, a hard disk 110, a network interface (I / F) 116, and a data reader 118. Including. These components are connected to each other via an internal bus 122 so that data communication is possible.

  In the load evaluation apparatus 100, the processor 102 as an arithmetic unit executes a load evaluation program 114 preliminarily stored in the hard disk 110 or the like after developing it in the main memory 104 or the like, thereby executing a load of plant operation as described later. Realize the evaluation method. The load evaluation program 114 is read from an arbitrary recording medium (card device 120) via the data reader 118, an optical reader (not shown), or the like, and is installed in the hard disk 110. Alternatively, the load evaluation program 114 may be downloaded from an external server via the network via the network interface 116.

  In addition to the load evaluation program 114, an operating system (OS) 112 is also installed on the hard disk 110. Basically, the load evaluation program 114 is executed in an environment provided by the processor 102 executing the operation system. The load evaluation program 114 may be coded so as to perform processing using a module provided by the operation system 112. In this case, the module provided by the operation system 112 is not included in the load evaluation program 114 itself. It is obvious that even such a load evaluation program 114 is included in the scope of the present invention.

  The display 106 displays various information to the user. The display 106 typically includes a CRT (Cathode Ray Tube), a liquid crystal display, a plasma display, an organic EL (Electro Luminescence) display, and the like.

  The input device 108 is a device that receives an operation from a user. The input device 108 typically includes a keyboard and a mouse.

  The network interface 116 exchanges data with external devices through the network. The network interface 116 is typically composed of Ethernet (registered trademark) or the like. In the present embodiment, the load evaluation apparatus 100 acquires plant data (alarm / event log file 150) from the plant control monitoring system SYS. This plant data may be input to the load evaluation device 100 via the network interface 116.

  The data reader 118 is configured to be detachable from the card device 120, reads data stored in the card device 120, outputs the data to the processor 102, and writes data given from the processor 102 or the like to the card device 120. The card device 120 typically includes an SD (Secure Digital) card, a compact flash (registered trademark), or the like. Plant data (alarm / event log file 150) from the plant control monitoring system SYS may be input to the load evaluation apparatus 100 via the card device 120.

<F. Control structure>
Next, a control structure mounted on the load evaluation apparatus 100 will be described.

  FIG. 5 is a block diagram showing a control structure provided by load evaluation device 100 according to the embodiment of the present invention. Each functional block shown in FIG. 5 is typically realized by the processor 102 of the load evaluation apparatus 100 executing a load evaluation program 114 stored in the hard disk 110 or the like.

  The load evaluation apparatus 100 includes, as its control structure, an acquisition unit 160, a preprocessing unit 162, an alarm generation interval distribution calculation unit 168, an alarm response time distribution calculation unit 170, a treatment time distribution calculation unit 172, and a load index. And a calculation unit 174.

  The acquisition unit 160 acquires the alarm / event log file 150 from the plant control monitoring system SYS (display operation device 300). The alarm / event log file 150 is history data indicating a history related to an alarm in the plant control monitoring system SYS. As described above, the alarm / event log file 150 may be transferred via the network interface 116 via the network, or may be input while being carried on a recording medium such as the card device 120.

  The preprocessing unit 162 performs preprocessing for calculating the alarm occurrence interval distribution and the alarm response time distribution for the alarm / event log file 150 acquired by the acquisition unit 160. More specifically, the preprocessing unit 162 includes an aggregation unit 164 and an instantaneous return alarm exclusion unit 166.

  The aggregating unit 164 aggregates a plurality of alarms (hereinafter also referred to as “subordinate alarms”) having a dependency relationship among the alarms included in the alarm / event log file 150. More specifically, the aggregating unit 164 aggregates alarms generated due to the same measurement data, into a single alarm among a plurality of alarms generated within a predetermined period.

  The instantaneous return alarm exclusion unit 166 excludes, from the calculation target of the alarm corresponding time distribution, the collected alarms whose period from the occurrence to the return is within a predetermined time.

Details of these pre-processing will be described later.
The alarm occurrence interval distribution calculation unit 168 calculates the alarm occurrence interval distribution from the result of the preprocessing of the alarm / event log file 150 by the aggregation unit 164. Also, the alarm response time distribution calculation unit 170 calculates the alarm response time distribution from the result of the pre-processing by the aggregating unit 164 and the pre-processing by the instantaneous recovery alarm exclusion unit 166. These calculation units calculate the alarm generation interval distribution and the alarm response time distribution in the plant 1 based on the alarm occurrence time and / or return time included in the alarm / event log file 150.

