Device control assistance device, device control assistance method, and device control assistance program

Abstract

The present application relates to a device control assistance device for assisting device control using a plurality of process values of a device predicted by a plurality of prediction models. The device has a first prediction model learning unit for learning a plurality of first prediction models as the plurality of prediction models using operation data of the device, i.e., learning data. For a first prediction model of the plurality of first prediction models whose prediction accuracy is less than a reference value, a second prediction model is re-learned using operation data to which additional data is added as learning data. For a second prediction model whose prediction accuracy is less than the reference value, a third prediction model is re-learned using similar operation data similar to an operation point of the device from among the operation data as learning data. A prediction model whose prediction accuracy is the best among the first prediction model, the second prediction model, and the third prediction model is selected as a prediction model.

Description

Technical Field

[0001] This invention relates to an equipment control auxiliary device, an equipment control auxiliary method, and an equipment control auxiliary program.

[0002] This application claims priority based on Japan Patent Application No. 2024-070327 filed with the Japan Patent Office on April 24, 2024, the contents of which are incorporated herein by reference. Background Technology

[0003] Control devices for controlling equipment comprised of various machines are known. In such control devices, control is implemented based on process values ​​related to the characteristics of the equipment. However, when the process values ​​are parameters that are difficult to measure practically, or parameters that, even if they could be measured, are preferred to be avoided for reasons such as cost reduction, a predictive model simulating the equipment's behavior is sometimes used to obtain the process values ​​as computational predictions. Such predictive models can be constructed, for example, through machine learning using past operating data of the equipment as learning data. For instance, in Patent Document 1, by learning a predictive model for each operating condition of the equipment, process values ​​can be appropriately predicted even when switching operating conditions.

[0004] Previous technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent No. 7222943 Summary of the Invention

[0007] The technical problem to be solved by the invention

[0008] In the aforementioned patent document 1, a prediction model is constructed by learning from each operating condition. However, the number of data points available for use as learning data may vary depending on the operating conditions. In this case, depending on the operating conditions, there is a possibility that insufficient learning data may not be available, leading to a decrease in the prediction accuracy of the prediction model constructed through learning.

[0009] At least one embodiment of the present invention was made in view of the above circumstances, and its purpose is to provide an equipment control auxiliary device, an equipment control auxiliary method, and an equipment control auxiliary program, which can achieve good prediction accuracy of process values ​​based on prediction models regardless of the operating conditions of the equipment.

[0010] means for solving technical problems

[0011] To address the aforementioned issues, at least one embodiment of the present invention provides an equipment control auxiliary device for assisting equipment control that utilizes multiple process values ​​of equipment predicted by multiple prediction models, the equipment control auxiliary device comprising: The first prediction model learning unit is used to learn multiple first prediction models, which are multiple prediction models, using the operating data of the device, i.e., the learning data. The second prediction model learning unit is used to relearn the second prediction model by using the running data with added data as the learning data for the first prediction model among the plurality of first prediction models whose prediction accuracy is less than the benchmark value. The third prediction model learning unit is configured to, for the second prediction model whose prediction accuracy is less than the benchmark value, use similar operating data in the operating data that are similar to the operating points of the device as the learning data to relearn the third prediction model; and The prediction model selection unit selects the prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model as the prediction model.

[0012] To address the aforementioned issues, at least one embodiment of the present invention provides an equipment control support method for assisting equipment control that utilizes multiple process values ​​of equipment predicted by multiple prediction models. The equipment control support method comprises the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

[0013] To address the aforementioned issues, at least one embodiment of the present invention provides an equipment control support program for assisting equipment control that utilizes multiple process values ​​of equipment predicted using multiple prediction models. The equipment control support program enables a computer device to perform the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

[0014] Invention Effects

[0015] According to at least one embodiment of the present invention, an equipment control auxiliary device, an equipment control auxiliary method, and an equipment control auxiliary program can be provided, which can achieve good prediction accuracy of process values ​​based on prediction models regardless of the operating conditions of the equipment. Attached Figure Description

[0016] Figure 1 This is an overall structural diagram of the equipment control system of one implementation method.

[0017] Figure 2 It is a schematic diagram representing the virtual space of the specified operating points of the equipment.

[0018] Figure 3 This indicates that the storage in Figure 1 A diagram illustrating the process of calculating the predicted values ​​of the prediction model in the prediction model storage section.

