Electricity consumption forecasting method and electricity consumption forecasting device

Abstract

To provide a device for predicting a power use amount, which can accurately predict a power use amount at a plant using an operation plan.SOLUTION: A device 100 for predicting a power use amount comprises: operation plan acquisition means 10 for acquiring an operation plan of a plant regarding operation time; operation achievement acquisition means 20 for acquiring an operation achievement of the plant regarding the operation time within prescribed past time; operation achievement change calculation means 30 for calculating the change degree of the operation achievement with respect to the operation plan within the prescribed past time; operation plan correction means 40 for correcting a future operation plan on the basis of the change level of the operation achievement; and power use amount prediction means 50 for predicting a future power use amount at the plant on the basis of the corrected operation plan.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] This invention relates to a power consumption prediction method and power consumption prediction device that can accurately predict the power consumption of factories such as steel mills using operational plans. [Background technology]

[0002] Factories that use a lot of electricity, such as steel mills, enter into electricity purchase agreements with power companies and purchase electricity from them. Generally, in electricity purchase agreements, the contracted amount of electricity is determined based on the maximum amount of electricity used within a specified time (for example, 30 minutes), and a penalty is imposed if the actual amount of electricity used exceeds the contracted amount. On the other hand, even if the actual amount of electricity used falls significantly below the contracted amount, the amount equivalent to the contracted amount is still paid. Under such contracts, it is economically important for factories to adjust their electricity usage to keep it stable at all times, and to do so, it is necessary to accurately predict the factory's future electricity usage up to a specified time in advance.

[0003] For example, the rolling mill accounts for a large portion of the total electricity consumption of a steel mill, and furthermore, the amount of electricity consumed varies greatly depending on the manufacturing conditions of the rolled material. Therefore, in order to accurately predict future electricity consumption up to a predetermined time, it is important to accurately grasp the factory's operating plan, which contains information about the rolled material to be rolled within that predetermined time, and to predict electricity consumption based on that plan. For example, Patent Documents 1 to 3 propose methods for predicting electricity consumption based on the factory's operating plan.

[0004] Here, if actual operations (actual operating performance) deviate from the plan due to factory equipment troubles, advancement of operating hours, or bottlenecks in upstream and downstream processes of the rolling mill (heating furnace, coil storage area, etc.), the rolling time of the material to be rolled will deviate from the planned rolling time in the operating plan, and the accuracy of electricity consumption prediction will deteriorate significantly. For this reason, when there is a discrepancy between actual operating performance and the operating plan, the operating plan should be revised, as proposed in Patent Documents 1 and 2. However, Patent Document 1 only states that the initial operation plan should be modified by adding the manufacturing progress and the amount of adjustments from the operators, but does not propose any specific modification methods. Furthermore, Patent Document 2 only proposes a method of uniformly adjusting the planned rolling start time for future rolled materials by the difference between the planned rolling start time and the actual rolling start time (for example, if the actual rolling start time is 6 minutes later than the planned rolling start time, the planned rolling start time for all future rolled materials should be delayed by 6 minutes).

[0005] If the actual rolling start time is delayed or advanced by a certain period due to factory equipment trouble or advancement of operating hours, it is acceptable to uniformly adjust the future rolling start time as described in Patent Document 2. However, if the processes before and after the rolling mill are bottlenecks (for example, if the heating time of the material to be rolled in the heating furnace of the hot rolling mill in the preceding process is insufficient and the rolling mill has to wait until the material to be rolled reaches the target temperature, or if the actual rolling start time is delayed because the coil storage area for the completed rolled material (coils) in the subsequent process is full), it is necessary to adjust the operation so that the time interval from the completion time of rolling of one material to the scheduled rolling start time of the next material to be rolled is longer than usual until the bottleneck is resolved (hereinafter referred to as "pitch-down operation"). In such cases, instead of uniformly adjusting the scheduled rolling start time, it is necessary to revise the operation plan taking into account the longer time interval between materials to be rolled.

[0006] As described above, when using the factory's operating plan to predict electricity consumption at factories such as steel mills, the accuracy of the prediction deteriorates if there is a discrepancy between actual operating conditions and the operating plan. To mitigate this deterioration in accuracy, future operating plans can be revised and used for electricity consumption predictions. In doing so, it is important to make appropriate revisions depending on the cause of the discrepancy between actual operating conditions and the operating plan. However, Patent Documents 1 and 2 do not offer any solutions regarding this point. [Prior art documents] [Patent Documents]

[0007] Patent Document 1 International Publication No. 2015 / 178256 Patent Document 2 Japanese Unexamined Patent Application Publication No. 2000-217253 Patent Document 3 Japanese Unexamined Patent Application Publication No. 2017-70134 Summary of the Invention Problems to be Solved by the Invention

[0008] The present invention has been made to solve the problems of the prior art as described above, and an object thereof is to provide a power consumption prediction method and a power consumption prediction device capable of accurately predicting the power consumption of a factory such as a steel mill using an operation plan. Means for Solving the Problems