  The treatment time distribution calculation unit 172 calculates a treatment time distribution for the alarm based on the alarm occurrence interval distribution and the alarm response time distribution. The load index calculation unit 174 calculates a value (load index) indicating the operation load by the operator in the plant 1 based on the alarm generation interval distribution and the treatment time distribution. More specifically, the load index calculation unit 174 uses an average alarm occurrence interval (1 / λ) that is a parameter of the alarm occurrence interval distribution and an average treatment time (1 / μ) that is a parameter of the treatment time distribution. Then, the operation load factor ρ and the operation load index ξ are calculated according to the above equations (6) to (8). The load index calculation unit 174 outputs these calculation results as load index results.

<G. Overall procedure>
Next, an overall processing procedure for quantitatively evaluating the load of the plant operation executed in the load evaluation apparatus 100 will be described.

  FIG. 6 is a flowchart showing an overall processing procedure executed in load evaluation device 100 according to the embodiment of the present invention. Each step shown in FIG. 6 is executed by the processor 102 of the load evaluation apparatus 100.

  First, the processor 102 (acquisition unit 160) of the load evaluation apparatus 100 acquires plant data (alarm / event log file 150) from the plant control monitoring system SYS (step S1). Subsequently, the processor 102 (preprocessing unit 162) performs preprocessing for calculating the alarm occurrence interval distribution and the alarm response time distribution for the acquired alarm / event log file 150 (step S2).

  The processor 102 (alarm occurrence interval distribution calculation unit 168) calculates an average alarm occurrence interval, which is a parameter of the alarm occurrence interval distribution, from the result of preprocessing the alarm / event log file 150 (step S3). In parallel with the process of step S3, the processor 102 (alarm response time distribution calculation unit 170) calculates an average alarm response time that is a parameter of the alarm response time distribution from the result of preprocessing the alarm / event log file 150. (Step S4). Note that step S3 and step S4 may be executed in series, and in that case, the execution order may be any.

  Then, the processor 102 (treatment time distribution calculation unit 172) calculates an average treatment time that is a parameter of the treatment time distribution for the alarm based on the average alarm occurrence interval and the average alarm response time (step S5). Further, the processor 102 (load index calculation unit 174) calculates a value (operation load index) indicating the operation load by the operator in the plant 1 based on the average alarm occurrence interval and the average action time (step S6).

  Finally, the processor 102 outputs a load index result including the operation load index (step S7). Then, the process ends.

<H. Pretreatment>
Next, details of the above-described preprocessing (step S2 shown in FIG. 6) will be described.

(H1: Aggregation process (exclusion process of dependency))
As described above, queuing theory is based on the premise of independent random events, so it is necessary to exclude the dependency of the target data. Hereinafter, alarm dependency in an actual plant and a method for eliminating the dependency will be described.

  FIG. 7 is a diagram for explaining the aggregation processing in the load evaluation method for plant operations according to the embodiment of the present invention. FIG. 7 shows an example of drum level control (LC). Upper limit (hereinafter referred to as “Hi”), upper upper limit (hereinafter referred to as “HH”), lower limit (hereinafter referred to as “Lo”), lower and lower limit (hereinafter referred to as “Hi”) for the measured value in drum liquid level control Assume that a total of four alarms (denoted as “LL”) are set.

  When the measured value by the liquid level sensor exceeds or falls below any of the alarm setting values due to process fluctuations such as disturbance, an alarm is generated and the corresponding alarm data is output to the alarm / event log file 150. The

  The operator checks the status of the process when any alarm occurs, silences the alarm generated by a confirmation operation (typically, pressing the confirmation button), and takes action (operation to return the process to a normal state). ). When the process returns to the normal range, the alarm is recovered (hereinafter also referred to as “Recover”), and data indicating the recovery is output to the alarm / event log file 150.

  When the process fluctuation (liquid level fluctuation) as shown in the trend graph of FIG. 7 occurs, the operator treats it as a single liquid level fluctuation phenomenon, but in the alarm / event log file 150, the operator is dependent on each other. It is recorded that some four alarms occurred and returned.

  Therefore, when calculating the alarm occurrence distribution from the alarm data (alarm / event log file 150), it is preferable to eliminate at least the dependency of these data.

  In the present embodiment, for a series of alarms that are subordinate to each other, the series of alarms is collected into the alarm that has occurred first. More specifically, among the series of alarms at the same location (measurement data / tag) that occurred during a predetermined time (for example, 2 minutes), only the first alarm is regarded as a valid alarm occurrence. Assuming other alarms are invalid alarms.

  On the other hand, the alarm response time is the time from the occurrence of a series of alarms to normal recovery, assuming that event-driven operation is performed. That is, as shown in FIG. 7, the time is from when the first alarm (Li alarm of the liquid level control LC) is generated until the last return data (Lo alarm Recover of the liquid level control LC) is obtained.

