JP2011107760A - Device of detecting plant abnormality - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device of detecting plant abnormality and a method of detecting a plant abnormality, for preventing the occurrence of abnormality from being erroneously detected. <P>SOLUTION: The device for detecting an abnormality is configured to detect abnormality which occurs in a plant by comparing a simulation result using a model in a normal time of the plant with the actual state of the plant, and provided with: a simulation trial means for trying the simulation of the plant after applying a normal time factor of the generation of an error in the simulation to the model; and a normality determination means for determining that any abnormality has not been generated in the plant, when the simulation result of the simulation trial means is matched with the actual state of the plant. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an abnormality detection apparatus and an abnormality detection method for detecting an abnormality occurring in a plant by comparing a simulation result using a normal model of the plant with an actual state in the plant.

  In the overtime work of raw materials such as petroleum refining, petrochemicals, steel, and papermaking, manufacturing plants are becoming larger and more complex. When an abnormality occurs in such a large and complicated plant, the measured value indicating the state of the plant is affected over a wide range of the plant, and a large number of alarms are simultaneously issued. Therefore, it is difficult to quickly and accurately grasp the location of an abnormality and the content of the abnormality based solely on the experience of the operator.

  Therefore, in recent years, when an abnormality occurs in a plant, a method for detecting the occurrence of an abnormality or an abnormal part using a plant model has been developed. The plant model includes a transfer function model, a polynomial model, a physical model, and the like, and all models are constructed through model parameters for simulating the output of an actual plant. The model parameters are adjusted so that the model output matches the plant output under normal conditions.

  In a system using such a plant model, the actual plant input is also given to the plant model, the plant output (output from the actual plant) and the model output (output from the plant model) are compared, and the two differ It is determined that a plant abnormality has occurred.

JP 2005-332360 A Japanese Patent Laid-Open No. 08-189846

  In the above system, in order to accurately determine the presence or absence of an abnormality, a plant model that accurately simulates the plant output is required. However, approximation of the model is unavoidable when constructing the plant model, and an error in the model output due to the modeling error occurs. Moreover, since the limitation of the system to be modeled is unavoidable, an error occurs between the plant output and the model output due to external factors outside the system (for example, outside air temperature). Furthermore, a phenomenon occurs in which the plant output deviates from the model output due to the secular change of the plant, such as piping contamination and catalyst deterioration, and the error gradually increases.

  For this reason, in a system that detects an abnormality based on the error between the plant output and the model output, the above error occurs even though no abnormality actually occurs in the plant, and the occurrence of the abnormality is erroneously detected. There is a risk of being.

  An object of the present invention is to provide a plant abnormality detection device and a plant abnormality detection method that can prevent the occurrence of abnormality from being erroneously detected.

An abnormality detection apparatus according to the present invention is an abnormality detection apparatus that detects an abnormality occurring in a plant by comparing a simulation result using a model of the plant in a normal state with an actual state in the plant. In consideration of the normal factor causing the occurrence of the simulation, the simulation trial means for trying the simulation of the plant, and the simulation results by the simulation trial means coincide with the actual state in the plant, Normality determining means for determining that no abnormality has occurred in the plant.
According to this abnormality detection device, a simulation of the plant is tried after taking into account the normal factors, and if the simulation result matches the actual state of the plant, there is no abnormality in the plant. Since the determination is made, it is possible to prevent erroneous detection of occurrence of abnormality.

  An abnormality factor specifying unit that specifies a factor of abnormality of the plant by trying the simulation of the plant after taking into account the factor at the time of abnormality in the model, the abnormality factor specifying unit is based on the normality determining unit When the determination is negative, the factor may be specified.

  The abnormality factor specifying means may try the simulation of the plant after adding the factor at the time of abnormality and the factor at the time of normality to the model.

  The model at the normal time may be created by matching a simulation result using the model with an actual state in the plant.

The abnormality detection method of the present invention is an abnormality detection method for detecting an abnormality occurring in a plant by comparing a simulation result using a normal model of the plant with an actual state in the plant. In the case where the simulation result of the step of trying the simulation of the plant and the step of trying the simulation coincides with the actual state in the plant after taking into account the normal factor causing the error in the simulation in the model And determining that no abnormality has occurred in the plant.
According to this abnormality detection method, a simulation of the plant is tried after taking into account the normal factors in the model, and if the simulation result matches the actual state of the plant, there is no abnormality in the plant. Since the determination is made, it is possible to prevent erroneous detection of occurrence of abnormality.