[0009] To solve the above problems, the present invention includes an operation plan acquisition step of acquiring an operation plan of a factory related to operation time, an operation result acquisition step of acquiring an operation result of the factory related to operation time within a predetermined past time, an operation result change calculation step of calculating the degree of change in the operation result with respect to the operation plan within the predetermined past time, an operation plan correction step of correcting the future operation plan based on the degree of change in the operation result, and a power consumption prediction step of predicting the future power consumption of the factory based on the corrected operation plan. Furthermore, if the degree of change in the operational performance is within a predetermined range, the future operational plan is revised in the operational plan revision step by uniformly using the degree of change in the operational performance; if the degree of change in the operational performance is not within a predetermined range, the future operational plan is revised in the operational plan revision step by assuming that the degree of change in the operational performance is monotonically increasing or decreasing with respect to the operating time. A power consumption prediction method is provided.

[0010] In the present invention, the "operation plan of a factory related to operation time" means an operation plan related to the operation time in the operation plan of the factory, such as a planned start time of operation, a planned end time of operation, and a planned number of operations per predetermined time. For example, when the factory is a rolling mill of a steel mill, examples of the "operation plan of the factory related to operation time" include a planned start time of rolling for each coil of the material to be rolled, a planned end time of rolling for each coil of the material to be rolled, and a planned value of the number of coils of the material to be rolled to be rolled per predetermined time. Furthermore, in the present invention, "the operational performance of the factory related to the operating time" means the operational performance of the factory related to the operating time, such as the actual start time of operation, the actual end time of operation, and the number of operations per predetermined time. For example, if the factory is a rolling mill of a steel mill, examples of "the operational performance of the factory related to the operating time" include the actual start time of rolling for each coil of rolled material, the actual end time of rolling for each coil of rolled material, and the actual number of coils of rolled material rolled per predetermined time. Furthermore, in the present invention, "the degree of change in the actual operation relative to the operation plan" means the difference between the actual operation and the operation plan, or the ratio between the actual operation and the operation plan. For example, if the factory is a rolling mill in a steel mill, examples of "the degree of change in the actual operation relative to the operation plan" include the difference between the actual start time and the planned start time of rolling for each coil of rolled material, the difference between the actual end time and the planned end time of rolling for each coil of rolled material, and the actual value of the number of coils of rolled material rolled per predetermined time / the planned value of the number of coils of rolled material to be rolled per predetermined time. According to the present invention, in the step of calculating changes in operational performance, the degree of change in operational performance compared to the operational plan within a predetermined period in the past is calculated, and in the step of modifying the operational plan, the future operational plan is modified based on the degree of change in operational performance. In other words, the degree of change in operational performance, which may vary depending on the cause of the discrepancy between operational performance and the operational plan, is calculated, and the future operational plan is modified based on this degree of change in operational performance. Therefore, the future operational plan can be appropriately modified, and consequently, the future power consumption of the factory can be accurately predicted based on the modified operational plan.

[0012] This inventionIn this context, "when the degree of change in the operational results is within a predetermined range" means that the degree of change in operational results relative to the operational plan is approximately constant. Specifically, this means that the difference between the operational results and the operational plan is within a predetermined range, or that the ratio of operational results to the operational plan is within a predetermined range. For example, if the factory is a rolling mill in a steel mill, this could mean that the cumulative value of the difference between the actual rolling start time and the planned rolling start time for each coil of rolled material within a predetermined time is within a predetermined range, or that the cumulative value of the difference between the actual rolling end time and the planned rolling end time for each coil of rolled material within a predetermined time is within a predetermined range, or that the actual value of the number of coils of rolled material rolled per predetermined time / the planned value of the number of coils of rolled material to be rolled per predetermined time is within a predetermined range. Also, This invention In this context, "when the degree of change in the operational results is not within a predetermined range" means when the degree of change in operational results relative to the operational plan is not considered to be approximately constant. Specifically, this means when the difference between operational results and the operational plan is not within a predetermined range, or when the ratio of operational results to the operational plan is not within a predetermined range. For example, if the factory is a rolling mill in a steel mill, this could mean that the cumulative value of the difference between the actual rolling start time and the planned rolling start time for each coil of rolled material over a predetermined period of time is not within a predetermined range, or that the actual value of the number of coils of rolled material rolled per predetermined period of time / the planned value of the number of coils of rolled material to be rolled per predetermined period of time is not within a predetermined range. This inventionAccording to the guidelines, if the degree of change in actual operating performance falls within a predetermined range, it is assumed that the discrepancy between actual operating performance and the operating plan is due to equipment trouble at the factory or early start of operations. In the operating plan revision step, the degree of change in actual operating performance is uniformly used to revise the future operating plan. For example, if the factory is a rolling mill in a steelworks, the scheduled start time for rolling each coil of rolled material will be uniformly delayed by the degree of change in actual operating performance. If the degree of change in actual operating performance falls outside a predetermined range, it is assumed that the discrepancy between actual operating performance and the operating plan is due to pitch-down operations. In the operating plan revision step, it is assumed that the degree of change in actual operating performance is monotonically increasing or decreasing with respect to the operating time, and the future operating plan is revised accordingly. For example, if the factory is a rolling mill in a steel mill, assuming that the difference between the actual start time and the scheduled start time for each coil of rolled material increases monotonically with respect to the operating time, adjustments would be made to increase the delay in the scheduled start time as the operating time is later, i.e., as the coil is rolled later in the sequence. therefore, This invention According to this, the method of revising future operational plans will be categorized according to the cause of the discrepancy between actual operating results and the operational plan, making it possible to appropriately revise future operational plans.

