JP7089438B2 - Plant operation support equipment - Google Patents

Description

The present invention relates to a plant operation support device that supports the operation of a steel plant or a chemical plant, and more particularly to a plant operation support device that indicates optimal plant operation to an operator from a physical model of the plant and past operation data.

In steel plants and chemical plants, it may not be possible to directly measure the quality and condition of products. Therefore, in the past, the operator decided the plant operation such as the processing time and the amount of input materials based on the experience and intuition from the indirect measured values. Therefore, the quality and throughput of the product varied depending on the operator.

On the other hand, a method has been proposed in which the quality and state of products are estimated from physical models such as plant simulations and indirect measured values, and the plant operation is presented based on the results.

As a background technique in this technical field, there is Japanese Patent Application Laid-Open No. 2011-256407 (Patent Document 1). In this publication, the amount of power input required for one charge can be determined without excess or deficiency based on the mass balance and heat balance of the substance entering the arc furnace for steel and the substance leaving the arc furnace for steel. The power input control method of is described.

Further, as a background technique in this technical field, Non-Patent Document 1 describes a physical model of an arc furnace.

Japanese Unexamined Patent Publication No. 2011-256407

Johannes Gerhardt Bekker, Ian Keith Craig, and Petrus Christiaan Pistorius. "Modeling and simulation of an electric arc furnace process." ISIJ international vol.39 (1999) no.1 pp23-32.

The above-mentioned Patent Document 1 describes a power input control method for an arc furnace for steel, which can determine a required input power amount without excess or deficiency. However, in the power input control method in a plant such as the arc furnace for steel described in Patent Document 1, the so-called plant operation support method, uncertainty about plant operation estimated by plant simulation or the like is not taken into consideration.

Therefore, the present invention provides a highly reliable plant operation support device in consideration of uncertainty regarding plant operation estimated by plant simulation or the like.

In order to solve the above problems, the plant operation support device of the present invention uses the physical model of the plant to calculate the estimated value of the optimum plant operation of the plant, and the optimum plant operation estimation means and the optimum plant operation. Estimated values are given as past operation data, an operator model given as a probability distribution representing the characteristics of the operation by the operator from the ideal operation timing calculated when all the internal state of the plant is accurately known , and the internal state of the plant. Create a modified model by modifying based on the estimated error model given as a probability distribution that represents the error from the estimated value of the optimal plant operation for the ideal operation timing calculated when all are accurately known. Means, optimal plant operation correction means for correcting estimates of optimal plann operation using a correction model, and operation data acquisition means for acquiring operation data from the plant control device that controls the plant during plant operation. Online optimal plant operation estimation means that sequentially calculates the estimated value of the optimum plant operation of the plant using the operation data acquired from the plant controller, and optimal plant that displays the estimated value corrected by the optimum plant operation correction means. It has an operation display means.

According to the present invention, it is possible to provide a highly reliable plant operation support device in consideration of uncertainty for plant operation estimated by plant simulation or the like.

It is explanatory drawing explaining the structure of the plant operation support apparatus by this Example. It is explanatory drawing explaining the data flow of the plant operation support apparatus by this Example. It is explanatory drawing explaining the flow of the process from the optimum plant operation estimation means to the correction model creation means by this Example. It is explanatory drawing explaining the structure of the optimum plant operation estimation means by this Example. It is explanatory drawing explaining the image of the estimated value error model and the operator model by this Example. It is explanatory drawing explaining the flow of processing at the time of plant operation by this Example. It is explanatory drawing explaining the structure of the online optimum plant operation estimation means by this Example. It is an image diagram of the input part of the plant operation support device by this embodiment. It is an image diagram of the output part of the plant operation support device by this embodiment. It is explanatory drawing explaining the structure of the plant operation support apparatus by 2nd Embodiment. It is explanatory drawing explaining the data flow of the plant operation support apparatus by 2nd Example. It is explanatory drawing explaining the flow of the process from the optimal plant operation estimation means to the correction model creation means by the 2nd Example.

Hereinafter, examples will be described with reference to the drawings.

