Data drift correction method, system, device, medium and program product

Abstract

The invention provides a data drift correction method, system and device, a medium and a program product. The correction method comprises the following steps: acquiring real-time operation data of target equipment; based on the real-time operation data, the benchmark reference data set and the energy efficiency weight, determining first operation data with data drift; correcting the first operation data by adopting a preset correction mode to obtain third operation data; obtaining a first energy efficiency score of the benchmark reference data set and a second energy efficiency score of the third operation data according to a preset energy efficiency evaluation model; and in response to the condition that the difference value between the first energy efficiency score and the second energy efficiency score is greater than a preset energy efficiency score deviation threshold value, adjusting the energy efficiency weight, returning to re-determine the first operation data and the second operation data until the difference value meets a preset condition, and taking the latest third operation data as the target operation data of the target equipment. According to the invention, the dual rationality of the correction result in physical significance and energy efficiency is ensured.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of sensor calibration, in particular to a data drift correction method, system, device, medium and program product. BACKGROUND

[0002] For the problem of boiler data drift, the common correction methods mainly include threshold judgment method, statistical process control method (SPC) and correction method based on physical model or data driven model. Among them, the threshold judgment method identifies and replaces abnormal data by setting fixed upper and lower limits, which is simple and easy to implement, but it is difficult to cope with the coupling relationship between multiple measuring points, and it is easy to cause "misjudgment" or "omission"; the statistical process control method relies on the mean and variance of historical data to construct a control interval, which improves adaptability to a certain extent, but the response ability to working condition change is limited, especially in the actual scene of frequent load fluctuation and variable coal quality, the benchmark is easy to lose accuracy; the correction method based on single physical model or data driven model often ignores the internal relationship between system overall energy efficiency and each measuring point, resulting in unreasonable correction results in physical sense, and even introducing secondary deviation.

[0003] It can be seen that the existing correction methods for boiler data drift lack multi-measuring point collaborative analysis capability, do not fully consider the interaction between variables, have poor adaptability to dynamic changes of operating conditions, are difficult to establish a flexible and effective correction benchmark, and the correction process is often independent of the energy efficiency evaluation system, resulting in disconnection between data drift correction and energy efficiency target. SUMMARY

[0004] The technical problem to be solved by the present disclosure is to overcome the defects in the prior art that the correction of boiler data drift lacks multi-measuring point collaborative analysis capability, has poor adaptability to dynamic changes of operating conditions, and the data drift correction is disconnected with the energy efficiency target, and provide a data drift correction method, system, device, medium and program product.

[0005] The present disclosure solves the above technical problems by the following technical solutions:

[0006] In a first aspect, the present disclosure provides a data drift correction method, which comprises:

[0007] obtaining real-time operating data including different preset detection parameters in a target device;

[0008] determining first operating data with data drift and second operating data without data drift based on the real-time operating data, a benchmark reference data set, and an energy efficiency weight corresponding to each preset detection parameter; the benchmark reference data set is pre-constructed based on normal historical operating data of different preset detection parameters;

[0009] The first operation data is corrected by using a preset correction method to obtain corrected operation data, and the corrected operation data and the second operation data are combined to obtain third operation data.

[0010] A first energy efficiency score of the reference data set and a second energy efficiency score of the third operation data are obtained according to a preset energy efficiency evaluation model;

[0011] In response to a difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, the energy efficiency weight corresponding to each preset detection parameter is adjusted, and the step of determining the first operation data with data drift and the second operation data without data drift based on the real-time operation data, the reference data set, and the energy efficiency weight corresponding to each preset detection parameter is returned to be executed until the difference is less than or equal to the preset energy efficiency score deviation threshold or a maximum iteration number is reached, and the latest obtained third operation data is taken as the target operation data of different preset detection parameters in the target device.

[0012] Optionally, the step of correcting the first operation data by using a preset correction method to obtain corrected operation data comprises:

[0013] The first detection parameter corresponding to the first operation data and the second detection parameter corresponding to the second operation data are obtained;

[0014] Based on the reference data set, first normal historical operation data corresponding to the first detection parameter and second normal historical operation data corresponding to the second detection parameter are obtained, and a data drift correction model is constructed;

[0015] The second operation data is input into the data drift correction model to obtain the corrected operation data.