  FIG. 8 is a diagram for explaining specific data processing of preprocessing in the load evaluation method for plant operation according to the embodiment of the present invention. Hereinafter, with reference to FIG. 8, a method for processing alarms generated independently of each other will be described in detail with reference to a specific example.

  Referring to FIG. 8, the alarm / event log file 150 includes, in addition to the target process alarm and its return data, logs such as operations performed by the operator and control mode change of the control device 200. The alarm / event log file 150 is stored in the hard disk 311 as a sequential file. Each record of the alarm / event log file 150 includes a code indicating a type of alarm / event such as occurrence or return of an alarm, a time, a tag, and a message sentence including the type of alarm. In order to simplify the explanation, the alarm / event log file 150 shown in FIG. 8 shows only the alarm occurrence and its return data.

  The load evaluation device 100 reads the alarm occurrence and return data included in the alarm / event log file 150 one record at a time. If the read data is the alarm occurrence or return data, the contents are stored in the master table 152. Registered sequentially. The master table 152 stores a combination of an occurrence time and a return time for each alarm. More specifically, each record of the master table 152 includes an alarm occurrence time, a return time, a tag, and an alarm type. The return time is written in the corresponding alarm data by searching from the back to the front of the master table 152 using the tag and alarm type as keys, that is, searching from the newest time to the oldest direction. . In this way, the master table 152 in which the occurrence time and the return time correspond to each alarm is generated.

  Next, among the alarms included in the master table 152, a series of alarms that are subordinate to each other are aggregated. Referring to the master table 152 shown in FIG. 8, the Hi alarm of the liquid level control LC is generated at 0:00:27, and the HH alarm of the liquid level control LC is generated at 0:00:30. The occurrence interval between the two is 3 seconds, which is less than a predetermined time (for example, 2 minutes). Therefore, it is determined that the alarm is in a dependent relationship, and the occurrence of the alarm is integrated into the first Hi alarm. Similarly, the Lo alarm of the liquid level control LC is generated at 0:02:29, and the LL alarm of the liquid level control LC is generated at 0:02:31. Since these occurrence intervals are also less than a predetermined time (for example, 2 minutes), the occurrence of the alarm is integrated into the first Lo alarm.

  In contrast, the Hi alarm of the liquid level control LC generated at 0:08:16 has an occurrence interval of 5 minutes 45 seconds with the alarm of the liquid level control LC generated before that, Since it exceeds the specified time (for example, 2 minutes), it is regarded as an independent alarm.

By such a procedure, alarm occurrence interval data 156 is generated.
FIG. 9 is a diagram for describing the effect of the aggregation processing in the load evaluation method for plant operations according to the embodiment of the present invention. More specifically, FIG. 9 shows the above-described (2) based on the frequency distribution of the alarm occurrence intervals acquired from the actual plant and the average alarm occurrence interval calculated from the alarm occurrence intervals acquired from the actual plant. ) And an exponential distribution curve drawn according to the equation. FIG. 9A shows a result when the above-described dependency elimination processing (aggregation processing) is not performed, and FIG. 9B illustrates the above-described dependency exclusion processing (aggregation processing). The result of performing is shown.

  As is clear from a comparison between FIG. 9 (A) and FIG. 9 (B), by performing processing (aggregation processing) that excludes alarms dependent on the same location (measurement data / tag), the occurrence interval of alarms is randomized. It can be regarded as an event, and the alarm generation interval can be approximated more accurately by an exponential distribution.

(H2: Instantaneous return alarm elimination processing)
Next, a description will be given of the instantaneous return alarm elimination processing that is executed to calculate the alarm response time distribution more accurately. When calculating the alarm response time distribution, in addition to the above-described processing for eliminating alarms that are generated depending on each other, it is necessary to eliminate the instantaneous return alarm that causes noise. “Instantaneous recovery alarm” means that after an alarm occurs, it will return to a normal state in a short period of time, and it cannot be considered that the operator has completed the treatment within the time from when the alarm occurs until it returns. It is an alarm.

  FIG. 10 is a diagram for explaining an instantaneous return alarm elimination process in the load evaluation method for plant operation according to the embodiment of the present invention. FIG. 10 shows an example of drum temperature control (TC). It is assumed that an alarm of a change rate of a measured value (hereinafter referred to as “Vel”) in drum temperature control is set. Hereinafter, with reference to FIG. 10, a state of an instantaneous return alarm generated by a change rate alarm and a method of eliminating it will be described.