FIG. 1 is an explanatory diagram illustrating a configuration of a plant operation support device according to the present embodiment.

The plant operation support device 1 has an arithmetic processing unit 101 and a storage unit 102, and is connected to an input unit 103 and an output unit 104.

Further, the plant operation support device 1 acquires operation data from the plant control device 106, and the plant control device 106 controls the plant 105.

The operation data includes sensor data (measurement data) from sensors installed in the plant, plant operation data of the operator (control command to the plant by the operator (operation history)), amount of raw materials input to the plant, and the like. Includes operating condition data.

Specifically, the storage unit 102 is a storage device such as a hard disk, and has an operation data DB (database) 107.

Past operation data is recorded in the operation data DB 107.

Specifically, the input unit 103 is various input devices such as a keyboard and a mouse, and is used when an operator inputs some information to the arithmetic processing unit 101.

Specifically, the output unit 104 is an output device such as a display, and the process and result of processing by the arithmetic processing unit 101 are output. Alternatively, information for interactive processing between the plant operation support device 1 and the operator is displayed.

Specifically, the arithmetic processing unit 101 is a CPU (Central Processing Unit) and executes information processing. A plant control device 106 that controls the plant 105 is connected to the arithmetic processing unit 101.

The arithmetic processing unit 101 includes an optimum plant operation estimation means 108, a correction model creation means 109, an operation data acquisition means 110, an online optimum plant operation estimation means 111, an optimum plant operation correction means 112, and an optimum plant operation display means 113. Each of these is a program, and for convenience of explanation, each of these functions is defined and described as a means.

FIG. 2 is an explanatory diagram illustrating a data flow of the plant operation support device 1 according to the present embodiment.

The optimum plant operation estimation means 108 estimates the optimum plant operation in the past operation mainly by using the operation condition data among the past operation data recorded in the operation data DB 107, and outputs it as the optimum plant operation estimation value. ..

The modified model creating means 109 mainly uses the operator's plant operation data, the operator model, and the estimation among the past operation data recorded in the operation data DB 107 for the optimum plant operation estimation value estimated by the optimum plant operation estimation means 108. Modify using the value error model and create a modified model.

The operation data acquisition unit 110 acquires operation data from the plant control device 106. That is, the sensor data acquired from the sensor installed in the plant 105, the plant operation data input by the operator to the plant control device 106, and the operation condition data such as the amount of the raw material input to the plant 105 are acquired. ..

Then, these sensor data, plant operation data, and operation condition data are passed to the online optimum plant operation estimation means 111 and the operation data DB 107 as operation data.

The online optimum plant operation estimation means 111 acquires operation data from the operation data acquisition means 110, estimates the optimum plant operation, and outputs it as an online optimum plant operation estimation value.

The optimum plant operation correction means 112 corrects the online optimum plant operation estimation value output by the online optimum plant operation estimation means 111 in real time by using the correction model output by the correction model creation means 109, and outputs it as the optimum plant operation. ..

The optimum plant operation display means 113 presents the optimum plant operation output by the optimum plant operation correction means 112 to the operator via the output unit 104 (see FIG. 1).

FIG. 3 is an explanatory diagram illustrating a flow of processing from the optimum plant operation estimation means to the correction model creation means according to the present embodiment.

What kind of operation was optimal for the optimum plant operation estimation means 108, mainly using the operation condition data among the past operation data recorded in the operation data DB 107 (optimal plant operation in the past operation)? Is estimated (S301).

Next, the modified model creating means 109 uses the optimum plant operation estimation value estimated by the optimum plant operation estimation means 108 and the past operation data (mainly the plant operation data) recorded in the operation data DB 107. , Create a modified model (S302).

FIG. 4 is an explanatory diagram illustrating the configuration of the optimum plant operation estimation means according to the present embodiment.

As shown in FIG. 4, the optimum plant operation estimation means 108 estimates the internal state of the plant (for example, the degree of melting of iron inside the arc furnace) using the operation data and the physical model of the plant, and the value (for example). The optimum plant operation is estimated using the estimated value (predicted value) of the internal state of the plant, and what kind of operation is optimal is estimated as the optimum plant operation estimated value (virtual plant operation estimated value). ..