[0016] Optionally, the step of obtaining, based on the reference data set, first normal historical operation data corresponding to the first detection parameter and second normal historical operation data corresponding to the second detection parameter, and constructing a data drift correction model comprises:

[0017] The second normal historical operation data is taken as input, and the first normal historical operation data is taken as output to train a preset model to obtain the data drift correction model;

[0018] The preset model comprises any one of a multiple linear regression model, a support vector regression model, and a neural network learning model.

[0019] Optionally, the reference data set includes a plurality of different working condition data subsets, and the different working condition data subsets correspond to different operating conditions.

[0020] The step of determining the first operating data with data drift and the second operating data without data drift based on the real-time operating data, the reference data set, and the energy efficiency weight corresponding to each preset detection parameter includes:

[0021] acquiring a target operating condition corresponding to the real-time operating data, and determining a matching target working condition according to the target operating condition;

[0022] determining the first operating data and the second operating data based on the real-time operating data, a target working condition data subset corresponding to the target working condition, and the energy efficiency weight.

[0023] Optionally, the step of determining the first operating data and the second operating data based on the real-time operating data, the target working condition data subset corresponding to the target working condition, and the energy efficiency weight includes:

[0024] calculating Mahalanobis distance between the real-time operating data and the normal historical operating data in the target working condition data subset;

[0025] in response to the Mahalanobis distance being greater than a second preset threshold, determining the first operating data and the second operating data based on the real-time operating data, the normal historical operating data in the target working condition data subset, and the energy efficiency weight corresponding to each preset detection parameter;

[0026] or,

[0027] in response to the Mahalanobis distance being less than or equal to the second preset threshold, determining that the real-time operating data all do not have data drift.

[0028] Optionally, the first operating data satisfies the following formula:

[0029] ;

[0030] wherein, denotes real-time operating data corresponding to the jth preset detection parameter, denotes a mean value of the normal historical operating data corresponding to the jth preset detection parameter in the target working condition data subset, denotes the energy efficiency weight corresponding to the jth preset detection parameter, denotes a preset drift threshold.

[0031] A second aspect of the present disclosure provides a data drift correction system, the correction system comprising:

[0032] The acquisition module is used to acquire real-time operating data from the target device, including different preset detection parameters;

[0033] The detection module is used to determine the first operating data that has experienced data drift and the second operating data that has not experienced data drift based on the real-time operating data, the benchmark reference dataset, and the energy efficiency weight corresponding to each preset detection parameter; the benchmark reference dataset is pre-constructed based on normal historical operating data with different preset detection parameters;

[0034] The calibration module is used to calibrate the first running data using a preset calibration method to obtain calibrated running data, and to merge the calibrated running data and the second running data to obtain third running data;

[0035] The verification module is used to obtain the first energy efficiency score of the benchmark reference dataset and the second energy efficiency score of the third running data according to the preset energy efficiency evaluation model.

[0036] The determination module is used to adjust the energy efficiency weight corresponding to each preset detection parameter in response to the difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, and return to call the detection module until the difference is less than or equal to the preset energy efficiency score deviation threshold or the maximum number of iterations is reached, and the latest obtained third operating data is used as the target operating data for different preset detection parameters in the target device.

[0037] A third aspect of this disclosure provides an electronic device including a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein the processor, when executing the computer program, implements the data drift correction method described in the first aspect of this disclosure.

[0038] A fourth aspect of this disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the data drift correction method described in the first aspect of this disclosure.

[0039] A fifth aspect of this disclosure provides a computer program product including a computer program that, when executed by a processor, implements the data drift correction method described in the first aspect of this disclosure.

[0040] Based on common knowledge in the field, the above optional conditions can be combined arbitrarily to obtain the optional examples of this disclosure.

