JP6641867B2 - Power consumption prediction method, apparatus and program - Google Patents

Description

  The present invention is a power consumption amount of a power system of a rolling mill in a rolling mill such as a hot rolling mill or a cold rolling mill, and a power consumption prediction method suitable for use in predicting the power consumption of a rolling mill. It relates to an apparatus and a program.

Business establishments that consume a large amount of power, such as steelworks, have contracts with power companies to determine the upper limit of power consumption per predetermined time (for example, 30 minutes or 60 minutes). This upper limit is called the contracted electric energy, and the basic electricity charge (the amount to be paid regardless of the electricity consumption) is determined based on the contracted electric energy. If the power consumption exceeds the currently set contracted power amount, the contracted power amount becomes the maximum consumption at that time for one year thereafter, and the basic fee increases accordingly.
In this way, if the power consumption exceeds the contracted power even for one hour, the basic fee for one year will increase, so the steelworks will always observe the operating status of all factories and predict the power consumption, When it is expected that the contracted amount of electricity will be exceeded, measures will be taken to temporarily suspend production at some factories. Although this is called power limitation, if the prediction accuracy of the power consumption is low, the factory will be stopped uselessly, so the power consumption is predicted with high accuracy and the factory is stopped only for the minimum necessary time. It is desirable.
Among the power consumption of the entire steelworks, the consumption of the rolling mills such as the hot rolling mill and the plate mill is large, and the consumption differs greatly depending on the material to be rolled. Therefore, it is important to accurately predict the power consumption of the rolling mill in order to grasp the power consumption of the entire steelworks.

  As this kind of technology, Patent Literature 1 discloses a method of estimating power consumption in a rolling process. In Patent Literature 1, the rolling power for each slab is predicted based on the rolling amount, and the power used is calculated by integrating the slab to be rolled within a predetermined time and the power required for rolling this slab. Predict.

  Patent Literature 2 discloses a method of estimating the amount of electric power used in a rolling mill. In Patent Literature 2, the material to be rolled in the time zone to be predicted is classified into three categories having a carbon content of 0.05% or less, 0.06 to 0.15%, and 0.16% or more. Calculate the expected rolling amount. In addition, a numerical value of 0 to 1 is determined according to the number of materials to be rolled in the time zone to be predicted and the operation status of each process (rolling process, refinement process, and pickling process) in the time zone to be predicted. Then, using these as inputs, the power consumption in the time zone to be predicted is calculated by the neural network.

JP-A-64-15201 JP-A-6-262223

  However, the method of Patent Document 1 predicts the rolling power for each slab based on the amount of rolling, and does not consider information such as the manufacturing specifications of the material to be rolled. Accuracy is reduced.

  Further, the method of Patent Document 2 predicts power consumption in a predetermined time after aggregating information such as a scheduled rolling amount classified according to a carbon content into a value in a predetermined time unit. The neural network is a non-linear model, but can only consider the non-linearity of the sum of information such as the expected rolling amount and the power consumption. Therefore, in principle, it is not possible to consider the effect of the non-linearity between the information such as the planned rolling amount for each material to be rolled and the power consumption, and the prediction accuracy of the power consumption is reduced. For example, the predicted value of the power consumption is the same for two slabs whose planned rolling amounts are 20 and 40 and two slabs whose planned rolling amounts are both 30.

  The present invention has been made in view of the above points, and an object of the present invention is to make it possible to accurately predict power consumption of a power system of a rolling mill in a rolling mill, and furthermore, power consumption of the rolling mill. And