[0019] Figure 4 It means in Figure 1 A diagram illustrating the calculation process of the indicators used for searching operating conditions in the operating condition search section.

[0020] Figure 5 This is a flowchart illustrating one embodiment of a device control auxiliary method. Detailed Implementation

[0021] Hereinafter, several embodiments of the present invention will be described with reference to the accompanying drawings. However, the dimensions, materials, shapes, relative arrangements, etc. of the structures described as embodiments or shown in the drawings are not intended to limit the scope of the present invention, but are merely illustrative examples.

[0022] First, the overall structure of the equipment control auxiliary system 1, which includes the equipment control auxiliary device 100 according to at least one embodiment of the present invention, will be described. Figure 1 This is an overall structural diagram of the equipment control system 1 according to one embodiment.

[0023] The equipment control auxiliary device 100 is a device used to assist the control implemented by the control device 200, which is a control unit for controlling the equipment 10. The equipment 10 is, for example, a boiler device in a thermal power generation plant for generating steam to be supplied to a steam turbine (not shown) connected to a generator, but this is only one example and is not limited thereto.

[0024] The control device 200 controls the operating state of the device 10 by sending and receiving various signals between itself and the device 10. The control performed by the control device 200 is carried out according to preset operating conditions, and it acquires operating parameters from the device 10 while simultaneously outputting control signals corresponding to those operating parameters to the device 10. Within the device 10, the control of each component machine is performed based on the control signals input from the control device 200.

[0025] Furthermore, the detailed structure of the control device 200 will be referenced from well-known examples and is omitted here.

[0026] The equipment control auxiliary device 100 includes an external input interface 102, a preprocessing unit 104, an operating data storage unit 106, a prediction model learning unit 108, a prediction model storage unit 110, a prediction accuracy determination unit 112, an operating condition search unit 114, a search range setting unit 116, and an external output interface 118.

[0027] External input interface 102 is an interface structure for inputting data from the control device 200, which is an external device serving as a device control auxiliary device 100. Various types of data processed in the control device 200 are input to external input interface 102. This data includes data processed by the preprocessing unit 104 to become running data used as learning data during machine learning of the prediction model M in the prediction model learning unit 108. Furthermore, data input from the control device 200 is repeatedly implemented at predetermined time intervals.

[0028] The preprocessing unit 104 is a structure for preprocessing the data input from the external input interface 102. The preprocessing includes at least the extraction of data from the input to the external input interface 102 and the processing of running data used as learning data when performing machine learning of the prediction model M in the prediction model learning unit 108, but the details of the processing are not particularly limited.

[0029] The operation data accumulation unit 106 is a structure (e.g., a database) for accumulating operation data created by the preprocessing unit 104. In the equipment control auxiliary device 100, each time data is input from the control device 200 to the external input interface 102, the preprocessing unit 104 performs preprocessing, thereby processing the operation data and accumulating it sequentially in the operation data accumulation unit 106.

[0030] The predictive model learning unit 108 is a structure used to perform machine learning on the running data accumulated in the running data accumulation unit 106 as learning data, thereby constructing a predictive model M. This machine learning specifies explanatory variables and objective functions from the parameters contained in the running data, and learns the correlation between the explanatory variables and the objective variable according to the machine learning algorithm, thereby constructing the predictive model M. The machine learning algorithm can be linear regression, neural networks, or random forests, but is not limited to these.

[0031] The prediction model learning unit 108 includes a first prediction model learning unit 120, a second prediction model learning unit 122, and a third prediction model learning unit 124.

[0032] The first prediction model learning unit 120 is a structure for constructing a prediction model (hereinafter appropriately referred to as "the first prediction model M1") by using machine learning to build a prediction model through machine learning by using the running data accumulated in the running data accumulation unit 106 as learning data. The machine learning performed by the first prediction model learning unit 120 is the machine learning performed before the machine learning performed by the second prediction model learning unit 122 and the third prediction model learning unit 124, and is used to construct the so-called original prediction model M.