FIG. 5 is an explanatory diagram illustrating an image of an estimated value error model and an operator model according to the present embodiment.

The image of the estimated value error model and the operator model will be described with reference to FIG. In FIG. 5, the horizontal axis shows time and the vertical axis shows probability. For example, the operator model represents the deviation or variation of the operation by the operator from the ideal operation timing calculated when all the internal states of the plant are accurately known, and the estimated value error model is the plant. The internal state of is the error of the optimal plant operation estimate for the ideal (true optimal) operation timing calculated when everything is known accurately. Then, using the operator model and the estimated value error model, a modified model is created from the optimum plant operation estimated value and the past operation data.

For example, let t ₀ be the ideal operation timing in the past operation, t _op be the operation timing by the operator, and t _est be the optimum plant operation estimation value.

At this time, the operator model can be given as a probability distribution P _op (t _op -t ₀ | θ _op ) of t _op -t ₀ . Note that θ _op is a parameter that defines the probability distribution P _op . Further, the estimated value error model can be given as a probability distribution _Pest (t _est -t ₀ | θ _est ) of t _est -t ₀ . Note that θ _est is a parameter that defines the probability distribution P _est .

Since t ₀ is an unknown value, instead, the probability distribution P _diff (t _op -t _est | θ) of the difference between the operator's operation timing t _op and the optimal plant operation estimate t _est that can be obtained t _op -t _est Consider _est , θ _op ), and estimate the parameters θ _op and θ _est of the probability distribution from t _op -t _est using maximum likelihood estimation.

Then, using a probability distribution that represents the uncertainty of the optimal plant operation estimate according to _{Pest (t est -t 0} | θ _est ), for example, the appropriate correction value Δt _est of t _est is calculated. Next, create a modified model that takes t _est as an input and outputs t _est + Δt _est .

FIG. 6 is an explanatory diagram illustrating a flow of processing during plant operation according to the present embodiment.

FIG. 6 shows the flow of processing at the time of plant operation after the modified model is created.

First, the operation data acquisition unit 110 acquires operation data from the plant control device 106 (S501).

Next, the online optimum plant operation estimation means 111 calculates the optimum plant operation based on the operation data acquired by the operation data acquisition means 110, and outputs the online optimum plant operation estimation value (S502).

Next, the optimum plant operation correction means 112 corrects the online optimum plant operation estimation value based on the correction model created in advance by the correction model creation means 109, and outputs the corrected value as the optimum plant operation (S503). ).

Finally, the optimum plant operation display means 113 presents the optimum plant operation to the operator via the output unit 104 (S504).

FIG. 7 is an explanatory diagram illustrating the configuration of the online optimum plant operation estimation means according to the present embodiment.

As shown in FIG. 7, the online optimum plant operation estimation means 111 estimates the estimated value (predicted value) of the current internal state of the plant based on the operation data up to the present by using the physical model of the plant, and presently. Based on the estimated value (predicted value) of the internal state of the plant, the optimum plant operation is estimated and output as the online optimum plant operation estimated value.

Here, the plant operation support device in the present embodiment that supports the operation of the arc furnace (plant) will be further described.

In an arc furnace that melts iron scraps and uses them as materials, iron scraps are recharged twice or three times at the timing when the melting is completed.

If the reloading is too early, additional work will be required to stuff the unmelted iron scraps. On the other hand, if the re-loading is too late, the inside of the arc furnace will be exposed and the arc furnace will be consumed. It also consumes extra power. In order to optimize the timing of recharge, it is necessary to know the melting state (solubility) of iron scraps, but it is difficult to directly measure this melting degree because the temperature inside the arc furnace is high.

In the past, from the viewpoint of work efficiency, re-loading was performed at the timing when the operator thought that the solution was sufficiently dissolved. Therefore, the arc furnace was heated excessively and used more electric power than necessary.

Therefore, in this embodiment, a plant operation support device that presents the optimum timing of iron scrap recharge (plant operation) to the operator is shown.