The gist of the present invention for solving the above-mentioned problems is as follows.
[1] A power consumption prediction method for predicting power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A plurality of types of information including, as explanatory variables, a first power consumption prediction step obtained by a mill motor power consumption prediction model that is a nonlinear model;
For a material to be rolled within a predetermined time, a sum calculation step of calculating the sum of the values for each piece of information and the sum of the power consumption predicted values of the mill motor obtained in the first power consumption prediction step,
The power consumption predicted value of the power system of the rolling mill during the predetermined time is calculated by summing the values for each piece of information obtained in the sum calculation step and the sum of the power consumption predicted values of the mill motor obtained in the sum calculation step. And a second power consumption prediction step of determining the power consumption of the rolling mill using a power consumption prediction model.
[2] The power consumption prediction method according to [1], wherein the mill motor power consumption prediction model is a random forest model.
[3] A power consumption prediction method for predicting power consumption of a rolling mill including a rolling mill and other equipment,
The power consumption prediction value of the power system of the rolling mill for the predetermined time obtained by the power consumption prediction method according to [1] or [2], and the other equipment obtained by the predetermined power consumption prediction method A power consumption prediction method for the rolling mill, wherein the sum of the power consumption prediction values of the respective power systems for the predetermined time is used as the power consumption prediction value of the rolling mill for the predetermined time.
[4] A power consumption prediction device for predicting power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A first power consumption predicting means obtained by using a mill motor power consumption prediction model, which is a nonlinear model, using a plurality of types of information including
A first sum calculating means for calculating a sum of values for each piece of information for a material to be rolled within a predetermined time;
A second sum calculating means for calculating a sum of power consumption predicted values of the mill motor obtained by the first power consumption predicting means, for a material to be rolled within the predetermined time;
The predicted value of the power consumption of the power system of the rolling mill during the predetermined time is calculated by summing up the values for each piece of information obtained by the first summation means and the consumption of the mill motor obtained by the second summation means. A power consumption predicting apparatus, comprising: a second power consumption predicting unit that obtains a power system power consumption prediction model of a rolling mill using a total of the power consumption predicted values as an explanatory variable .
[5] A program for predicting power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A plurality of types of information including, as explanatory variables, a first power consumption prediction process obtained by a mill motor power consumption prediction model that is a nonlinear model;
For a material to be rolled within a predetermined time, a sum calculation process of calculating a sum of values for each piece of information and a sum of predicted power consumption values of the mill motor obtained in the first power consumption prediction process;
The predicted value of the power consumption of the power system of the rolling mill for the predetermined time is the sum of the values for each piece of information obtained in the sum calculation processing, and the sum of the predicted power consumption of the mill motor obtained in the sum calculation processing. And a second power consumption prediction process that is performed using a power system power consumption prediction model of a rolling mill with the following as explanatory variables .

ADVANTAGE OF THE INVENTION According to this invention, the power consumption of the electric power system of a rolling mill in a rolling mill, and also the power consumption of a rolling mill can be predicted with high precision.

FIG. 2 is a diagram illustrating a functional configuration of a power consumption prediction device according to the embodiment. It is a figure which shows the example of a typical equipment arrangement | positioning of a hot rolling mill. FIG. 9 is a characteristic diagram showing a result when a linear model and a non-linear model are applied when obtaining a predicted value of power consumption of a finishing mill motor for each steel material. When estimating the hourly power consumption of the finishing mill motor, a model using the sum of steel material information as an input variable is used, and the power consumption prediction for each steel material is performed using the finishing mill motor power consumption prediction model as in the present invention. FIG. 9 is a characteristic diagram showing a result of a case where a value is obtained and prediction is performed by taking a sum thereof. FIG. 9 is a characteristic diagram illustrating a result of the prediction method according to the present invention in the example and a result of the conventional prediction method.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.
In the present embodiment, a description will be given of a hot rolling mill which consumes much power among rolling mills.
FIG. 2 shows an example of a typical equipment arrangement such as a hot rolling mill. The hot rolling factory is a factory that processes a steel slab called a slab as a material to be rolled into a predetermined width and thickness. Specifically, the slab is heated to a predetermined temperature in the heating furnace 1 and formed in the width direction by the sizing press 2. Next, the slab formed by the sizing press 2 is rolled by a rough rolling mill 3 and a finishing rolling mill 4 to have a predetermined size. Then, the rolled steel sheet is controlled to have a predetermined structure by a water cooling device called ROT 5, and is wound into a coil by a facility called a coiler 6. The hot-rolling plant also has a refining facility 7 for inspecting the surface of the steel sheet. In order to prevent scale generated on the surface of a slab or a steel plate (hereinafter, referred to as a steel material) from cutting into the surface due to processing, the exit side of the heating furnace 1, the entrance side of the rough rolling mill 3, and the finishing mill 4 A device for removing scale called a descaler 8 is installed on the entrance side.
Among the equipment of the hot rolling mill, the power consumption of the finishing mill 4, particularly the power consumption of the mill motor (hereinafter referred to as a finishing mill motor), is the largest, and about 40% of the power consumption of the entire hot rolling mill. Consume.