[0013] Furthermore, in order to solve the above-mentioned problems, the present invention includes: an operation plan acquisition step of acquiring an operation plan of a factory relating to operating times; an operation performance acquisition step of acquiring the factory's actual operating times relating to operating times within a predetermined period in the past; an operation performance change calculation step of calculating the degree of change in the actual operating times relative to the operation plan within the predetermined period in the past; an operation plan modification step of modifying the future operation plan based on the degree of change in the actual operating times; and a power consumption prediction step of predicting the future power consumption of the factory based on the modified operation plan. In the operation plan revision step, a machine learning-generated learning model is used, which takes the degree of change in past operation performance as input and the degree of revision of the future operation plan as output, to calculate the degree of revision of the future operation plan, and the future operation plan is revised based on the calculated degree of revision of the future operation plan. It is also provided as a method for predicting electricity usage. .

[0014] This inventionIn this context, "the degree of revision of the future operational plan" means the difference between the revised operational plan and the original operational plan, or the ratio between the revised operational plan and the original operational plan. For example, if the factory is a rolling mill in a steel mill, examples include the difference between the revised and original scheduled start times for rolling each coil of rolled material, the difference between the revised and original scheduled end times for rolling each coil of rolled material, and the ratio of the revised number of rolled material coils rolled per predetermined time to the original number of rolled material coils rolled per predetermined time. The discrepancy between actual operations and the planned operations is not necessarily limited to a single cause; it may involve a combination of factors within the scheduled timeframe, such as equipment malfunctions at the factory, earlier start times, or reduced operating hours. This invention According to the report, when revising future operational plans, a machine learning-generated learning model is used, which takes the degree of change in operational performance as input and the degree of revision of the future operational plan as output. Therefore, it is expected that future operational plans can be appropriately revised even when there are multiple causes of discrepancies. Furthermore, when generating a learning model using machine learning, it is sufficient to use a combination of previously acquired operational plans and operational results as training data. Specifically, the degree of change in operational results calculated from operational plans and operational results acquired in the past prior to a certain operational time should be used as input for training data, and the degree of modification of the operational plan (the degree of modification of the operational plan after a certain operational time necessary to match the operational results after a certain operational time) calculated from operational plans and operational results acquired in the past prior to a certain operational time should be used as output for training data.

[0015] To solve the aforementioned problems, the present invention provides: an operation plan acquisition means for acquiring an operation plan of a factory relating to its operating hours; an operation performance acquisition means for acquiring the factory's actual operating hours within a predetermined period in the past; an operation performance change calculation means for calculating the degree of change in the actual operating hours relative to the operation plan within the predetermined period in the past; an operation plan modification means for modifying the future operation plan based on the degree of change in the actual operating hours; and a power consumption prediction means for predicting the future power consumption of the factory based on the modified operation plan. picture , If the degree of change in the operational performance is within a predetermined range, the operational plan modification means modifies the future operational plan by uniformly using the degree of change in the operational performance; if the degree of change in the operational performance is not within a predetermined range, the operational plan modification means modifies the future operational plan by assuming that the degree of change in the operational performance is monotonically increasing or decreasing with respect to the operating time. It is also offered as a power consumption forecasting device.

[0017] Furthermore, in order to solve the above problems, the present invention comprises: operation plan acquisition means for acquiring an operation plan of a factory relating to operating times; operation performance acquisition means for acquiring the actual operating times of the factory relating to operating times within a predetermined period in the past; operation performance change calculation means for calculating the degree of change in the actual operating times relative to the operation plan within the predetermined period in the past; operation plan modification means for modifying the future operation plan based on the degree of change in the actual operating times; and power consumption prediction means for predicting the future power consumption of the factory based on the modified operation plan. The operation plan modification means comprises a learning model generated by machine learning, which takes the degree of change in past operation performance as input and the degree of future modification of the operation plan as output. The operation plan modification means calculates the degree of future modification of the operation plan using the learning model and modifies the future operation plan based on the calculated degree of future modification of the operation plan. It is also provided as a power consumption forecasting device. . [Effects of the Invention]

[0018] According to the present invention, future operational plans can be appropriately revised according to the reasons for discrepancies between actual operating results and operational plans, and consequently, future power consumption of the factory can be accurately predicted based on the revised operational plan. Therefore, by adjusting factory operations and self-generation based on the accurately predicted power consumption so that this power consumption does not exceed the contracted power consumption, the risk of being charged penalties is reduced. Consequently, the manufacturing costs of the factory can be kept to a minimum. [Brief explanation of the drawing]