  An alarm is generated when the rate of temperature change exceeds the set value of the alarm due to process fluctuations such as disturbance, and the alarm is restored when the normal state is restored. The change rate alarm is often used for the purpose of quickly detecting a change in the process, and the time from the occurrence of the alarm to the return of the alarm is not always the same as the time when the operator actually takes action.

  Since such an event becomes noise in the calculation of the alarm response time distribution, it is preferable to eliminate the instantaneous return alarm.

  Note that the instantaneous return alarm may include an upper / lower limit alarm that is not actually accompanied by an operation, which is caused by sensor noise or the influence of batch operation, in addition to the change rate alarm as described above.

  In the present embodiment, the alarm response time record whose alarm response time is less than a predetermined time (for example, 15 seconds) is excluded. Referring to FIG. 8 again, the temperature control TC which is the second record in the processing table 154 has an alarm response time because the alarm occurrence time is 0:00:42 and the alarm return time is 0:00:45. The time is 3 seconds. Since this alarm response time is less than a predetermined time (for example, 15 seconds), it is skipped when calculating the alarm response time distribution. That is, in the alarm response time data 158, the record of the temperature control TC is invalidated.

  FIG. 11 is a diagram for illustrating the effect of the instantaneous return alarm elimination processing in the load evaluation method for plant operation according to the embodiment of the present invention. More specifically, FIG. 11 shows the above-described (5) based on the frequency distribution of the alarm response time acquired from the actual plant and the average alarm response time calculated from the alarm response time acquired from the actual plant. ) And an exponential distribution curve drawn according to the equation. FIG. 11A shows the result when the above-described instantaneous return alarm exclusion process is not performed, and FIG. 11B shows the result when the above-described instantaneous return alarm exclusion process is performed.

  As is clear from a comparison between FIG. 11A and FIG. 11B, after performing processing (aggregation processing) excluding alarms dependent on the same location (measurement data / tag), an instantaneous return alarm is further generated. By performing the process of removing, the alarm response time can be regarded as a random event, whereby the alarm response time can be approximated more accurately with an exponential distribution.

(H3: Processing procedure)
Next, a processing procedure according to the preprocessing (step S2 shown in FIG. 6) for calculating the above-described average alarm occurrence interval and average alarm response time will be described.

  FIG. 12 is a flowchart showing a more detailed procedure of the preprocessing (step S2) shown in FIG. Referring to FIG. 12, processor 102 (preprocessing unit 162) of load evaluation apparatus 100 generates master table 152 from alarm / event log file 150 (step S21). Subsequently, the processor 102 (aggregation unit 164) generates the processing table 154 while aggregating the master table 152 (step S22). Then, the processor 102 (instantaneous recovery alarm exclusion unit 166) generates alarm generation interval data 156 from the processing table 154, and generates alarm response time data 156 while excluding the instantaneous recovery alarm (step S23). Then, the process returns.

  Hereinafter, a more detailed processing procedure of steps S21, S22, and S23 will be described. FIG. 13 is a flowchart showing a more detailed procedure of the generation process (step S21) of the master table 152 of FIG. FIG. 14 is a flowchart showing a more detailed procedure of the generation process (step S22) of the process table 154 of FIG. FIG. 15 is a flowchart showing a more detailed procedure of the generation process (step S23) of the alarm generation interval data 156 and the alarm response time data 158 of FIG.

  First, the generation process of the master table 152 will be described with reference to FIG. The processor 102 (preprocessing unit 162) of the load evaluation apparatus 100 reads the first record from the alarm / event log file 150 (step S211). The processor 102 (preprocessing unit 162) determines the type of the read record (step S212). That is, it is determined whether the read record is “alarm data” (alarm occurrence), “return data” (alarm return), or “other”.

  When the read record is “alarm data” (“alarm data” in step S212), the processor 102 (preprocessing unit 162) extracts information on the occurrence time, tag, and type from the record, and the master table. Write to 152 (step S213). Then, the process proceeds to step S215.

  On the other hand, when the read record is “return data” (“return data” in step S212), the processor 102 (pre-processing unit 162) uses the tag and alarm type included in the record as keys. A search is performed from the back to the front of the master table 152 to identify the corresponding alarm in the master table 152, and the return time of the record is written in the master table 152 in association with the identified alarm (step S214). Then, the process proceeds to step S215.

  If the read record is neither “alarm data” nor “return data” (“other” in step S212), the process proceeds to step S215 without any processing.

  In step S215, the processor 102 (preprocessing unit 162) determines whether or not the read record is the last record of the alarm / event log file 150 (step S215). If the read record is not the last record of the alarm / event log file 150 (NO in step S215), the processor 102 (preprocessing unit 162) reads the next record from the alarm / event log file 150. Read (step S216). And the process after step S212 is repeated.