[0033] The second prediction model learning unit 122 is a structure for constructing a prediction model (hereinafter appropriately referred to as "the second prediction model M2") by relearning through supplementary running data. In supplementary running data relearning, the running data used as learning data during the construction of the first prediction model M1 can be used as learning data for relearning, thereby constructing the second prediction model M2. As described above, running data is sequentially accumulated in the running data accumulation unit 106, and supplementary data is the running data accumulated in the running data accumulation unit 106 after the learning of the first prediction model M1. In supplementary running data relearning, by using the running data whose data volume has increased through supplementary data in this way, a second prediction model M2 with improved prediction accuracy compared to the first prediction model M1 can be constructed.

[0034] The third prediction model learning unit 124 is a structure for constructing a prediction model (hereinafter appropriately referred to as "the third prediction model M3") through similar operating data relearning. In similar operating data relearning, the third prediction model M3 can be constructed by implementing similar operating data relearning, in which similar operating data belonging to a defined operating area of ​​the operating point including the device 10 in the operating data stored in the operating data accumulation unit 106 is used as learning data.

[0035] Here, for reference Figure 2 The similar running data processed as learning data in the learning unit 124 of the third prediction model will be explained in detail. Figure 2 This is a schematic diagram representing the virtual space V of the specified operating point of device 10. Figure 2 In the virtual space V, where multiple parameters representing the operating point are defined as spatial axes, the current operating point of the device is shown (in...). Figure 2 (In the illustration, for ease of understanding, the case where the dimension of the virtual space V is "3" is shown.)

[0036] exist Figure 2 In the virtual space V defined by three spatial axes x, y, and z, the coordinates A corresponding to the current running point of the device are set as (x1, y1, z1), and the coordinates B of any point in the virtual space are set as (x2, y2, z2). The deviation L between the two is expressed by the following formula.

[0037] L = {(x1-x2)} 2 +(y1-y2) 2 +(z1-z2) 2} 0.5

[0038] Similar operating data is defined, for example, as data within the operating data that is contained within a similar operating range where the deviation L is within a specified value L0 (in Figure 2 The diagram illustrates the similar operating range as shown by the interior of a sphere represented by dashed lines. Therefore, in the relearning of similar operating data, by using the operating data of operating points with similar operating states as learning data, the prediction accuracy of the prediction model M can be appropriately improved.

[0039] Furthermore, this example illustrates the use of Euclidean distance to define the deviation L, but the deviation L can also be defined using Mahalanobis distance. Moreover, similar running data can be defined as data extracted from the running data in ascending order of deviation L until a predetermined proportion or quantity is reached. As mentioned earlier, when similar running data is defined as data with a deviation L within a predetermined value L0, there is a possibility that insufficient learning data cannot be extracted when there is not enough data within this range, making it impossible for the prediction model M to learn. In contrast, in this approach, by defining similar running data as data extracted until a predetermined proportion or quantity is reached, even when there is a limited amount of running data with small deviation L, insufficient data required for learning can be prevented, and similar running data relearning can be performed (this has the same meaning as setting the predetermined value L0 as a variable value and deciding to extract running data with small deviation L from the running data in a predetermined quantity or proportion).

[0040] Back Figure 1The prediction model storage unit 110 is a structure for storing the prediction models M constructed in the prediction model learning unit 108. In the prediction model learning unit 108, at least one of the first prediction model M1, the second prediction model M2, or the third prediction model M3 is constructed as a prediction model M, and these prediction models M are stored in the prediction model storage unit 110, thereby becoming a state that can be appropriately utilized.

[0041] The prediction accuracy determination unit 112 is a structure used to determine the prediction accuracy of each prediction model M (first prediction model M1, second prediction model M2, or third prediction model M3) stored in the prediction model storage unit 110. The prediction accuracy of the prediction model M can be calculated, for example, by statistically analyzing the mean absolute error of the predicted value relative to the actual measured value or other verification data, or by employing any known method, and is not limited to these methods. In the prediction accuracy determination unit 112, by comparing the calculated prediction accuracy with a reference value pre-stored in the tolerance accuracy database 126, it is possible to determine whether the prediction accuracy of the prediction model M is less than the reference value.

[0042] The operation condition search unit 114 is a structure for searching the operation conditions of the control device 200 using the prediction model M stored in the prediction model storage unit 110. The search for operation conditions is carried out by calculating an index corresponding to the performance of the device 10 based on the predicted value calculated using the prediction model M, and performing the search within the search range set by the search range setting unit 116 to optimize the index.