[0019] [Figure 1] This is a block diagram showing a schematic configuration of a power consumption prediction device according to an embodiment of the present invention. [Figure 2] This flowchart shows the general procedure for the power consumption prediction method performed using the power consumption prediction device shown in Figure 1. [Figure 3]This is a flowchart showing the general procedure of the operation plan modification step ST4 performed by the operation plan modification means 40 of the first embodiment. [Figure 4] This is a timing chart showing an example of an operation plan modification step ST4 executed by the operation plan modification means 40 of the first embodiment. [Figure 5] This is a flowchart showing the general procedure of the operation plan modification step ST4 executed by the operation plan modification means 40 of the second embodiment. [Figure 6] This figure conceptually shows an example of a learning model (input / output items of the learning model) provided by the operation plan modification means 40 of the second embodiment. [Modes for carrying out the invention]

[0020] Hereinafter, embodiments of the present invention (first and second embodiments) will be described with reference to the attached drawings as appropriate, using the case where the factory is a rolling mill of a steelworks as an example.

[0021] <Common matters> First, we will describe the matters that are common to the first and second embodiments. Figure 1 is a block diagram showing a schematic configuration of a power consumption prediction device according to an embodiment of the present invention. Figure 2 is a flowchart showing a schematic procedure for a power consumption prediction method to be performed using the power consumption prediction device shown in Figure 1. As shown in Figure 1, the power consumption forecasting device 100 according to this embodiment includes an operation plan acquisition means 10, an operation performance acquisition means 20, an operation performance change calculation means 30, an operation plan modification means 40, and a power consumption forecasting means 50. The power consumption forecasting device 100 comprises, for example, one or more hardware processors such as a CPU (Central Processing Unit), one or more memories such as RAM (Random Access Memory) and ROM (Read Only Memory), and performs various calculations by executing one or more programs stored in the memories using one or more hardware processors. As a result, the power consumption forecasting device 100 functions as an operation plan acquisition means 10, an operation performance acquisition means 20, an operation performance change calculation means 30, an operation plan modification means 40, and a power consumption forecasting means 50. The power consumption forecasting device 100 may be a PLC (Programmable Logic Controller) or may be implemented using dedicated hardware such as an ASIC (Application Specific Integrated Circuit).

[0022] The operation plan acquisition means 10 executes the operation plan acquisition step ST1 shown in Figure 2. Specifically, the operation plan acquisition means 10 acquires and stores in advance the operation plan of the rolling mill related to the operating time (for example, the scheduled start time of rolling for each coil of rolled material, the scheduled end time of rolling for each coil of rolled material, the planned number of coils of rolled material to be rolled per predetermined time, etc.) from, for example, a computer (not shown) that manages the production of the rolling mill or a process computer (not shown) that controls the operation. The operation plan related to the operating time acquired by this operation plan acquisition means 10 is an operation plan that spans before and after the current time (the time when the power consumption prediction means 50 predicts (performs prediction calculations) the amount of power consumption). In addition to the operation plan related to the operating time described above, the operation plan acquisition means 10 also acquires and stores in advance the operating conditions necessary for predicting electricity consumption by the electricity consumption prediction means 50 described later (for example, the material of each coil of rolled material, the thickness, width, and length before rolling, the thickness, width, and length after rolling, etc.) as part of the operation plan.

[0023] The operational performance acquisition means 20 executes the operational performance acquisition step ST2 shown in Figure 2. Specifically, the operational performance acquisition means 20 sequentially acquires and stores the factory's operational performance related to the operating time within a predetermined time period prior to the current time (for example, the past 30 minutes) (for example, the actual start time of rolling for each coil of rolled material, the actual end time of rolling for each coil of rolled material, the actual value of the number of coils of rolled material rolled per predetermined time, etc.).

[0024] The operational performance change calculation means 30 executes the operational performance change calculation step ST3 shown in Figure 2. Specifically, the operational performance change calculation means 30 calculates the degree of change in the operational performance relative to the operational plan for a predetermined period of time in the past, based on the operational plan acquired by the operational plan acquisition means 10 and the operational performance for a predetermined period of time in the past acquired by the operational performance acquisition means 20 (for example, the difference between the actual rolling start time and the planned rolling start time for each coil of rolled material, the difference between the actual rolling end time and the planned rolling end time for each coil of rolled material, the actual value of the number of coils of rolled material rolled per predetermined time / the planned value of the number of coils of rolled material to be rolled per predetermined time, etc.).

[0025] The operation plan modification means 40 executes the operation plan modification step ST4 shown in Figure 2. Specifically, the operation plan modification means 40 modifies the future operation plan based on the degree of change in the operational performance calculated by the operational performance change calculation means 30. As described later, the specific details of the operation plan modification step ST4 by the operation plan modification means 40 differ between the first embodiment and the second embodiment.