  On the other hand, if the read record is the last record of the alarm / event log file 150 (YES in step S215), the process returns.

  Through the above processing, the master table 152 is generated from the alarm / event log file 150.

  Next, processing for generating the processing table 154 will be described with reference to FIG. The processor 102 (aggregation unit 164) of the load evaluation apparatus 100 reads the first record from the master table 152 (step S221). The processor 102 (aggregation unit 164) extracts a tag from the read record, and the same tag as the extracted tag from the record in the master table 152 until a predetermined time (for example, 2 minutes). A record (alarm) having “” is searched (step S222). Then, the processor 102 (aggregation unit 164) determines whether or not there is a record (alarm) having the same tag as the tag extracted until a predetermined time (step S223).

  When there is a record (alarm) having the same tag as the tag extracted until a predetermined time (YES in step S223), the processor 102 (aggregation unit 164) selects the same tag. Of the plurality of records, the later return time is overwritten to the corresponding return time position in the processing table 154 (step S224). That is, the processor 102 (aggregation unit 164) determines that the record (alarm) is a dependent alarm, and updates only the corresponding return time in the processing table 154. As the return time to be updated, the return time of the previous record is compared with the return time of the current record, and the later time is written.

  On the other hand, when there is no record (alarm) having the same tag as the tag extracted until a predetermined time (NO in step S223), the processor 102 (aggregation unit 164) A new record corresponding to (alarm) is added to the processing table 154 (step S225). That is, if there is no alarm having the same tag, the processor 102 (aggregation unit 164) determines that the alarm is an independent new alarm and writes the content in the processing table 154.

  After execution of step S224 or S225, the processor 102 (aggregation unit 164) calculates an alarm occurrence interval and an alarm response time for the target record, and writes the values in the processing table 154 (step S226). That is, the alarm occurrence interval that is the difference from the occurrence time of the previous alarm and the alarm response time that is the difference between the return time and the occurrence time are written.

  Subsequently, the processor 102 (aggregation unit 164) determines whether or not the read record is the last record in the master table 152 (step S227). If the read record is not the last record in the master table 152 (NO in step S227), the processor 102 (preprocessing unit 162) reads the next one record from the master table 152 (step S228). And the process after step S222 is repeated.

  On the other hand, if the read record is the last record in the master table 152 (YES in step S227), the process returns.

Through the above processing, the processing table 154 is generated from the master table 152.
Finally, with reference to FIG. 15, the generation processing of the alarm occurrence interval data 156 and the alarm response time data 158 will be described. Basically, the alarm occurrence interval data can be generated by extracting the tag and alarm occurrence interval data from the processing table 154. On the other hand, when calculating the alarm response time data 158, an alarm response time record whose alarm response time is less than a predetermined time (for example, 15 seconds) is excluded.

  The processor 102 (preprocessing unit 162) of the load evaluation device 100 reads the first record from the processing table 154 (step S231). The processor 102 (preprocessing unit 162) extracts an alarm occurrence interval and a tag from the read record and outputs them as alarm occurrence interval data 156 (step S232).

  Subsequently, the processor 102 (preprocessing unit 162) determines whether or not the alarm response time of the record read in the processing table 154 is less than a predetermined time (for example, 15 seconds) (step S233).

  If the alarm response time is not less than the predetermined time (NO in step S233), the processor 102 (preprocessing unit 162) extracts the alarm response time and the tag and outputs them as alarm response time data 158. (Step S234).

  On the other hand, if the alarm response time is less than the predetermined time (YES in step S233), the process of step S234 is skipped.

  Subsequently, the processor 102 (preprocessing unit 162) determines whether or not the read record is the last record in the processing table 154 (step S235). If the read record is not the last record in the processing table 154 (NO in step S235), the processor 102 (preprocessing unit 162) reads the next one record from the processing table 154 (step S236). And the process after step S232 is repeated.

  On the other hand, if the read record is the last record in the processing table 154 (YES in step S235), the process returns.

  Through the above processing, alarm generation interval data 156 and alarm response time data 158 are generated from the processing table 154.

<I. Calculation process>
Next, details of the calculation processing of the average alarm occurrence interval and the average alarm response time shown in step S4 of FIG. 6 will be described.