[0043] Here, Figure 3 This indicates that the storage in Figure 1 A diagram illustrating the process of calculating the predicted value P of the prediction model M in the prediction model storage unit 110. Figure 4 It means in Figure 1 A diagram illustrating the calculation process of the indicators used for searching operating conditions in the operating condition search unit 114.

[0044] like Figure 3 As shown, the prediction model M outputs a predicted value P corresponding to the target variable by taking parameters corresponding to the explanatory variables as input. For example, in the case of a boiler unit, the predicted value P is a process value related to the operating state of the boiler unit, more specifically, the temperature, pressure, metal temperature of steam in various parts of the boiler unit, or the NOx concentration emitted from the boiler unit. The explanatory variables are various operating parameters of the boiler unit, more specifically, the angle of the fuel-injecting burner, the opening degree of various dampers for adjusting air volume, etc. In the prediction model learning unit 108, prediction models M are constructed according to the types of these target variables, and multiple prediction models M according to the types of target variables are stored in the prediction model storage unit 110.

[0045] Furthermore, regarding the prediction model M corresponding to a certain objective variable, if a second prediction model M2 or a third prediction model M3 is also stored in addition to the first prediction model M1, the prediction model with the best prediction accuracy among these prediction models is used to calculate the predicted value P. That is, the predicted value P of multiple objective variables is calculated using the corresponding multiple prediction models M, but each prediction model M uses the prediction model with the best prediction accuracy among the prediction models related to the corresponding objective variable (at least one of the first prediction model M1, the second prediction model M2, or the third prediction model M3).

[0046] The operation condition search unit 114 inputs input parameters (initial values) corresponding to pre-set operation conditions to multiple prediction models M stored in the prediction model storage unit 110, thereby obtaining their respective prediction values ​​P (hereinafter appropriately referred to as "predicted values ​​P1, P2, ..."). These prediction values ​​P are converted into an index IN for evaluating the prediction values ​​P using a prescribed conversion formula. Specifically, as follows... Figure 4 As shown, the operation condition search unit 114 converts each predicted value P1, P2, ... into corresponding scores SC1, SC2, ... respectively, and calculates the index IN (=SC1+SC2+...) as the sum of these. Then, the operation condition search unit 114 searches for operation conditions to make the index IN optimal (e.g., minimum or maximum) (that is, by repeatedly changing the operation conditions and calculating the index IN, the operation conditions that make the index IN optimal are determined).

[0047] Furthermore, the search for operating conditions in the operating condition search unit 114 is performed within a search range set by the search range setting unit 116. By default, this search range is set to a first search range, but as described later, when performing search range limiting processing, the search range is limited to a second search range that is specified as narrower than the first search range. By limiting the search range of operating conditions to a range where the prediction accuracy of the prediction model M is good, or to a range that does not significantly impair the operating state, the prediction accuracy or operational reliability can be substantially improved.

[0048] The operating conditions searched by the operating condition search unit 114 are output to the control device 200 via the external output interface 118. The control device 200 controls the device 10 according to the operating conditions output in this manner, thereby appropriately controlling the device 10.

[0049] Next, the equipment control assistance method implemented using the equipment control assistance device 100 having the above-described structure will be described. Figure 5 This is a flowchart illustrating one embodiment of a device control auxiliary method.

[0050] First, operational data is collected in the operational data accumulation unit 106 (step S100). In step S100, data from the control device 200 controlling the operating device 10 is input to the device control auxiliary device 100 via the external input interface 102. This data is preprocessed by the preprocessing unit 104, thereby being processed into operational data. The processed operational data is collected by being sequentially accumulated in the operational data accumulation unit 106. As described above, this accumulation of operational data is performed sequentially. In the following description, it is assumed that the operational data accumulation unit 106 has accumulated enough operational data to learn the prediction model M (first prediction model M1).

[0051] Next, the prediction model learning unit 108 constructs a first prediction model M1 by using the running data accumulated in the running data accumulation unit 106 as learning data (step S101). In step S101, the first prediction model M1 is constructed as prediction model M by the first prediction model learning unit 120 in the prediction model learning unit 108. In the first prediction model learning unit 120, the first prediction model M1 is constructed based on at least one pre-prepared machine learning algorithm. In this embodiment, in particular, multiple machine learning algorithms are pre-prepared in the first prediction model learning unit 120, and multiple prediction model candidates corresponding to the first prediction model M1 are generated for each objective variable through multiple machine learning algorithms. Then, in the first prediction model learning unit 120, the prediction accuracy of these multiple prediction model candidates corresponding to the first prediction model M1 is calculated, and the prediction model with the best prediction accuracy is adopted as the first prediction model M1 for the objective variable. The first prediction model M1 adopted in this way is stored in the prediction model storage unit 110. The first prediction model M1 stored in the prediction model storage unit 110 can be appropriately accessed by the operation condition search unit 114.