[0026] The power consumption forecasting means 50 performs the power consumption forecasting step ST5 shown in Figure 2. Specifically, the power consumption forecasting means 50 predicts the future power consumption of the rolling mill based on the operation plan modified by the operation plan modification means 40. Furthermore, as a method for predicting electricity consumption based on the revised operating plan, for example, as described in Patent Document 3, it is possible to use a method that involves adding together the predicted electricity consumption value calculated based on a time-series prediction model using actual electricity consumption values ​​for a predetermined period in the past, obtained from past data, and the predicted electricity consumption value calculated based on a multiple regression model using the rolling conditions for each coil of rolled material obtained as part of the operating plan, i.e., the planned rolling time (planned rolling time obtained from the revised operating plan), material, thickness, width, and length before rolling, and thickness, width, and length after rolling. As a method for predicting electricity consumption based on the revised operating plan, it is possible to use conventionally known methods as appropriate, not limited to the method described in Patent Document 3, so a detailed explanation is omitted here.

[0027] <First Embodiment> The first embodiment of the present invention will be described below. Figure 3 is a flowchart showing the general procedure of the operation plan modification step ST4 performed by the operation plan modification means 40 of the first embodiment. As shown in Figure 3, the operation plan modification means 40 of the first embodiment determines whether the degree of change in the operational performance calculated by the operational performance change calculation means 30 (calculated in the operational performance change calculation step ST3) is within a predetermined range (ST41 in Figure 3). Then, if the degree of change in operational performance is within a predetermined range (if "YES" is shown in ST41 in Figure 3), the operational plan modification means 40 modifies the future operational plan using the degree of change in operational performance uniformly (ST42 in Figure 3). On the other hand, if the degree of change in operating performance is not within a predetermined range (in the case of "NO" in ST41 in Figure 3), the operation plan modification means 40 modifies the future operation plan by assuming that the degree of change in operating performance is monotonically increasing or monotonically decreasing with respect to operating time (ST43 in Figure 3).

[0028] Figure 4 is a timing chart showing an example of the operation plan modification step ST4 executed by the operation plan modification means 40 of the first embodiment. In Figure 4, the leftmost column of the "No. 1" block represents the rolling start time for the No. 1 coil of the rolled material (for the blocks representing the operation plan and revised operation plan, this is the scheduled rolling start time; for the block representing the actual operation, this is the actual rolling start time). The same applies to the No. 2 to No. 20 coils of the rolled material.

[0029] The example shown in Figure 4(a) illustrates a case where, among coils No. 1 to No. 10 of the rolled material that were started rolling within a predetermined time period prior to the current time (the time when the power consumption prediction means 50 predicts (performs prediction calculations)) (in the example shown in Figure 4(a), the past 30 minutes), the actual rolling start times for coil No. 2 and later were uniformly delayed by 2 minutes compared to the scheduled rolling start time. For example, if we define the "degree of change in operational performance" as the difference between the actual rolling start time and the scheduled rolling start time for each coil of rolled material, and define the "predetermined range" as 10% or less of the past predetermined time (3 minutes or less if the predetermined time is 30 minutes), then in the example shown in Figure 4(a), the cumulative value of the delay time of the actual rolling start time over the past 30 minutes is Δt. D Since this is 2 minutes and will be 3 minutes or less, it is determined that the degree of change in operational performance is within a predetermined range, and the operational plan modification means 40 modifies the future operational plan using the degree of change in operational performance uniformly (ST42 in Figure 3). Specifically, as shown in Figure 4(a), the future operational plan (scheduled start time of rolling from coil No. 11 onwards) is modified by the cumulative value of the delay time Δt D The adjustment is made to uniformly delay the process by (2 minutes). In other words, the operation plan adjustment means 40 uses the following formula (1) to determine the scheduled start time t of rolling for the n (n=1, 2, ...) coils of rolled material that are scheduled to be rolled from the current time onward. n to n、rev Correct it to this. t n、rev =t n +Δt D ...(1) Furthermore, if the actual start time of rolling for materials that started rolling in the past 30 minutes was the same as the scheduled start time, that is, if the operation was in line with the operation plan, then Δt Dbecomes 0, and since the degree of change in operation results is within a predetermined range (in the above example, 3 minutes or less), the operation plan correction means 40 will correct the future operation plan using the degree of change in operation results uniformly (ST42 in FIG. 3). However, in this case, since the degree of change in operation results is 0, the corrected operation plan will be the same as the previously obtained operation plan (that is, "correction" in ST42 in FIG. 3 includes the case where the correction amount = 0).