  According to the abnormality detection device of the present invention, a plant simulation is tried after taking into account the normal factor, and if the simulation result matches the actual state of the plant, an abnormality has occurred in the plant. Therefore, it is possible to prevent erroneous detection of occurrence of abnormality.

The block diagram which shows the structure of the plant abnormality detection apparatus of one Embodiment. The flowchart which shows operation | movement of a plant abnormality detection apparatus. The block diagram which shows the structure regarding the parameter adjustment of the plant normal model in a plant abnormality detection apparatus. The flowchart which shows the procedure of the parameter adjustment of the plant normal model in a plant abnormality detection apparatus.

  Hereinafter, an embodiment of a plant abnormality detection apparatus according to the present invention will be described.

  FIG. 1 is a block diagram showing the configuration of the plant abnormality detection device 4 of the present embodiment.

  As shown in FIG. 1, the plant 1 is controlled by a plant control device 2 and monitored by a control monitoring device 3 via the plant control device 2. As the plant control device 2, for example, a distributed control device in which the plant 1 is distributedly controlled via a plurality of field controllers is used.

  The operator issues a control command to the plant control device 2 via the control monitoring device 3. The plant control device 2 executes calculation according to the control law based on the plant output from the plant 1 and the control command received via the plant control device 2, and outputs the calculation result to the plant 1 as a plant input.

  In the plant 1, each operation end is operated by a plant input, and a plant output from each installed measuring instrument is fed back to the plant control device 2. The fed back plant output is used for feedback calculation, alarm calculation, and the like in the plant control apparatus 2. The plant output, alarm information, and the like are displayed as plant information on the control monitoring device 3, and the operator monitors the plant information displayed on the control monitoring device 3 to operate the plant.

  As shown in FIG. 1, the plant abnormality detection device 4 includes an input unit 41, an abnormality detection unit 42, an abnormality diagnosis unit 43, and a display unit 44.

  In addition, a plant normal model 51, a modeling error / external factor model 52, and a plant failure model 53 are mounted on the abnormality detection device 4. As these models, a transfer function model, a polynomial model, a physical model, or the like can be used.

  The input unit 41 of the plant abnormality detection device 4 acquires a plant input and a plant output from the plant control device 2.

  The plant normal model 51 receives a plant input via the input unit 41 and calculates a model normal output that simulates the plant output of the plant 1 at the normal time.

  The modeling error / external factor model 52 receives the plant input via the input unit 41 and calculates the sensitivity of each modeling error and external factor with respect to the plant output of the plant 1 at the normal time.

  Here, the modeling error and the external factor error are errors that the model normal output of the plant normal model 51 has, and the modeling error is an error caused by the approximation of the model in the plant normal model 51, and dirt and catalyst on the piping. This includes errors caused by secular changes of the plant 1 such as deterioration of the plant. An external factor error is an error caused by an external factor outside the system that is not included in the system to be modeled. The modeling error / external factor model 52 defines a range of modeling error or an external factor error that occurs in a normal state. If an output error described later is within the range, the occurrence of abnormality in the plant 1 is denied. Is done.

  The abnormality detection unit 42 receives the plant input via the input unit 41 and the model normal output from the plant normal model 51, respectively, and calculates an output error between the two signals. The output error is solved by the modeling error / external factor sensitivity obtained from the modeling error / external factor model 52, and it is determined whether the output error can be explained by the modeling error or the external factor.

  The plant failure model 53 receives a plant input via the input unit 41 and calculates the sensitivity of each failure factor with respect to the plant output.

  The abnormality diagnosing unit 43 uses the output error obtained from the abnormality detecting unit 42 as a modeling error / modeling error / external factor sensitivity obtained from the external factor model 52 and a fault factor sensitivity obtained from the plant fault model 53. The location where an abnormality has occurred is diagnosed by extracting the cause of the failure that can be solved using and extracting the output error well.

  The display unit 44 receives and displays the abnormality detection result calculated by the abnormality detection unit 42 and the abnormality diagnosis result calculated by the abnormality diagnosis unit 43, thereby notifying the operator of these results.