(I1: Alarm generation interval and average alarm generation interval)
Next, processing for calculating the alarm occurrence interval and the average alarm occurrence interval will be described. As described with reference to FIGS. 8 and 15 and the like, the alarm occurrence interval data 156 is generated from the processing table 154 as shown in FIG. The alarm generation interval data 156 stores the elapsed time from the timing at which the immediately preceding alarm is generated for each alarm. That is, the load evaluation apparatus 100 calculates the occurrence interval (alarm occurrence interval) from the occurrence of each alarm to the occurrence of the subsequent alarm for the alarm included in the history data (alarm / event log file 150). Then, the load evaluation apparatus 100 typically calculates an average alarm occurrence interval from each alarm occurrence interval using one of the following methods.

(1) Arithmetic average of alarm occurrence intervals When each alarm occurrence interval is x _i (where i = 1,..., N _a (N _a is the total number of alarms after dependent alarm aggregation processing)), the alarm occurrence interval Can be calculated according to the following equation (10).

  Note that the alarm generation interval distribution shown in FIG. 9 is calculated using an arithmetic average.

  The average alarm occurrence interval calculated by the arithmetic average corresponds to the reciprocal of the total number of alarms generated per unit time. Therefore, for example, the total number of alarms generated per day may be counted, and a value obtained by dividing one day by the total number may be set as the average alarm generation interval.

(2) Logarithmic average of alarm occurrence intervals When the alarm occurrence interval distribution slightly deviates from the exponential distribution, the alarm occurrence intervals are biased. In this case, the simple arithmetic averaging method as described above is easily affected by data having a large alarm generation interval. In such a case, the average alarm occurrence interval may be calculated using the logarithmic average of the alarm occurrence intervals. Specifically, as shown in the following equation (11), the average alarm occurrence interval can be calculated by taking the logarithm of the data, obtaining the average value, and then restoring it using an exponential function.

(3) Median Alarm Generation Interval As a method that can be expected to achieve the same effect as in (2) logarithmic averaging described above, the median alarm generation interval may be adopted as the average alarm occurrence interval. If the function for obtaining the median is median, the average alarm occurrence interval can be calculated according to the following equation (12).

  As described above, (3) by calculating the average alarm occurrence interval using the median value, the processing amount related to the average alarm occurrence interval can be reduced.

  As described above, the load evaluation apparatus 100 calculates the average alarm occurrence interval for each calculated alarm occurrence interval by any one of the arithmetic average, logarithmic average, and median method.

(I2: Alarm response time and average alarm response time)
Similar to the above-described processing for calculating the average alarm occurrence interval, the average alarm response time is also calculated using one of the following methods.

  First, the load evaluation apparatus 100 calculates a corresponding time (alarm corresponding time) from the occurrence of each alarm to the return for each alarm included in the history data (alarm / event log file 150). More specifically, as described with reference to FIGS. 8 and 15, the alarm response time data 158 is generated from the processing table 154 as shown in FIG. 8. The alarm response time data 158 stores the time required from occurrence to recovery for each alarm.

(1) Interval Arithmetic Average of Alarm Response Times Each alarm response time is represented by t _i (where i = 1,..., N _b (N _b is the total number of alarms after subordinate alarm aggregation processing and instantaneous return alarm processing) ), And t ′ _j (where j = 1,..., N _b ′ (N _b ′ is a subordinate)) is a set of data of the section time (in this case, the alarm response time is 0 to less than 25 minutes) Assuming that the number of alarms is less than 25 minutes among the alarms after the alarm aggregation process and the instantaneous return alarm elimination process), the interval arithmetic average of the alarm response time can be calculated according to the following equation (13).

  Note that the alarm response time distribution shown in FIG. 11 described above is calculated using the interval arithmetic average.

(2) Logarithmic average of alarm response time Similar to the description of the alarm occurrence interval distribution described above, if the alarm response time distribution slightly deviates from the exponential distribution, the alarm response time is biased. In this case, the simple interval arithmetic average as described above is easily affected by data having a long alarm response time. In such a case, the average alarm response time may be calculated using the logarithmic average of the alarm response time. Specifically, as shown in the following equation (14), the average alarm response time can be calculated by calculating the average value by taking the logarithm of the data and returning it to the original value using an exponential function. According to this method, it is not necessary to consider the interval width as in the case of using (1) interval arithmetic average described above.

(3) Median Alarm Response Time As a method that can be expected to have the same effect as when using (2) logarithmic average, the median alarm response time may be adopted as the average alarm response time. If the function for obtaining the median is median, the average alarm response time can be calculated according to the following equation (15).

  Thus, by calculating the average alarm response time using (3) the median value, the processing amount related to the average alarm response time can be reduced.

  As described above, the load evaluation apparatus 100 calculates an average alarm response time for each calculated alarm response time by any one of the method of arithmetic arithmetic mean, logarithmic average, and median.

(I3: calculation example)
An example of the average alarm occurrence interval and the average alarm response time calculated by using the respective methods as described above will be shown.