[0030] The example shown in FIG. 4(b) is among Coils No. 1 to No. 10 of the rolled material whose rolling start was scheduled within a predetermined time in the past from the current time (in the example shown in FIG. 4(b), the past 30 minutes). Since the heating time of the rolled material in the heating furnace, which is the previous process of the rolling mill, was insufficient, the extraction from the heating furnace was awaited until the rolled material was heated to the target temperature. During this period, the rolling mill was on standby, the actual rolling start time was delayed, and the delay time gradually increased. In the example shown in FIG. 4(b), the cumulative value Δt of the delay time of the actual rolling start time in the past 30 minutes D is 6 minutes, that is, a 6-minute delay has occurred at the current time, and among Coils No. 1 to No. 10 scheduled in the past 30 minutes, actually only 8 coils, No. 1 to No. 8, have been rolled. In the example shown in FIG. 4(b), the cumulative value Δt of the delay time of the actual rolling start time in the past 30 minutes D is 6 minutes and does not become 3 minutes or less. Therefore, it is determined that the degree of change in operation results is not within the predetermined range. The operation plan correction means 40 assumes that the degree of change in operation results is monotonically increasing or decreasing with respect to the operation time (in the example shown in FIG. 4(b), it is assumed to be monotonically increasing) and corrects the future operation plan (the scheduled rolling start time after Coil No. 9) (ST43 in FIG. 3). Specifically, the operation plan correction means 40 uses the following formula (2) to correct the scheduled rolling start time t of the coil of the rolled material to be rolled the n (n = 1, 2, ···)th time after the current time n to t [[ID=**13**]] n、rev Here, N shown in the following formula (2) is the actual number of coils of the rolled material rolled within a predetermined time in the past (in the example shown in FIG. 4(b), the past 30 minutes) (in the example shown in FIG. 4(b), N = 8). t n、rev =t n +Δt D +(n-1)·Δt D / N···(2) In the example shown in Figure 4(b), for example, for coil No. 11, Δt is added to equation (2) above. D By substituting =6 minutes, n=3, and N=8, we get t n、rev =t n This will result in a +7.5 minute adjustment to the scheduled start time of rolling (t n、rev -t n If we truncate the decimal part of (2), the correction amount becomes 7 minutes, which means we will correct the scheduled start time of rolling for coil No. 11 by 7 minutes. Similarly, for coil No. 17, for example, we add Δt to equation (2) above. D By substituting =6 minutes, n=9, and N=8, t n、rev =t n This adds 12 minutes, so the correction amount will be 12 minutes, meaning the scheduled start time for rolling coil No. 17 will be delayed by 12 minutes.

[0031] Note that in the example shown in Figure 4(b), the planned start time of rolling is considered in 1-minute increments, so the correction amount (t) calculated using equation (2) is used. n、rev -t n If N is not an integer, the decimal part is truncated and the scheduled rolling start time is corrected, but it is also possible to round to the nearest integer. Furthermore, it is possible to make the time resolution finer and set the scheduled rolling start time to 1-second increments. In addition, the N shown in equation (2) does not necessarily have to be an integer. As shown in the example in Figure 4(a), if coil No. 10 is in the process of rolling at the current time, the coils rolled in the past 30 minutes are coils No. 1 to No. 9 plus 1 / 3 of the total rolling time of coil No. 10, so it is sufficient to calculate with N = 9.33. The same applies if a coil was in the process of rolling 30 minutes ago.

[0032] As described above, the operation plan modification means 40 of the first embodiment can reduce the difference between the scheduled start time of rolling and the actual start time of rolling of a coil of rolled material scheduled to be rolled in the future (for example, 30 minutes from the present time) by executing the operation plan modification step ST4 as shown in Figure 4.

[0033] The following describes an example of the results of an evaluation of whether the future operation plan can be appropriately modified by the operation plan modification means 40 of the first embodiment. The actual number of coils of rolled material rolled in the 30 minutes following a certain time t0 is M. act Let M be the planned number of coils of rolled material to be rolled in the 30 minutes from time t0, calculated based on the planned rolling start time before the correction. M ≤ t0+30 <t M+1 This is the planned number of coils that satisfies the condition. Here, t M This is the scheduled start time for rolling the Mth coil of rolled material that will be rolled after time t0. Similarly, t M+1 This is the scheduled start time for rolling the coil of rolled material that is scheduled to be rolled M+1th after time t0. Then, by changing time t0 in 10-minute increments, the M of k groups act (M for each group) act to M i act This is written as follows: i = 1, 2, ..., k) and k sets of M (M for each set is M i This is expressed as follows. Calculate i = 1, 2, ..., k) and calculate the mean squared error RMSE shown in equation (3) below.

number

[0034] Similarly, the planned rolling start time t after modification by ST42 or ST43 in Figure 3 n、rev Based on this, the planned number of coils of rolled material to be rolled in the 30 minutes from time t0 onwards (planned number of coils) is calculated as M rev Let's assume that. M rev is, t M rev≤ t0+30 <t M+1 rev This is the planned number of coils that satisfies the condition. Here, t M rev This is the revised scheduled rolling start time for the Mth rolled material coil that is scheduled to be rolled after time t0. Similarly, t M+1 rev This is the revised rolling start time for the coil of rolled material that is scheduled to be rolled M+1th after time t0. Then, by changing time t0 in 10-minute increments, the M of k groups rev (M for each group) rev to M i rev This is expressed as follows. Calculate i = 1, 2, ..., k) and calculate the mean squared error RMSE shown in equation (4) below. Note that M in equation (4) i act M in equation (3) i act It has the same value as [the other value].

number

[0035] Using data from k=259 sets, the RMSEs shown in equations (3) and (4) were calculated, and it was confirmed that the RMSE shown in equation (4) was 14.5% lower than the RMSE shown in equation (3). This result demonstrates that the operation plan modification means 40 of the first embodiment can appropriately modify future operation plans, thereby reducing the difference between the scheduled start time of rolling and the actual start time of rolling for coils of rolled material scheduled for future rolling.