[0052] Furthermore, the learning of the first prediction model M1 in step S101 can be performed at predetermined intervals T1. That is, at predetermined intervals T1, the operating data accumulated in the operating data accumulation unit 106 can be used as learning data for machine learning, thereby updating the first prediction model M1 based on the latest operating data.

[0053] Next, the prediction accuracy determination unit 112 determines whether the prediction accuracy of the first prediction model M1 constructed in step S101 is less than a preset reference value (step S102). In step S102, the first prediction model M1 stored in the prediction model storage unit 110 is accessed every predetermined period T2, and the prediction accuracy of the first prediction model M1 is calculated and compared with the reference value. The predetermined period T2 may be different from the aforementioned predetermined period T1, and in particular, it may be shorter than the predetermined period T1.

[0054] Next, if the prediction accuracy is less than the baseline value (step S102: Yes), the prediction model learning unit 108 performs additional running data relearning through the second prediction model learning unit 122, thereby learning a second prediction model M2 (step S103). In the additional running data relearning, the running data used as learning data during the construction of the first prediction model M1 in step S101 is used as learning data for relearning, thereby constructing the second prediction model M2. Therefore, if the prediction accuracy of the first prediction model M1 is less than the baseline value, by using the running data with increased data volume through additional running data relearning, a second prediction model M2 with improved prediction accuracy compared to the first prediction model M1 can be constructed.

[0055] Furthermore, the construction of the second prediction model M2 in the second prediction model learning unit 122 can be performed using at least one machine learning algorithm, similar to the first prediction model learning unit 120 described above. In this embodiment, particularly in the second prediction model learning unit 122, multiple machine learning algorithms are prepared in advance, and multiple prediction model candidates corresponding to the second prediction model M2 are generated for each objective variable using multiple machine learning algorithms. Then, in the second prediction model learning unit 122, the prediction accuracy of these multiple prediction model candidates corresponding to the second prediction model M2 is calculated, and the prediction model with the best prediction accuracy is adopted as the second prediction model M2 for the objective variable. The constructed second prediction model M2, replacing the aforementioned first prediction model M1 or appending to the aforementioned first prediction model M1, is stored in the prediction model storage unit 110. The second prediction model M2 stored in the prediction model storage unit 110 can be appropriately accessed by the operation condition search unit 114.

[0056] Next, the prediction accuracy determination unit 112 determines whether the prediction accuracy of the second prediction model M2 constructed in step S103 is less than a preset benchmark value (step S104). In step S104, referring to the aforementioned step S102, the prediction accuracy of the second prediction model M2 is calculated, and the prediction accuracy is compared with the benchmark value.

[0057] Furthermore, the reference value used as the determination criterion in step S104 may be the same as or different from the reference value used as the determination criterion in step S102.

[0058] Next, if the prediction accuracy is less than the baseline value (step S104: Yes), the prediction model learning unit 108 performs similar operation data relearning through the third prediction model learning unit 124 (step S105), or the search range setting unit 116 performs search range limitation processing (step S106). Figure 1As shown, one embodiment of the equipment control auxiliary device 100 includes a display unit 130 and a selection unit 132, thereby enabling the operator to selectively perform one of steps S105 or S106.

[0059] Display unit 130 is a display device such as a monitor used to display various information required by the operator. The information displayed in display unit 130 can broadly include information used by the operator when selecting one of steps S105 or S106, for example, it can include the prediction accuracy calculated during the determination process performed in prediction accuracy determination unit 112.

[0060] The selection unit 132 is a structure for providing an operator with a choice between step S105 or S106 based on the information displayed on the display unit 130. It can be an input device such as a keyboard, mouse, or touch panel. The operator refers to the information displayed on the display unit 130 and operates the selection unit 132 based on their own judgment. This allows them to select which step S105 or S106 to perform if the prediction accuracy of the second prediction model M2 is less than a reference value.