  FIG. 2 is a flowchart showing the operation of the plant abnormality detection device 4.

  In step S <b> 1 of FIG. 2, a plant input and a plant output are acquired via the input unit 41. The acquired plant output is given to the abnormality detection unit 42.

  Next, in step S <b> 2, a plant input is given to the plant normal model 51, and the model normal output is calculated in the plant normal model 51.

  Next, in step S3, the abnormality detection unit 42 compares the plant output acquired in step S1 with the model normal output obtained from the plant normal model 51, and calculates an output error between the two signals.

  Next, in step S4, it is determined whether or not the output error calculated in step S3 is smaller than a predetermined threshold value. If the determination is affirmative, the processing is terminated as being normal, and if the determination is negative. Proceed to step S5.

  In step S5, the plant input acquired in step S1 is given to the modeling error / external factor model 52, and the modeling error / external factor model 52 calculates the sensitivity of each modeling error / external factor with respect to the plant output. .

  Next, in step S6, the modeling error and external factor sensitivity obtained in step S5 are used to solve the modeling error and external factor that minimize the residual of the output error obtained in step S3 by the least square method. .

  Next, in step S7, it is determined whether or not the output error can be explained by a modeling error / external factor. Specifically, it is determined whether or not the output error residual (difference) minimized in step S6 is smaller than a predetermined threshold. If this determination is affirmed, the process is terminated assuming that there is no abnormality, and if the determination is denied, it is determined that there is an abnormality and the process proceeds to step S8.

  In step S8, the plant input acquired in step S11 is given to the plant failure model 53, and the plant failure model 53 calculates the sensitivity of the plant output for each failure factor.

  Next, in step S9, one of the failure factor sensitivities (failure factor i) obtained in step S8 is selected, and the failure factor for minimizing the residual of the output error for the failure factor i by the least square method. Solve.

  Next, in step S10, it is determined whether or not the output error can be explained by the failure factor i. Specifically, it is determined whether or not the output error residual (difference) minimized in step S9 is smaller than a predetermined threshold. If this determination is positive, the process proceeds to step S11, and if the determination is negative, the process skips to step S12.

  In step S11, failure factor i is recorded as an abnormality candidate, and the process proceeds to step S12.

  In step S12, it is determined whether or not the processing in steps S8 to S11 has been completed for all failure factors. If the determination is affirmative, the process proceeds to step S13. If this determination is denied, the process returns to step S8. In this case, the process proceeds to the process (step S8 to step S11) for the next failure factor i + 1.

  In step S13, it is diagnosed that an abnormality due to the failure factor that can minimize the residual of the output error in step S9 among the failure factors recorded as abnormality candidates in step S11, and the process ends. Based on this processing, a diagnostic result indicating an abnormality occurrence location or the like in the plant 1 is displayed on the display unit 44.

  As described above, in the plant abnormality detection device according to the present embodiment, when an abnormality of the plant 1 is detected, it is possible to determine and explain whether or not the output error can be explained by a model that takes into account modeling errors and external factors. In this case, it is determined that there is no abnormality (steps S5 to S7). For this reason, it is possible to effectively prevent erroneous detection of abnormality when the output error is enlarged due to modeling errors or external factors.

  In addition, in the plant abnormality detection device of the present embodiment, when extracting a failure factor, it is determined whether or not an output error can be explained for each failure factor using a model that takes into account modeling errors and external factors. Uses the failure factor as an abnormality candidate (steps S8 to S10). For this reason, it is possible to appropriately diagnose the abnormal part and the degree of influence of the failure factor.

  In this way, by performing a simulation using a model that takes into account the effects of modeling errors and external factors, even if an error occurs between the actual plant value and the calculated value in the normal model, the abnormal Detection and diagnosis can be performed.

  2 is the function of the simulation trial unit, the process of step S7 is the function of the normality determination unit, and the processing of step S8 to step S12 is the function of the abnormality factor identification unit. Equivalent to.

  FIG. 3 is a block diagram showing a configuration relating to parameter adjustment of the plant normal model 51 in the plant abnormality detection device 4. As shown in FIG. 3, the plant abnormality detection device 4 is provided with a parameter adjustment unit 46 as a component for parameter adjustment and a plant adjustment model 54.