[0036] <Second Embodiment> A second embodiment of the present invention will be described below. Figure 5 is a flowchart showing the general procedure of the operation plan modification step ST4 performed by the operation plan modification means 40 of the second embodiment. The operation plan modification means 40 of the second embodiment includes a machine learning-generated learning model that takes the degree of change in past operation performance calculated by the operation performance change calculation means 30 (calculated in the operation performance change calculation step ST3) as input and outputs the degree of modification of the future operation plan. This learning model is generated by machine learning using combinations of operational plans and operational results acquired in the past as training data, and is stored in the operational plan modification means 40 (ST6 in Figure 5). Specifically, it is generated by machine learning using the degree of change in operational results calculated from operational plans and operational results acquired in the past prior to a certain operational time as input for training data, and the degree of modification of the operational plan calculated from operational plans and operational results acquired in the past prior to a certain operational time (the degree of modification of the operational plan after a certain operational time necessary to match the operational results after a certain operational time) as output for training data. The operation plan modification means 40 of the second embodiment calculates the degree of modification of the future operation plan using the learning model described above (ST44 in Figure 5). Then, based on the calculated degree of modification of the future operation plan, it modifies the future operation plan (ST45 in Figure 5).

[0037] Figure 6 is a conceptual diagram showing an example of a learning model (input / output items of the learning model) provided by the operation plan modification means 40 of the second embodiment. In the example shown in Figure 6, the operation plan modification means 40 of the second embodiment comprises three learning models: learning model 1, which shows input / output items in Figure 6(a); learning model 2, which shows input / output items in Figure 6(b); and learning model 3, which shows input / output items in Figure 6(c). The learning models 1-3 shown in Figure 6 use the ratio of the actual number of coils rolled in 10 minutes to the planned number of coils as the degree of change in past operational performance (if the operational plan or actual performance is expressed in terms of the rolling start time, the number of coils that will start rolling in 10 minutes should be converted to the planned number of coils or actual number of coils and used), that is, actual number of coils / planned number of coils (actual number of coils / planned number of coils for the time from the present to 10 minutes ago, 10 minutes ago to 20 minutes ago, and 20 minutes ago to 30 minutes ago), and these are used as inputs (explanatory variables) to each learning model 1-3. The inputs to each learning model 1-3 are the same.

[0038] Furthermore, the learning models 1 to 3 shown in Figure 6 use the correction ratio of the planned number of coils (planned number of coils of the rolled material after correction to be rolled in 10 minutes / planned number of coils of the rolled material before correction to be rolled in 10 minutes) as the degree of correction of the future operation plan. The output (dependent variable) of learning model 1 is the correction ratio of the planned number of coils from the current time to 10 minutes later, the output (dependent variable) of learning model 2 is the correction ratio of the planned number of coils from 10 minutes later to 20 minutes later, and the output (dependent variable) of learning model 3 is the correction ratio of the planned number of coils from 20 minutes later to 30 minutes later. When performing machine learning on each of the learning models 1 to 3, the output of the training data for learning model 1 will be the actual number of coils of rolled material obtained in the past from a certain operating time to 10 minutes later / the planned number of coils of rolled material before correction. The output of the training data for learning model 2 will be the actual number of coils of rolled material from 10 minutes to 20 minutes later / the planned number of coils of rolled material before correction. The output of the training data for learning model 3 will be the actual number of coils of rolled material from 20 minutes to 30 minutes later / the planned number of coils of rolled material before correction.

[0039] Using the learning models 1-3 generated as described above, the correction ratios for the planned number of coils from the current time to 10 minutes later, from 10 minutes later to 20 minutes later, and from 20 minutes later to 30 minutes later are calculated. The operation plan correction means 40 uses these correction ratios for the planned number of coils to correct the future operation plan (ST45 in Figure 5). Specifically, if the correction ratio calculated by each learning model 1-3 is r, the operation plan correction means 40 corrects the planned rolling start time after correction to x / r minutes from the current time, compared to the planned rolling start time before correction (let's say x minutes after the current time). In other words, the operation plan correction means 40 corrects the planned rolling start time t of the coil of rolled material that is to be rolled n (n=1, 2, ...) from the current time onward. n Using the following equation (5), t n、rev Correct it to this. t n、rev =t n +x / rx ···(5)

[0040] For learning models 1-3, for example, a random forest regression model can be used, but it is not limited to this; other machine learning models such as neural networks and support vector machines can also be used.