  FIG. 1 shows a functional configuration of a power consumption estimation device according to the present embodiment. In the present embodiment, an example will be described in which the power consumption of the hot rolling mill described in FIG. 2 for a predetermined time is predicted. Normally, in a hot rolling plant, power is supplied by dividing the power system for each facility. As described in detail below, the power consumption prediction device obtains a predicted value of power consumption for a predetermined time for each of the power systems of the facilities of the hot rolling plant, and sums up the sum of the power consumption values for the predetermined time of the hot rolling plant. Assume a predicted value. The predetermined time may be, for example, a time (for example, 30 minutes or 60 minutes) that defines an upper limit of the amount of power consumption contracted with the power company.

Reference numeral 100 denotes a power consumption prediction unit of the finishing mill, which predicts the power consumption of the power system of the finishing mill 4 (hereinafter, also referred to as a finishing main machine system) for a predetermined time.
The power consumption predicting unit 100 of the finishing mill includes a finishing mill motor power consumption predicting unit 101, an adder 102, and a finishing main machine power consumption predicting unit 103.

The finishing mill motor power consumption estimating unit 101 calculates, for each steel material to be rolled, a predicted power consumption value w _a of the finishing mill motor required for rolling, as shown in equation (1a), as shown in equation (1a). (Hereinafter referred to as steel material information) using a finishing mill motor power consumption prediction model. The finishing mill motor power consumption prediction unit 101 is an example of a first power consumption prediction unit according to the present invention.
The finishing mill motor power consumption prediction model is a non-linear model, and f _a (•) is _a non-linear function representing the relationship between steel material information and the power consumption of the finishing mill motor. As the nonlinear model, for example, a random forest model can be cited, but a neural network or a kernel method may be applied.
Further, in the present embodiment, the steel material information includes a size (a set plate thickness, a set plate width, and the like), a component (for example, a content of carbon, silicon, manganese, and the like), and the presence or absence of controlled rolling. Not something. For example, the weight of the steel material, the strength of the steel material, and the like may be included.
When the finishing mill is composed of a plurality of stands, a prediction model may be constructed for each stand, or a plurality of stands may be collectively regarded as one finishing mill to construct a prediction model.
Predicted power consumption w _{a of} finishing mill motor
= F _a (size, component,..., Controlled rolling presence / absence) (1a)

The adder 102 calculates the sum of the predicted power consumption w _a of the finishing mill motor, which is obtained by the finishing mill motor power consumption predicting unit 101, as shown in Expression (2a), for the steel material rolled within a predetermined time. The adder 102 is an example of the second sum total calculating means in the present invention.
Sum = Σw _a power consumption amount prediction value w _a finishing mill motor ... (2a)

Here, reference numeral 200 denotes an adder, which calculates the sum of the steel material information for the steel material rolled within a predetermined time, as shown in equations (3-1) to (3-m). The adder 200 is an example of the first sum total calculating means in the present invention.
Sum of set plate thickness = Σ set plate thickness ... (3-1)
Sum of setting plate width = Σ setting plate width ... (3-2)
Sum of actual X component = actual X component (3-3)
...
Sum of control rolling presence = Σ control rolling presence ... (3-m)

Finishing main engine system power consumption prediction unit 103, a finishing power consumption prediction value of a predetermined time of the power system of the rolling mill 4 (power consumption amount prediction value of finishing the main machine line) W _A, as shown in equation (4a) , the sum of the steel information obtained by the adder 200, with the sum of the power consumption amount predicted value w _a of the mill motor finish obtained by the adder 102, determined by the main motor system power consumption prediction model finish. The finish main machine power consumption prediction unit 103 is an example of a second power consumption prediction unit according to the present invention.
It should be noted that the power model for predicting the power consumption of the finished main system may be a linear model such as a multiple regression or a nonlinear model such as a neural network. In the present embodiment, g _a (·) is represented by a nonlinear function.
Estimated power consumption value W _{A of the} finishing engine system
= G _a (total of power consumption predicted value w _a of finishing mill motor, total of set plate thickness, total of set plate width, total of actual X component, ..., total of control rolling presence / absence) ... (4a )

Reference numeral 300 denotes a power consumption prediction unit of the rough rolling mill, which predicts power consumption of a power system of the rough rolling mill 3 (hereinafter, also referred to as a rough main machine system) for a predetermined time.
Like the power consumption prediction unit 100 of the finishing mill, the power consumption prediction unit 300 of the rough rolling mill includes a rough mill motor power consumption prediction unit 301, an adder 302, and a rough main machine power consumption prediction unit 303. And