[0061] Furthermore, generally speaking, the computation time for search range limitation processing tends to be shorter than that for relearning from similar running data. Therefore, as a selection criterion for the selection unit 132, for example, if a high-quality prediction model M with good prediction accuracy is required, step S105 can be selected; on the other hand, if minimizing computation time is a priority, step S106 can be selected.

[0062] When similar running data relearning is performed in step S105, the third prediction model M3 is constructed by using similar running data as learning data through the third prediction model learning unit 124. In similar running data relearning, by using running data of running points with similar running states as learning data for relearning, a third prediction model M3 with improved prediction accuracy can be obtained.

[0063] Furthermore, the construction of the third prediction model M3 in the third prediction model learning unit 124 can be performed using at least one machine learning algorithm, similar to the construction of the first prediction model M1 in the first prediction model learning unit 120 and the construction of the second prediction model M2 in the second prediction model learning unit 122. In this embodiment, particularly in the third prediction model learning unit 124, multiple machine learning algorithms are prepared in advance, and multiple prediction model candidates corresponding to the third prediction model M3 are generated for each objective variable using multiple machine learning algorithms. Then, in the third prediction model learning unit 124, the prediction accuracy of these multiple prediction model candidates corresponding to the multiple third prediction models M3 is calculated, and the prediction model with the best prediction accuracy is adopted as the third prediction model M3 for the objective variable. The constructed third prediction model M3, either replacing the aforementioned second prediction model M2 or appending to the aforementioned second prediction model M2, is stored in the prediction model storage unit 110. The third prediction model M3 stored in the prediction model storage unit 110 can be appropriately accessed by the operation condition search unit 114.

[0064] On the other hand, when the search range limitation process is implemented in step S106, the executability of the operation optimization using the prediction model can be improved by limiting the search range when searching for operating conditions in step S110 (described later) to a specified range. For example, when the search range is limited to a range where the prediction accuracy in the prediction model M is above a threshold, the prediction accuracy can be substantially improved. Furthermore, when the search range is limited to a range that does not significantly impair the operating state, the operation optimization using the prediction model M can be performed while maintaining safe operation.

[0065] Next, for each of the multiple prediction models M, the prediction model with the best prediction accuracy among the first prediction model M1, the second prediction model M2, or the third prediction model M3 is selected (step S107). As described above, multiple prediction models are prepared according to the type of objective variable, and for each prediction model, at least one of the first prediction model M1, the second prediction model M2, or the third prediction model M3 is constructed in the aforementioned steps. In step S107, the prediction model with the best prediction accuracy among the first prediction model M1, the second prediction model M2, or the third prediction model M3 is adopted as each prediction model M.

[0066] Next, the operating conditions are searched using the prediction models M selected in step S107 (step S108), and the searched operating conditions are output to the control device 200 via the external output interface 118 (step S109). In step S108, as described above, referencing Figure 4 Using the prediction models M selected in step S107, the index IN is calculated based on each predicted value P, and the operating conditions are searched in a way that makes the index IN optimal.

[0067] Furthermore, if a search range limitation process is implemented in step S106, the search range of the operation conditions in step S108 is limited to a specified range.

[0068] Furthermore, without departing from the spirit of the present invention, the constituent elements in the above embodiments can be appropriately replaced with known constituent elements, and the above embodiments can also be appropriately combined.

[0069] The contents described in the above embodiments can be understood, for example, as follows.

[0070] (1) An equipment control auxiliary device of one type is used to assist equipment control that uses multiple process values ​​of equipment predicted by multiple prediction models, the equipment control auxiliary device comprising: The first prediction model learning unit is used to learn multiple first prediction models, which are multiple prediction models, using the operating data of the device, i.e., the learning data. The second prediction model learning unit is used to relearn the second prediction model by using the running data with added data as the learning data for the first prediction model among the plurality of first prediction models whose prediction accuracy is less than the benchmark value. The third prediction model learning unit is configured to, for the second prediction model whose prediction accuracy is less than the benchmark value, use similar operating data in the operating data that are similar to the operating points of the device as the learning data to relearn the third prediction model; and The prediction model selection unit selects the prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model as the prediction model.