  FIG. 4 is a flowchart showing a procedure for parameter adjustment of the plant normal model 51 in the plant abnormality detection device 4.

  In step S <b> 21 of FIG. 4, a plant input and a plant output are acquired via the input unit 41. The acquired plant output is given to the parameter adjustment unit 46.

  Next, in step S22, a plant input is given to the plant normal model 51, and the model normal output is calculated in the plant normal model 51.

  Next, in step S23, the parameter adjustment unit 46 compares the plant output acquired in step S21 with the model normal output obtained from the plant normal model 51, and calculates an output error between the two signals.

  Next, in step S24, the plant input acquired in step S1 is given to the plant adjustment model 54, and the sensitivity of each adjustment parameter with respect to the plant output is calculated in the plant adjustment model 54.

  Next, in step S25, using the adjustment parameter sensitivity obtained in step S24, an adjustment parameter value that minimizes the residual of the output error obtained in step S23 is solved by the least square method.

  Next, in step S26, it is determined whether or not the output error can be appropriately explained by the adjustment parameter. Specifically, it is determined whether the output error residual (difference) minimized in step S25 is smaller than a predetermined threshold. If this determination is affirmed, the process ends. If the determination is negative, the process proceeds to step S27.

  In step S27, the adjustment parameter value obtained in step S25 is set in the plant normal model 51 and presented to the operator via the display device 44, and the process ends.

  Thus, the adjustment parameter of the plant normal model 51 is adjusted so that the model normal output of the plant normal model 51 is matched with the plant output. An appropriate plant normal model 51 can be constructed by executing such an adjustment parameter setting procedure when an apparatus is introduced or when a plant is updated.

  Further, after the introduction of the apparatus, the above procedure is repeated as appropriate to update the plant normal model 51, so that the adjustment parameters can always be optimized in accordance with the time-dependent change of the plant 1. In this case, since the output error between the plant output and the model normal output obtained from the plant normal model 51 is always suppressed, for example, the determination in step S4 in FIG. The processing time when an abnormality is detected can be shortened.

  As described above, according to the abnormality detection device and the abnormality detection method of the present invention, when a simulation of a plant is tried after taking into account the factors of normality, the simulation result matches the actual state in the plant. Therefore, since it is determined that no abnormality has occurred in the plant, erroneous detection of the occurrence of abnormality can be prevented.

  The scope of application of the present invention is not limited to the above embodiment. The present invention is widely applied to an anomaly detection apparatus and an anomaly detection method for detecting an anomaly occurring in a plant by comparing a simulation result using a normal model of the plant with an actual state in the plant. can do.

42 Abnormality detection unit (simulation trial means, normality determination means)
43 Abnormality diagnosis unit (Abnormality factor specifying means)
51 Plant normal model (simulation trial means)
52 Modeling error / external factor model (simulation trial means)
53 Plant failure model (Abnormality factor specifying means)

Claims (5)

In an anomaly detection device that detects an anomaly occurring in a plant by comparing a simulation result using a normal model of the plant with an actual state in the plant,
A simulation trial means for trying a simulation of the plant after taking into account the normal factor causing an error in the simulation,
When the simulation result by the simulation trial unit coincides with the actual state in the plant, normal determination unit that determines that no abnormality has occurred in the plant,
An abnormality detection device comprising: After adding the factor at the time of abnormality to the model, it comprises an abnormality factor identifying means for identifying the factor of the plant abnormality by trying the simulation of the plant,
The abnormality detection device according to claim 1, wherein the abnormality factor identification unit identifies the factor when the determination by the normality determination unit is negative. The abnormality detection device according to claim 2, wherein the abnormality factor specifying unit tries the simulation of the plant in consideration of the factor at the time of abnormality and the factor at the time of normality in the model. The abnormality detection device according to any one of claims 1 to 3, wherein the normal model is created by matching a simulation result using the model with an actual state in the plant. . In an abnormality detection method for detecting an abnormality occurring in a plant by comparing a simulation result using a normal model of the plant and an actual state in the plant,
Computer
Adding a normal factor causing an error in simulation to the model, and then trying to simulate the plant;
If the simulation result by the step of trying the simulation matches the actual state in the plant, determining that no abnormality has occurred in the plant;
The abnormality detection method characterized by performing.

Images