The crude mill motor power consumption prediction section 301, for each steel, the power consumption prediction value w _b of the coarse mill motor required for rolling, as shown in Equation (1b), using a steel information, coarse mill motor power consumption Determined by the quantity prediction model.
The rough mill motor power consumption prediction model is a non-linear model, and f _b (·) is a non-linear function representing the relationship between steel material information and the power consumption of the coarse mill motor. As the nonlinear model, for example, a random forest model can be cited, but a neural network or a kernel method may be applied.
Estimated power consumption w _{b of} coarse mill motor
= F _b (size, component,..., Controlled rolling presence / absence) (1b)

The adder 302, the steel is rolled in a predetermined time, as shown in equation (2b), obtaining the sum of the power consumption amount predicted value w _b of the coarse mill motor determined in crude mill motor power consumption prediction section 301.
Total power consumption amount predicted value w _b of the rough mill motor = Σw _b ... (2b)

Crude main engine system power consumption prediction section 303, the power consumption amount prediction value of a predetermined time of the power system of the rough rolling mill 3 (power consumption amount prediction value of the coarse main engine system) W _B, as shown in equation (4b) , the sum of the steel information obtained by the adder 200, with the sum of the power consumption amount predicted value w _b of the coarse mill motor obtained by the adder 302, determined by the coarse main engine system power consumption prediction model.
The rough main system power consumption prediction model may be a linear model such as a multiple regression or a nonlinear model such as a neural network. In the present embodiment, g _b (·) is represented by a non-linear function.
Predicted power consumption value W _{B of} coarse main engine system
= G _b (rough mill motor total power consumption amount predicted value w _b of the sum of the set plate thickness, the sum of the set plate width, the sum of the actual X-component, ..., controlled rolling whether the sum of) ... (4b )

Reference numeral 400 denotes a heating furnace power consumption prediction unit that predicts the power consumption of the power system of the heating furnace 1 (hereinafter, also referred to as a heating furnace system) for a predetermined time.
The heating furnace power consumption prediction unit 400 includes a heating furnace system power consumption prediction unit 401.
The heating furnace system power consumption prediction unit 401 calculates the power consumption prediction value (power consumption prediction value of the heating furnace system) W _C of the power system of the heating furnace 1 for a predetermined time as shown in Expression (4c). The sum of the steel material information obtained by the adder 200 is used to obtain a heating furnace system power consumption prediction model.
The heating furnace system power consumption prediction model may be a linear model such as a multiple regression or a nonlinear model such as a neural network. In the present embodiment, g _c (·) is represented by a non-linear function.
Predicted power consumption W _{C of} heating furnace system
= G _c (total of set plate thickness, total of set plate width, total of actual X component,..., Total of presence or absence of controlled rolling) (4c)

The other equipment such as the sizing press 2, the ROT 5, the coiler 6, the refining equipment 7, and the desk 8 is also configured with a power consumption prediction unit. The power consumption of the system for a predetermined time is predicted.
For example, reference numeral 500 denotes a power consumption prediction unit for the conditioning system, which predicts the power consumption of a power system including the conditioning equipment 7 and the coiler 6 (hereinafter, also referred to as the conditioning system) for a predetermined time.
The adjustment system power consumption prediction unit 500 includes a adjustment system power consumption prediction unit 501.
The refining system power consumption predicting unit 501 calculates the power consumption predicted value (predicted power consumption value of the refining system) W _M of the power system of the refining facility 7 for a predetermined time as shown in Expression (4m). , Using the sum total of the steel material information obtained by the adder 200, and making a prediction using a refined system power consumption prediction model.
It should be noted that the refined system power consumption prediction model may be a linear model such as a multiple regression or a nonlinear model such as a neural network. In the present embodiment, g _m (·) is represented by a nonlinear function.
Predicted power consumption value W _{M of the} refining system
= G _m (total of set plate thickness, total of set plate width, total of actual X component, ..., total of control rolling presence / absence) ... (4m)
In addition, the power system has a different configuration for each hot rolling plant, and in addition to the above, there are, for example, a rough auxiliary system including the sizing press 7, a cooling facility system including the ROT 5, a desk system including the desk 8 and the like. Similarly to the heating furnace system power consumption prediction unit 400 and the refined system power consumption prediction unit 500, a power consumption prediction unit for each power system can be constructed.