[0071] According to the method described in (1) above, operating data is used as learning data to construct multiple first prediction models corresponding to multiple process values ​​as prediction objects. For these first prediction models whose prediction accuracy is less than the benchmark value, additional operating data is added for relearning, and the additional operating data is used as learning data for relearning, thereby constructing a second prediction model with improved prediction accuracy. If the prediction accuracy of the second prediction model is still less than the benchmark value, similar operating data is used for relearning, and similar operating data that is similar to the operating point of the equipment is used as learning data for relearning, thereby improving the prediction accuracy of a third prediction model with improved prediction accuracy. In the equipment control device, the prediction model with the best prediction accuracy among these first, second, or third prediction models is selected as the prediction model used to predict multiple process values ​​respectively. As a result, each process value can be predicted with high accuracy, and the control performance of the equipment control device can be appropriately improved.

[0072] (2) In another manner, in the manner described in (1) above, At least one of the first prediction model learning unit, the second prediction model learning unit, or the third prediction model learning unit generates multiple prediction model candidates using multiple machine learning algorithms for each of the plurality of prediction models, and selects the prediction model with the best prediction accuracy among the plurality of prediction model candidates as the first prediction model, the second prediction model, or the third prediction model.

[0073] According to the method described in (2) above, multiple prediction model candidates are generated using various machine learning algorithms in at least one of the first prediction model learning unit, the second prediction model learning unit, or the third prediction model learning unit. Then, by selecting the prediction model with the best prediction accuracy among these multiple prediction model candidates as the prediction model, a prediction model with excellent prediction accuracy can be constructed.

[0074] (3) In another manner, in the manner described in (1) or (2) above, The similar operating data is constructed by extracting a specified proportion or a specified number of data with small deviations from the operating point from the operating data and the supplementary data.

[0075] According to the method described in (3) above, the similar operating data used as learning data in similar operating data relearning is composed of a specified proportion or a specified number of data with small deviations from the operating point extracted from the operating data and supplementary data. Therefore, in similar operating data relearning, by performing relearning using such similar operating data, the prediction accuracy of the prediction model can be appropriately improved.

[0076] (4) In another manner, in any of the manners (1) to (3) above, The additional data is data collected from the device since the last learning of the first prediction model.

[0077] According to the method described in (4) above, the learning data used in the supplementary running data relearning includes the running data from the previous learning of the prediction model and the supplementary data collected from the device from the previous learning to the relearning. Thus, by using relearning with more learning data than the previous learning, the prediction accuracy of the prediction model can be appropriately improved.

[0078] (5) In another embodiment, in any of the embodiments (1) to (4) above, it further comprises: The indicator calculation unit is used to calculate the indicators related to the performance of the equipment based on the multiple process values ​​predicted using the multiple prediction models; Operating condition search unit, configured to search for operating conditions of the device based on the indicators; and A control unit is used to control the device based on the operating conditions.

[0079] According to the above (5), by using the predicted value of the process value obtained by the prediction model with improved prediction accuracy in the above manner, the operating conditions of the equipment can be appropriately searched to control the equipment.

[0080] (6) In another manner, in any of the manners (1) to (5) above, The equipment is a boiler unit.

[0081] According to the above (6), a prediction model with good prediction accuracy can be used to appropriately predict process values ​​related to the characteristics of the boiler unit.

[0082] (7) An equipment control assistance method is used to assist equipment control that uses multiple process values ​​of equipment predicted by multiple prediction models, and the equipment control assistance method includes the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

[0083] According to the method described in (7) above, operating data is used as learning data to construct multiple first prediction models corresponding to multiple process values ​​as prediction objects. For these first prediction models whose prediction accuracy is less than the benchmark value, relearning is performed by adding operating data, using the added operating data as learning data, thereby constructing a second prediction model with improved prediction accuracy. If the prediction accuracy of the second prediction model is still less than the benchmark value, relearning is performed by using similar operating data, using similar operating data that is similar to the operating point of the equipment as learning data, thereby improving the prediction accuracy of the third prediction model with improved prediction accuracy. In the equipment control device, the prediction model with the best prediction accuracy among these first, second, or third prediction models is selected as the prediction model used to predict multiple process values ​​respectively. Thus, it is possible to predict each process value with high accuracy and to appropriately improve the control performance of the equipment control device.