In this embodiment, first, the optimum plant operation estimation means 108 uses the past operating condition data such as the amount of iron scrap input to obtain, for example, a physical model of an arc furnace as described in Non-Patent Document 1. It is used to estimate the amount of unmelted iron scrap and melted iron scrap inside the arc furnace, and output the optimum timing of iron waste recharge in the past operation as the optimum plant operation estimate.

Next, the modified model creating means 109 uses the operator model and the estimated value error model to determine the optimum iron scrap reloading timing output by the optimum plant operation estimation means 108 and the control command to the plant by the operator (the operator gives a control command to the plant. A modified model is created using the operation data of the operation history), that is, the operation data at the timing when the operator actually recharges the iron scrap into the arc furnace.

For example, the optimal re-loading timing of iron scrap output by the optimum plant operation estimation means 108 is t _op , and the timing when the operator actually re-loads the iron scrap into the arc furnace is t _op , which is ideal. The timing of charging is t ₀ .

As the operator model, a model as shown in (Equation 1) can be considered.

Here, Δt is an extra time commonly taken by all operators to sufficiently dissolve iron scraps, and w is a random variable representing operator variation, for example, following a lognormal distribution.

On the other hand, as the estimated value error model, a model like (Equation 2) can be considered.

Here, v is a random variable representing the uncertainty of estimation of the optimum iron scrap recharge timing output by the optimum plant operation estimation means 108 for the ideal recharge timing, and follows, for example, a normal distribution. ..

(Equation 3) is derived from (Equation 1) and (Equation 2).

From the data related to (Equation 3) and t _op , t _est , the parameters of the probability distribution of Δt, w, v are estimated by using maximum likelihood estimation or the like.

Then, the uncertainty of the optimum plant operation estimation value (a random variable representing the uncertainty of the estimation of the optimum iron scrap recharge timing with respect to the ideal recharge timing), and the online optimum plant operation estimation. Using the estimated standard deviation σ _v of v, which is also the uncertainty of the value, we output a modified model such that t _est is t _est + σ _v .

At the time of plant operation, the operation data acquisition means 110 obtains sensor data such as power consumption, measured values such as component ratios of carbon and silicon in molten iron, and operation condition data such as input scrap amount in real time from the plant control device 106. And pass it to the online optimal plant operation estimation means 111.

Next, the online optimum plant operation estimation means 111 predicts the amount of unmelted iron scrap and melted iron based on the operation data, estimates the optimum iron waste recharge timing based on the prediction, and estimates the online optimum plant operation. The operation estimate is calculated and output.

Next, the optimum plant operation correction means 112 corrects the timing of iron scrap reloading, which is an online optimum plant operation estimation value, using the correction model created in advance by the correction model creation means 109, and optimizes the value. Output as a plant operation.

Finally, the optimal plant operation display means 113 presents the optimal plant operation to the operator.

FIG. 8 is an image diagram of an input unit of the plant operation support device according to the present embodiment.

FIG. 8 shows an input unit (input screen image) for inputting a physical model, an operator model, and an estimated value error model of the plant into the plant operation support device in this embodiment.

That is, the physical model of the plant is given in the form of an execution program of the plant simulation. Further, the operator model and the estimation error model can be selected from a plurality of probability distribution candidates displayed in the pull-down menu and the like. In this embodiment, it is shown that the lognormal distribution + offset is selected as the operator model and the normal distribution is selected as the estimation error model.

FIG. 9 is an image diagram of an output unit of the plant operation support device according to the present embodiment.

FIG. 9 shows an output unit (output) in which the next charge amount of iron scrap (next charge amount) and the recharge timing (time until recharge) of the plant operation support device in this embodiment are output. Screen image) is shown.

That is, in this embodiment, it is shown that 20.2 tons is presented as the next charge amount and 00:20:30 is presented as the time until recharge.

As described above, in this embodiment, the uncertainty of the optimum plant operation estimated value estimated using the physical model of the plant is taken into consideration, and appropriate correction (correction) is made to make an error in the physical model of the plant. Even if there is, it is possible to present the optimum plant operation in consideration of an appropriate safety factor.