[0041] The following describes an example of the results of an evaluation of whether the future operation plan can be appropriately modified using the operation plan modification means 40 of the second embodiment. Using the learning models 1-3 provided by the operation plan modification means 40 of the second embodiment, the operation plan for the next 30 minutes was modified based on the operation results and operation plan for the past 30 minutes from the current time. As a result, it was confirmed that the mean square error RMSE between the actual number of coils and the planned number of coils for the next 30 minutes was reduced by 27.6% after the modification compared to before the modification. Specifically, the RMSE between the actual number of coils and the planned number of coils from the current time to 10 minutes later was reduced by 16.2%, the RMSE between the actual number of coils and the planned number of coils from 10 minutes to 20 minutes later was reduced by 9.4%, and the RMSE between the actual number of coils and the planned number of coils from 20 minutes to 30 minutes later was reduced by 10.6%. Here, the calculation of the above RMSE was performed using the same k=259 data sets as in the first embodiment.

[0042] Furthermore, the revision of the operation plan is not limited to 30 minutes from the current time; for example, it is possible to revise the operation plan up to 100 minutes from the current time. Also, although the above explanation describes creating multiple (3) learning models 1-3 with different outputs (future revision times), it is not limited to this; it is possible to create only one learning model, or two or four or more learning models.

[0043] By having the operation plan modification means 40 of the second embodiment execute the operation plan modification step ST4 as shown in Figure 5, it is expected that future operation plans can be appropriately modified even if there are a mixture of causes for discrepancies between actual operations and the operation plan within a predetermined time, such as factory equipment troubles, advancement of operating hours, and pitch-down operations. [Explanation of Symbols]

[0044] 10. Methods for obtaining operational plans 20. Methods for obtaining operational performance data 30. Method for calculating changes in operational performance 40. Means of modifying the operation plan 50. Electricity Consumption Prediction Method 100...Electricity usage prediction device

Claims

1. The operation plan acquisition step involves obtaining the factory's operation plan related to operating hours, An operation record acquisition step to acquire the operation record of the aforementioned factory related to the operating time within a specified period in the past, A step for calculating changes in operational performance, which calculates the degree of change in the operational performance relative to the operational plan within the predetermined past time period, Based on the degree of change in the aforementioned operational performance, the operational plan revision step involves revising the future operational plan, The process includes a power consumption forecasting step that predicts the future power consumption of the factory based on the revised operating plan, If the degree of change in the aforementioned operational performance falls within a predetermined range, the future operational plan is revised in the operational plan revision step by uniformly using the degree of change in the aforementioned operational performance. If the degree of change in the aforementioned operational performance is not within a predetermined range, in the operational plan revision step, the future operational plan is revised by assuming that the degree of change in the aforementioned operational performance is monotonically increasing or monotonically decreasing with respect to the operating time. Methods for predicting electricity consumption.

2. An operation plan acquisition step for acquiring the operation plan of a factory relating to the operating time, An operation record acquisition step to acquire the operation record of the aforementioned factory related to the operating time within a specified period in the past, A step for calculating changes in operational performance, which calculates the degree of change in the operational performance relative to the operational plan within the predetermined past time period, Based on the degree of change in the aforementioned operational performance, the operational plan revision step involves revising the future operational plan, The process includes a power consumption forecasting step that predicts the future power consumption of the factory based on the revised operating plan, In the operation plan revision step, a machine learning-generated learning model is used, which takes the degree of change in past operation performance as input and the degree of revision of the future operation plan as output, to calculate the degree of revision of the future operation plan, and the future operation plan is revised based on the calculated degree of revision of the future operation plan. Methods for predicting electricity consumption.

3. A means for obtaining the operation plan of a factory that is related to the operating hours, An operation record acquisition means for acquiring the operation record of the aforementioned factory related to the operating time within a predetermined period in the past, An operational performance change calculation means for calculating the degree of change in the operational performance relative to the operational plan within the aforementioned predetermined time period in the past, An operational plan revision means for revising the future operational plan based on the degree of change in the aforementioned operational performance, The system includes a power consumption forecasting means that predicts the future power consumption of the factory based on the revised operating plan, If the degree of change in the operational performance is within a predetermined range, the operational plan modification means will uniformly use the degree of change in the operational performance to modify the future operational plan. If the degree of change in the operational performance is not within a predetermined range, the operational plan modification means modifies the future operational plan by assuming that the degree of change in the operational performance is monotonically increasing or decreasing with respect to the operating time. Electricity consumption forecasting device.

4. An operation plan acquisition means for acquiring an operation plan of a factory relating to the operating time, An operation record acquisition means for acquiring the operation record of the aforementioned factory related to the operating time within a predetermined period in the past, An operational performance change calculation means for calculating the degree of change in the operational performance relative to the operational plan within the aforementioned predetermined time period in the past, An operational plan revision means for revising the future operational plan based on the degree of change in the aforementioned operational performance, The system includes a power consumption forecasting means that predicts the future power consumption of the factory based on the revised operating plan, The aforementioned operation plan modification means comprises a learning model generated by machine learning, which takes the degree of change in past operation performance as input and the degree of modification of future operation plans as output. The operation plan modification means calculates the degree of modification of the future operation plan using the learning model, and modifies the future operation plan based on the calculated degree of modification of the future operation plan. Electricity consumption forecasting device.

Images