  FIG. 16 is a diagram showing a calculation example of an average alarm occurrence interval and an average alarm response time in the plant operation load evaluation method according to the embodiment of the present invention. FIG. 16A shows an example of an average alarm occurrence interval calculated using the arithmetic average, logarithmic average, and median, and FIG. 16B shows the interval arithmetic average, logarithmic average, and median, respectively. An example of the average alarm response time calculated using the above is shown.

(I4: correction process)
When the average alarm occurrence interval and the average alarm response time are calculated by any of the methods described above, and the calculated exponential distribution curve deviates from the frequency distribution acquired from the plant data, the correction parameter Using α and β, the exponential distribution curve may be corrected according to the following equations (16) and (17).

  Here, λ ′ indicates the average alarm occurrence interval calculated by any one of the above-described average alarm occurrence intervals, and (μ−λ) ′ is any one of the above-described average alarm response times. The calculated average alarm response time shall be indicated. The correction parameters α and β are adjustment parameters for setting so as to match the frequency distribution. When each of the average alarm occurrence interval and the average alarm response time matches the corresponding frequency distribution, it is set to 1 respectively, but when it does not match, it is set to a value suitable for correction. . As an example, the correction parameter α is set in a range of 0.5 <α <2.0, and the correction parameter β is set in a range of 0.5 <α <2.0.

<J. Evaluation process>
Next, details of the evaluation process shown in steps S5 and S6 of FIG. 6 will be described.

  According to the procedure as described above, the average alarm generation interval (1 / λ) and the average alarm response time (1 / (μ−λ)) are calculated from the data related to the generation and return of each alarm. The average treatment time (1 / μ) can be calculated from these average values. Further, the operation load factor ρ and the operation load index ξ of the target plant are calculated according to the above-described equations (6) and (9).

  An example in which the load evaluation method according to the present embodiment is applied using an alarm of plant data obtained from an actual plant and its return data will be described below.

  FIG. 17 is a diagram showing an example of an alarm generation interval distribution calculated using the load evaluation method for plant operation according to the embodiment of the present invention. FIG. 18 is a diagram showing an example of an alarm response time distribution calculated using the plant operation load evaluation method according to the embodiment of the present invention.

  In addition, for the four plants (Plant A, Plant B, Plant C, and Plant D) evaluated in FIGS. 17 and 18, the average alarm occurrence interval, the average alarm response time, the average treatment time excluding the waiting time, and the operation load of each plant The calculation results of the rate and operation load index are shown below.

  Each plant is actually operated by one operator, and the operation load factor and the operation load index show relatively low values. I can say that.

<K. Application example>
As described above, according to the plant operation load evaluation method according to the present embodiment, the plant operation load in each plant can be quantitatively evaluated. Furthermore, it is possible to evaluate whether or not the plant operations can be integrated using the result of such quantitative evaluation. For example, in a situation where a dedicated operator is arranged and operated in each of a plurality of plants, it is possible to evaluate whether or not a plurality of plants can be operated by only one of the operators.

  Hereinafter, as an application example according to the present embodiment, a process for predicting the load of the plant operation in the case where the above-mentioned Plant A and Plant B are integrated and in the case where the Plant C and Plant D are integrated will be described.

  Using the results for each plant described above, prediction results when Plant A and Plant B are operated by one operator, and when Plant C and Plant D are operated by one operator are shown below.

  The average alarm occurrence interval generated when Plant A and Plant B are integrated can be calculated as 205.7 seconds from the average alarm occurrence intervals (1 / λ) of Plant A and Plant B as shown in the above table. Similarly, the average alarm occurrence interval that occurs when Plant C and Plant D are integrated can be calculated as 367.3 seconds.

  On the other hand, since the average treatment time (1 / μ) can be estimated from Table 1 to approximately 180 seconds (3 minutes) to 240 seconds (4 minutes), it is assumed here to be 180 seconds as an example. Thus, the average alarm response time (1 / (μ−λ)) can be calculated, and specifically, 1440.0 seconds and 352.9 seconds, respectively.

  Furthermore, based on these numerical values, the operation load factor (ρ), the operation load index (ξ), and the average number of standing alarms (M) can be calculated. Specifically, when Plant A and Plant B are integrated, the operation load index becomes 0.963, which approaches the operation limit. In contrast, when Plant C and Plant D are integrated, the operation load index is 0.539, and there is still room for the operation limit. In this example, even if Plant C and Plant D are integrated, the load of plant operation is lower than Plant B.

  As a comparison, if the average treatment time (1 / μ) is estimated as 240 seconds uniformly, when Plant A and Plant B are integrated, the operation limit is exceeded, and one operator cannot cope with it. Therefore, the result that the effect by operation integration of a plant control monitoring system is not acquired is obtained.