  Reference numeral 600 denotes an adder, which calculates the total sum of the predicted power consumption values obtained by the power consumption prediction units 100, 300, 400,... The sum of the power consumption predicted values obtained for each facility obtained here becomes the power consumption predicted value of the hot rolling plant for a predetermined time.

As described above, when estimating the power consumption of the finishing mill 4 that consumes a large amount of power for a predetermined time, the power consumption of the finishing mill motor is predicted by a non-linear model (finishing mill motor power consumption prediction model) for each steel material. Find the value. Then, for the steel material rolled within a predetermined time, the sum of the steel material information and the sum of the predicted power consumption of the finishing mill motor for each steel material are obtained, and these are used as input variables to determine the predetermined value of the power system of the finishing mill 4. A predicted power consumption value for time is obtained. The same method is used for the rough rolling mill 3.
As described above, the two-stage prediction model is used, and by using the sum of the predicted values of the power consumption of the mill motor for each steel material as an input variable, it is possible to reflect the nonlinear relationship between the steel material information for each steel material and the power consumption of the mill motor. Therefore, the power consumption of the rolling mill for a predetermined time can be predicted with high accuracy.
When the sum of the predicted values of the power consumption of the mill motor for each steel material is not used as an input variable, the information on the steel materials of each steel material is aggregated as a sum and then used for predicting the power consumption of the rolling mill. In this case, if there is a non-linear relationship between the steel material information and the power consumption of the mill motor, the non-linear relationship cannot be reflected in the prediction of the power consumption of the rolling mill. For example, even if the sum of the power consumption of the two steel mill motors with the set thicknesses of 3 mm and 5 mm is different from the sum of the power consumption of the two steel mill motors with both the set thicknesses of 4 mm, Since both the sums are 8 mm, they cannot be distinguished from each other in the prediction of the power consumption of the rolling mill, and the predicted values of the power consumption become the same, resulting in a prediction error.

Further, by using the sum of the predicted values of the power consumption of the mill motor for each steel material as an input variable, the power consumption of the rolling mill for a predetermined time can be predicted with high accuracy also from the following viewpoints.
For example, the finishing main machine system includes the power consumption of the finishing mill motor, and the power consumption of equipment (auxiliary equipment) other than the finishing mill motor included in the finishing main motor system generally correlates with the power consumption of the finishing mill motor. It is thought that it is. For example, the power consumption of a finishing mill motor increases depending on the size and weight of steel, but the amount of cooling water between stands is proportional to the cross-sectional area of the steel, so the power consumption of a pump for supplying cooling water Has a strong correlation with the power consumption of the finishing mill motor. Since the power consumption of other auxiliary machines also increases depending on the size and weight of the steel material, the correlation with the power consumption of the finishing mill motor is high. As described above, when predicting the power consumption of the auxiliary machine, the prediction accuracy is improved by adding the predicted value of the power consumption of the finishing mill motor having a high correlation to the input variable.

In the example of FIG. 1, the sum of the predicted power consumption values w _a of the finishing mill motor for each steel material is used only as an input variable (explanatory variable) of the power consumption prediction model for the finishing main engine system, but from a physical point of view. If it is determined that it is necessary, it may be used as an input variable of a power consumption prediction model of a power system of another facility. The same applies to the sum of the power consumption amount predicted value w _b of the rough mill motor for each steel.
For example the total power consumption amount predicted value w _a steel each of the finishing mill motor, when used as input variables of the crude main engine system power consumption prediction model, so that the following formula (4b) '.
Predicted power consumption value W _{B of} coarse main engine system
= G _b ′ (total of predicted power consumption w _a of finishing mill motor, total of predicted power consumption w _b of coarse mill motor, total of set plate thickness, total of set plate width, total of actual X component, .., sum of the presence or absence of controlled rolling) (4b) '
Also, for example, the sum of the predicted power consumption value w _a of the finishing mill motor for each steel material and the sum of the predicted power consumption value w _b of the coarse mill motor for each steel material are used as input variables of the heating furnace system power consumption prediction model. And the following equation (4c) ′.
Predicted power consumption W _{C of} heating furnace system
= G _c ′ (total of predicted power consumption value w _a of finishing mill motor, total of predicted power consumption value w _b of coarse mill motor, total of set plate thickness, total of set plate width, total of actual X component, .., sum of the presence or absence of controlled rolling) (4c) '
However, in a power system that has a low correlation with the power consumption of the finishing mill motor and the coarse mill motor (for example, a refinement system), the sum of the power consumption of the finishing mill motor and the predicted value of the power consumption of the coarse mill motor for each steel material is added to the input variables. But the effect is low.