FIG. 10 is an explanatory diagram illustrating the configuration of the plant operation support device according to the second embodiment.
The configurations and functions already shown in FIG. 1 with the same reference numerals will be omitted. In this embodiment, an operation evaluation means 801 for evaluating a past operation is added to the arithmetic processing unit 101 based on the operation data.

FIG. 11 is an explanatory diagram illustrating a data flow of the plant operation support device according to the second embodiment. The description of the configurations and functions already shown in FIG. 2 with the same reference numerals will be omitted.

The operation evaluation means 801 evaluates the quality of the operation of a plurality of past operation data stored in the operation data DB 107 from the viewpoint of, for example, the throughput indicating the amount of work that can be processed per fixed time and the amount of energy consumption. Output as an operation evaluation.

The modified model creating means 109 includes the optimum plant operation estimation value output from the optimum plant operation estimation means 108, the operator's plant operation data (control command (operation history) for the plant by the operator) included in the operation data, and the operation evaluation. Using the operation evaluation output from the means 801 and the operator model and the estimated value error model, a modified model for modifying the optimum plant operation estimated value is created.

FIG. 12 is an explanatory diagram illustrating a flow of processing from the optimum plant operation estimation means to the correction model creation means according to the second embodiment.

The operation data evaluation means 801 evaluates the past operation data stored in the operation data DB 107 from the viewpoint of throughput, energy consumption, and the like (S1001).

Next, a modified model is constructed from the optimum plant operation estimation value estimated by the optimum plant operation estimation means 108, the operation evaluation output from the operation evaluation means 801 and the past operation data stored in the operation data DB 107 (S1002). ).

Note that S301 is the same as the process shown in FIG.

Specifically, when the operator model showing the deviation or variation of the operation by the operator from the ideal operation timing calculated when all the state quantities of the plant are accurately known, and when all the state quantities of the plant are accurately known. A modified model is created from the optimal plant operation estimates and past operation data using an estimated error model that represents the error of the optimal plant operation estimates for the calculated ideal operation timing.

For example, let t ₀ be the ideal operation timing in the past operation, t _op be the operation timing by the operator, and t _est be the optimum plant operation estimation value.

At this time, for example, with respect to the operator model, it can be given as a probability distribution P _op (t _op -t ₀ | θ _op ) of t _op -t ₀ . Note that θ _op is a parameter that defines the probability distribution P _op .

Further, the estimated value error model can also be given as the probability distribution _Pest (t _est -t ₀ | θ _est ) of t _est -t ₀ . Note that θ _est is a parameter that defines the probability distribution P _est .

Since t ₀ is unknown, instead, the probability distribution P _diff (t _op -t _est | θ _est , θ _op ) of the operator's operation timing t _op and the optimum plant estimation value t _est difference t _op -t _est that can be obtained is used. Consider, and estimate the parameters θ _op and θ _est of the probability distribution from t _op -t _est using maximum likelihood estimation. At this time, data with good operation evaluation are estimated with a large weight, and data with poor operation evaluation are given a small weight.

For example, in normal maximum likelihood estimation, θ _op and θ _est are adjusted so as to maximize the log-likelihood lh represented by (Equation 4).

Note that t ⁱ _op and t ⁱ _est represent data related to the operation timing by the i-th operator and the optimum plant operation estimated value, respectively. In addition, n represents the number of data to be used.

On the other hand, in the maximum likelihood estimation in this embodiment, the weighted log-likelihood lhw represented by (Equation 5) is maximized.

Here, ω _i is the weight determined from the operation evaluation of the i-th operation data.

Then, using a probability distribution that represents the uncertainty of the optimal plant operation estimate according to _Pest (t _est -t ₀ | θ _est ) using the adjusted θ _op , θ _est , for example, an appropriate modification of t _est . Value Δt _est Calculate.

Then, a modified model is created in which t _est is input and t _est + Δt _est is output.
Also in this embodiment, a plant operation support device that supports the operation of the arc furnace is considered. The configurations and functions already shown in the first embodiment having the same reference numerals will be omitted.