<L. Advantages of this embodiment>
According to the present embodiment, plant data can be acquired from a plant control monitoring system, and queuing theory can be applied to quantitatively evaluate the load of plant operation from the occurrence time and return time of each alarm included in the plant data. . By utilizing such a quantitatively evaluated plant operation load, the number of operators required for plant operation can be optimized. Furthermore, it is possible to estimate a plant operation load when a plurality of plants are integrated so that a common operator operates.

  By introducing such a quantitative evaluation method for plant operation load, it is possible to design a staffing arrangement for plant operation.

  10 drums, 12, 18 sensors, 14 pumps, 16 flow control valves, 100 load evaluation devices, 102, 202, 302 processors, 104, 204, 304 main memory, 106, 306 displays, 108 input devices, 110, 311 hard disks, 112 operation system, 114 load evaluation program, 116 network interface, 118 data reader, 120 card device, 122, 224, 314 internal bus, 150 alarm / event log file, 152 master table, 154 processing table, 156 alarm occurrence interval data, 158 Alarm response time data, 160 acquisition unit, 162 preprocessing unit, 164 aggregation unit, 166 Instantaneous recovery alarm exclusion unit, 168 Alarm occurrence interval Distribution calculation unit, 170 Alarm response time distribution calculation unit, 172 Treatment time distribution calculation unit, 174 Load index calculation unit, 200 Control device, 206 Process input unit, 208 Process output unit, 210 Non-volatile memory, 212 Plant control program, 214 Alarm set value, 216 log data, 220 fieldbus interface, 222, 312 local network interface, 300 display operation device, 308 input unit, 310 audio output unit, SYS plant control monitoring system.

Claims (8)

An acquisition means for acquiring history data indicating a history relating to an alarm in the plant control monitoring system from a plant control monitoring system that inputs measurement data from the plant and gives a control command to the plant;
Calculation means for calculating an alarm generation interval distribution and an alarm corresponding time distribution in the plant based on an alarm occurrence time and a return time included in the history data;
A load evaluation apparatus for plant operation, comprising: an evaluation means for obtaining a treatment time distribution for an alarm from the alarm occurrence interval distribution and the alarm response time distribution, and calculating a load of the plant operation from the alarm occurrence interval distribution and the treatment time distribution. .   The evaluation means calculates a value indicating an operation load by an operator in the plant based on an average alarm occurrence interval that is a parameter of the alarm occurrence interval distribution and an average treatment time that is a parameter of the treatment time distribution. The load evaluation apparatus for plant operation according to claim 1.   The plant operation load evaluation device according to claim 1, wherein the calculation unit of the alarm generation interval distribution includes an aggregation unit that aggregates a plurality of dependent alarms among alarms included in the history data.   The load evaluation of the plant operation according to claim 3, wherein the aggregation means aggregates alarms generated due to the same measurement data among a plurality of alarms generated within a predetermined period. apparatus. The alarm occurrence interval distribution calculating means includes:
Means for calculating an occurrence interval from the occurrence of each alarm to the occurrence of a subsequent alarm for the alarms included in the history data;
5. Means for calculating an average alarm occurrence interval which is a parameter of the alarm occurrence interval distribution by any one method of arithmetic average, logarithmic average, and median for each occurrence interval calculated. The load evaluation apparatus for plant operations according to any one of the above.   The alarm corresponding time distribution calculating means calculates, from alarms corresponding to the occurrence of alarm occurrence, alarms included in the history data that are within a predetermined time period from occurrence to recovery from the alarm. The load evaluation device for plant operation according to any one of claims 1 to 5, comprising means for excluding. The alarm response time distribution calculating means includes:
For the alarms included in the history data, means for calculating the time from the occurrence of each alarm to the return as the alarm response time,
Means for calculating an average alarm response time that is a parameter of the alarm response time distribution by any one of an arithmetic arithmetic mean, a logarithmic average, and a median value for each calculated response time. 7. The load evaluation device for plant operation according to any one of 6 above. A load evaluation program for plant operations,
Obtaining history data indicating a history of alarms in the plant control monitoring system from a plant control monitoring system that receives measurement data from the plant and gives a control command to the plant; and
Calculating an alarm generation interval distribution and an alarm response time distribution in the plant based on an alarm occurrence time and a return time included in the history data;
A plant operation load evaluation program for obtaining a treatment time distribution for an alarm from the alarm occurrence interval distribution and the alarm response time distribution, and calculating a plant operation load from the alarm occurrence interval distribution and the treatment time distribution. .

Images