FIG. 3 shows a result when a linear model and a non-linear model are applied when obtaining a predicted value of power consumption of a finishing mill motor for each steel material.
A prediction model was created from the operation results data of the hot rolling mill for four months (64937 steel materials), and the accuracy was evaluated using the operation results data for two months (17430 steel materials). FIG. 3 is a scatter diagram showing the predicted values and the actual values calculated using the operation result data for two months.
In the finishing mill motor of the finishing mill consisting of 7 stands, when calculating the predicted value of the power consumption of the finishing mill motor for each steel material, FIG. 3 (a) shows the finished mill motor power consumption prediction model as a linear model (multiple regression model). FIG. 3 (b) shows a case in which the finishing mill motor power consumption prediction model is a nonlinear model (random forest model). In FIG. 3, the power consumption of the finishing mill motor is scaled by setting the average power consumption of the hot rolling mill per hour to 100%.
As shown in FIG. 3, since the prediction accuracy differs greatly between the linear model and the non-linear model, it can be confirmed that the power consumption of the finishing mill motor for each steel material should be predicted by the non-linear model.
Note that the multiple regression model and the random forest model are models for predicting the power consumption unit as shown in equations (5a) and (5b). p (·) is a function for outputting the power consumption unit. The multiple regression model is a linear form, and the random forest model is a model composed of 500 decision trees.
Power consumption unit (kWh / t) = p (set plate thickness, set plate width, set plate length, set weight, set tensile strength, actual X component, presence or absence of controlled rolling) (5a)
Finishing mill motor power consumption (kWh) = Power consumption unit (kWh / t) × Set weight (t) (5b)

FIG. 4 shows a case where a model using the total sum of steel material information is used as an input variable when estimating the hourly power consumption of the finishing mill motor, and a case where the finishing mill motor power consumption predicting model is used for each steel material according to the present invention. A result of a case where a predicted value of the power consumption is obtained and the sum is obtained to perform the prediction is shown. Note that the target finishing mill, the scaling in FIG. 4, and the random forest model used as the model for predicting the power consumption of the finishing mill motor are the same as those described in FIG. 3, and the description thereof is omitted here.
FIG. 4 (a) shows that in equation (5a), instead of using the steel material information for each steel material as an input variable, using the one-hour sum of each steel material information as an input variable, the consumption of the steel material rolled in one hour The case where the electric energy is predicted is shown. The model used here was a random forest model composed of 500 decision trees.
FIG. 4B predicts the power consumption for each steel rolled in one hour by using the finishing mill motor power consumption prediction model as in the present invention, using the steel material information shown in the equation (5a) as an input variable. Here is a case where the sum is obtained.
As shown in FIG. 4, by a method of predicting the power consumption amount for each steel, the prediction accuracy R ² of the power consumption of 1 hour finishing mill motor can be improved from 0.9603% to 0.9934%.

FIG. 5 shows the result of the prediction method according to the present invention and the result of the conventional prediction method for the prediction of the hourly power consumption of the finishing main machine system including the seven stands of a hot rolling mill. Note that the target finishing mill, the scaling in FIG. 5, and the random forest model used as the model for predicting the power consumption of the finishing mill motor are as described in FIG. 3, and the description is omitted here.
FIG. 5 (a) shows a conventional prediction method, that is, the power consumption is predicted using the finishing main system power consumption prediction model using the sum of steel material information as an input variable without using the finishing mill motor power consumption prediction model. Here is the case. Referring to FIG. 1, there is a case where only the sum of the steel material information obtained by the adder 200 is input to the finishing main machine system power consumption prediction unit 103.
FIG. 5B shows the power consumption of the finishing main machine system using the prediction method according to the present invention, that is, using the finishing mill motor power consumption prediction model, and using the total sum of steel material information and the total power consumption of the finishing mill motor for each steel material as input variables. 5 shows a case in which power consumption is predicted using a power prediction model.
The finishing main engine system power consumption prediction model is a multiple regression model for obtaining a predicted value of power consumption of the finishing main engine system for a predetermined period of time using the sum of each piece of steel material information shown in equation (5a) as an input variable and operating for four months. Created from actual data (2372).
FIG. 5 is a scatter diagram showing a predicted value and an actual value with respect to the power consumption (638 pieces) of the finishing main machine system every hour for two months. As shown in FIG. 5, by using the prediction method according to the invention, it is possible to improve the prediction accuracy R ² from 0.9642 to 0.9828.