[0084] (8) A device control auxiliary program is used to assist device control that uses multiple process values ​​of the device predicted by multiple prediction models, wherein the device control auxiliary program enables a computer device to perform the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

[0085] According to the method described in (8) above, operating data is used as learning data to construct multiple first prediction models corresponding to multiple process values ​​as prediction objects. For these first prediction models whose prediction accuracy is less than the benchmark value, additional operating data is added for relearning, and the additional operating data is used as learning data for relearning, thereby constructing a second prediction model with improved prediction accuracy. If the prediction accuracy of the second prediction model is still less than the benchmark value, similar operating data is used for relearning, and similar operating data that is similar to the operating point of the equipment is used as learning data for relearning, thereby improving the prediction accuracy of the third prediction model with improved prediction accuracy. In the equipment control device, the prediction model with the best prediction accuracy among these first, second, or third prediction models is selected as the prediction model used to predict multiple process values ​​respectively. As a result, each process value can be predicted with high accuracy, and the control performance of the equipment control device can be appropriately improved.

[0086] Symbol Explanation

[0087] 1-Equipment control system, 10-Equipment, 100-Equipment control auxiliary device, 102-External input interface, 104-Preprocessing unit, 106-Operating data accumulation unit, 108-Predictive model learning unit, 110-Predictive model storage unit, 112-Predictive accuracy determination unit, 114-Operating condition search unit, 116-Search range setting unit, 118-External output interface, 120-First predictive model learning unit, 122-Second predictive model learning unit, 124-Third predictive model learning unit, 130-Display unit, 132-Selection unit, 200-Control device, M-Predictive model, M1-First predictive model, M2-Second predictive model, M3-Third predictive model, V-Virtual space.

Claims

1. An equipment control auxiliary device for assisting equipment control using multiple process values ​​of the equipment predicted by multiple prediction models, the equipment control auxiliary device comprising: The first prediction model learning unit is used to learn multiple first prediction models, which are multiple prediction models, using the operating data of the device, i.e., the learning data. The second prediction model learning unit is used to relearn the second prediction model by using the running data with added data as the learning data for the first prediction model among the plurality of first prediction models whose prediction accuracy is less than the benchmark value. The third prediction model learning unit is configured to, for the second prediction model whose prediction accuracy is less than the benchmark value, use similar operating data in the operating data that are similar to the operating points of the device as the learning data to relearn the third prediction model; and The prediction model selection unit selects the prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model as the prediction model.

2. The equipment control auxiliary device according to claim 1, wherein, At least one of the first prediction model learning unit, the second prediction model learning unit, or the third prediction model learning unit generates multiple prediction model candidates using multiple machine learning algorithms for each of the plurality of prediction models, and selects the prediction model with the best prediction accuracy among the plurality of prediction model candidates as the first prediction model, the second prediction model, or the third prediction model.

3. The equipment control auxiliary device according to claim 1 or 2, wherein, The similar operating data is constructed by extracting a specified proportion or a specified number of data with small deviations from the operating point from the operating data and the supplementary data.

4. The equipment control auxiliary device according to claim 1 or 2, wherein, The additional data is data collected from the device since the last learning of the first prediction model.

5. The equipment control auxiliary device according to claim 1 or 2, further comprising: The indicator calculation unit is used to calculate the indicators related to the performance of the equipment based on the multiple process values ​​predicted using the multiple prediction models; Operating condition search unit, configured to search for operating conditions of the device based on the indicators; and A control unit is used to control the device based on the operating conditions.

6. The equipment control auxiliary device according to claim 1 or 2, wherein, The equipment is a boiler unit.

7. A method for assisting equipment control, used to assist equipment control that utilizes multiple process values ​​of the equipment predicted by multiple prediction models, the method comprising the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

8. An equipment control auxiliary program for assisting equipment control using multiple process values ​​of the equipment predicted by multiple predictive models, the equipment control auxiliary program enabling a computer device to perform the following steps: The operating data of the device, i.e. the learning data, is used to learn multiple first prediction models, which are the multiple prediction models; For the first prediction model whose prediction accuracy is less than the baseline value among the plurality of first prediction models, the running data with added data is used as the learning data to relearn the second prediction model; For the second prediction model whose prediction accuracy is less than the benchmark value, similar operating data in the operating data that are similar to the operating points of the device are used as the learning data to relearn the third prediction model; and The prediction model with the best prediction accuracy among the first prediction model, the second prediction model, and the third prediction model is selected as the prediction model.

Images