First, the operation evaluation means 801 evaluates the quality of the operation based on the operation data stored in the operation data DB 107, and outputs the evaluation as the operation evaluation. The evaluation items of the operation include, for example, the power consumption with respect to the amount of iron waste input and the time until the iron waste is melted.

Next, the modified model creating means 109 outputs the optimum iron scrap re-loading timing output by the optimum plant operation estimation means 108, and the data of the timing actually re-loaded by the operator included in the operation data, the operation. A modified model is created using the operation evaluation, the operator model, and the estimated value error model output from the evaluation means 801.

For example, as in the example of the first embodiment to the arc furnace, the optimum timing for re-loading iron scrap by the plant operation estimation means 108 is set, and the timing when the operator actually re- _loads is t. If _op , the ideal reloading timing is set to t ₀ , and a similar operator model and estimated value error model are considered, (Equation 3) can be derived in the same way.

In this embodiment, the parameter of the lognormal distribution according to v is θ _v ∈ R ² , the parameter of the normal distribution according to w is θ _w ∈ R ² , and the weight of each data calculated from the operation evaluation is ω _i . Then, Δt, θ _v , and θ _w are calculated so as to maximize the weighted logarithmic likelihood lh _w represented by (Equation 6).

Then, using the standard deviation σ _v of the normal distribution according to the estimated v, a modified model is output such that _est is _stat + σ _v .

By doing so, also in this embodiment, the physical model of the plant is modified (corrected) appropriately in consideration of the uncertainty of the optimum plant operation estimated value estimated using the physical model of the plant. Even if there is an error, it is possible to present the optimum plant operation in consideration of an appropriate safety factor.

From these examples, it is possible to present the optimum plant operation that does not depend on the accuracy of the physical model of the plant used in the simulation or the type of measurement data.

Then, the operator can perform the optimum plant operation presented to the operator without considering the empirical factor of safety. Therefore, the operator reduces variations in product quality and throughput.

Therefore, this embodiment can provide a plant operation support device that presents a more optimum plant operation to an operator in consideration of the uncertainty of the optimum plant operation estimated by a plant simulation or the like.

The present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above-described embodiment has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to the one including all the described configurations.

1 Plant operation support device 101 Arithmetic processing unit 102 Storage unit 103 Input unit 104 Output unit 105 Plant 106 Plant control device

Claims (6)

An optimal plant operation estimation means that calculates an estimated value of the optimum plant operation of the plant using the physical model of the plant.
An operator model in which the estimated value of the optimum plant operation is given as a probability distribution representing the characteristics of the operation by the operator from the past operation data and the ideal operation timing calculated when all the internal states of the plant are accurately known . And, the modified model is modified based on the estimated error model given as a probability distribution that represents the error from the estimated value of the optimum plant operation for the ideal operation timing calculated when all the internal states of the plant are accurately known. How to create a modified model and
Optimal plant operation correction means for correcting the estimated value of the optimum plant operation using the modification model, and
An operation data acquisition means for acquiring operation data from a plant control device that controls the plant when the plant is in operation.
An online optimum plant operation estimation means that sequentially calculates an estimated value of the optimum plant operation of the plant using the operation data acquired from the plant control device.
The optimum plant operation display means for displaying the estimated value corrected by the optimum plant operation correction means, and the optimum plant operation display means.
A plant operation support device characterized by having. In the plant operation support device according to claim 1,
A plant operation support device characterized by having an operation evaluation means for evaluating past operations based on the operation data. In the plant operation support device according to claim 1,
A plant operation support device characterized by having an input unit for inputting a physical model of a plant, an operator model, and an estimated value error model. In the plant operation support device according to claim 1,
A plant operation support device characterized by having an output unit that outputs the next charge amount and the time until recharge. In the plant operation support device according to claim 1,
A plant operation support device characterized by having a database of operation data in which operation data is accumulated. In the plant operation support device according to claim 5,
The operation data is a plant operation support device including sensor data from a sensor installed in the plant, plant operation data of an operator, and operation condition data such as the amount of raw materials input to the plant.

Images