As described above, the present invention has been described with various embodiments. However, the present invention is not limited to these embodiments, and changes and the like can be made within the scope of the present invention.
The power consumption estimation device to which the present invention is applied is realized by a computer device having, for example, a CPU, a ROM, a RAM, and the like. The CPU expands a program stored in the ROM into the RAM and executes the program. The functions of the respective units are realized.
Further, the present invention supplies software (program) for realizing the function as the power consumption estimation device of the present invention to a system or an apparatus via a network or various storage media, and a computer of the system or the apparatus executes the program. It can also be realized by reading and executing.

100: Power consumption prediction unit for finishing mill 101: Finishing mill motor power consumption prediction unit 102: Adder 103: Power consumption prediction unit for finishing main machine 200: Adder 300: Power consumption prediction unit for rough rolling mill 301 : Coarse mill motor power consumption prediction unit 302: Adder 303: Coarse main machine power consumption prediction unit

Claims (5)

A power consumption prediction method for predicting power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A plurality of types of information including, as explanatory variables, a first power consumption prediction step obtained by a mill motor power consumption prediction model that is a nonlinear model;
For a material to be rolled within a predetermined time, a sum calculation step of calculating the sum of the values for each piece of information and the sum of the power consumption predicted values of the mill motor obtained in the first power consumption prediction step,
The power consumption predicted value of the power system of the rolling mill during the predetermined time is calculated by summing the values for each piece of information obtained in the sum calculation step and the sum of the power consumption predicted values of the mill motor obtained in the sum calculation step. And a second power consumption prediction step of determining the power consumption of the rolling mill using a power consumption prediction model.   The power consumption prediction method according to claim 1, wherein the mill motor power consumption prediction model is a random forest model. Rolling mill, a power consumption prediction method for predicting the power consumption of a rolling mill equipped with other equipment,
A power consumption prediction value of the power system of the rolling mill for the predetermined time obtained by the power consumption prediction method according to claim 1 or 2, and an electric power of the other equipment obtained by a predetermined power consumption prediction method. A power consumption prediction method, wherein the sum of the power consumption prediction values of the respective systems for the predetermined time is used as the power consumption prediction value of the rolling mill for the predetermined time. A power consumption prediction device for predicting power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A first power consumption predicting means obtained by using a mill motor power consumption prediction model, which is a nonlinear model, using a plurality of types of information including
A first sum calculating means for calculating a sum of values for each piece of information for a material to be rolled within a predetermined time;
A second sum calculating means for calculating a sum of power consumption predicted values of the mill motor obtained by the first power consumption predicting means, for a material to be rolled within the predetermined time;
The predicted value of the power consumption of the power system of the rolling mill during the predetermined time is calculated by summing up the values for each piece of information obtained by the first summation means and the consumption of the mill motor obtained by the second summation means. A power consumption predicting apparatus, comprising: a second power consumption predicting unit that obtains a power system power consumption prediction model of a rolling mill using a total of the power consumption predicted values as an explanatory variable . A program for predicting the power consumption of a power system of a rolling mill,
For each material to be rolled, the estimated power consumption of the mill motor required for rolling, the thickness after rolling, the width after rolling, the component, the presence or absence of controlled rolling, the weight, and at least one of the strengths A plurality of types of information including, as explanatory variables, a first power consumption prediction process obtained by a mill motor power consumption prediction model that is a nonlinear model;
For a material to be rolled within a predetermined time, a sum calculation process of calculating a sum of values for each piece of information and a sum of predicted power consumption values of the mill motor obtained in the first power consumption prediction process;
The predicted value of the power consumption of the power system of the rolling mill for the predetermined time is the sum of the values for each piece of information obtained in the sum calculation processing, and the sum of the predicted power consumption of the mill motor obtained in the sum calculation processing. And a second power consumption prediction process that is performed using a power system power consumption prediction model of a rolling mill with the following as explanatory variables .

Images