[0041] The positive progress effect of the disclosure is that a closed-loop correction system with an energy efficiency evaluation model as the core is constructed, the energy efficiency score deviation before and after correction is verified through iteration, and the weight of the preset detection parameter is dynamically adjusted, so that the correction result is always optimized around the energy efficiency optimal target, ensuring the dual rationality of the correction result in the physical sense and the energy efficiency level, and realizing the paradigm shift from "data-driven" to "target-driven"; the multivariate joint drift detection is performed through the Mahalanobis distance, which can capture the abnormal situation of the whole group of measurement points, realize the cooperative correction based on system correlation, and fundamentally overcome the disadvantages of the traditional method of "correcting one and disturbing others"; and based on the operating constraint condition division strategy, a dynamic reference system is established for each typical operating state, so that the drift judgment threshold can be adaptively adjusted with the operating condition, and the adaptability and robustness of the data drift correction in complex industrial scenes are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 A flowchart of the data drift correction method of the disclosure;

[0043] Figure 2 A flowchart of step S2 in the data drift correction method of the disclosure;

[0044] Figure 3 A flowchart of step S3 in the data drift correction method of the disclosure;

[0045] Figure 4 A flowchart of the data drift correction example of the 660MW boiler of the disclosure;

[0046] Figure 5 A module schematic diagram of the data drift correction system of the disclosure;

[0047] Figure 6 A structural schematic diagram of an electronic device of the disclosure. DETAILED DESCRIPTION

[0048] The disclosure will be further described below by way of examples, but the disclosure is not limited to the scope of the examples.

[0049] In the embodiments of the disclosure, prefix words such as "first", "second" are used only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of ordinal words and other prefix words in the embodiments of the disclosure to distinguish the described objects does not limit the described objects, and the description of the described objects in the context of the embodiments should not be construed as redundant limitations because of the use of such prefix words. In addition, in the description of the embodiments, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0050] Boiler combustion process is the core link in energy industry, and its running state is directly related to energy efficiency level, pollutant emission and equipment safety. In actual operation, boiler system relies on a large number of sensors to monitor key detection parameters in real time, such as exhaust gas temperature, oxygen content, air volume, fuel quantity, etc. However, due to factors such as sensor aging, environmental interference, transmission abnormalities, etc., the measured point data often appears systematic drift, which seriously affects the accuracy of energy efficiency evaluation and operation optimization. Therefore, how to effectively identify and correct data drift has become a key technical problem to improve the running efficiency and reliability of the boiler.

[0051] Compared with other industrial data, the power plant boiler data has the particularity of multi-measured point strong coupling, variable working condition and frequent data drift. Among them, multi-measured point strong coupling refers to the mutual influence between temperature, pressure, flow and other parameters in the boiler combustion process. If single measured point correction is performed, it is easy to distort, and the coupling relationship between parameters needs to be considered comprehensively; variable working condition refers to the fact that the boiler often runs under different loads and coal types. The drift correction of static threshold is difficult to adapt to different working conditions, and a dynamic reference needs to be constructed in the data drift correction process to improve the working condition adaptability. Frequent data drift refers to the fact that sensor aging, environmental interference, etc. frequently occur in the power plant scene, and physical rationality correction needs to be combined with the business mechanism of boiler combustion to avoid blind correction.

[0052] For a long time, the industrial field mainly relies on single measured point judgment method with fixed threshold, statistical process control method, and correction method based on mechanism model or data driven model to deal with the problem of data drift. The above three methods have played a role in a certain stage, but all have obvious limitations.

[0053] The single measured point judgment method based on fixed threshold sets an empirical upper and lower limit for each measured point. When the real-time data exceeds this range, it is considered abnormal and is replaced or corrected. It is simple to implement and has small computational complexity, and was widely used in distributed control system (DCS) in the early days. However, it ignores the coupling relationship between multiple measured points. The drift of a certain measured point may trigger a chain of distortion of other parameters. Separate correction may lead to system imbalance. The fixed threshold is difficult to adapt to the dynamic running characteristics of the boiler under different loads and different coal qualities, which may cause false positives or false negatives. In addition, it lacks consideration of the physical mechanism behind data drift, and the correction result often lacks rationality in terms of energy efficiency.

[0054] Compared with the fixed threshold method, the SPC method has certain dynamic adjustment capability and can set reasonable control limits according to the data distribution. However, it is still a univariate analysis method in nature and fails to fully consider the correlation between variables. In the boiler system which is a multivariate and strongly coupled system, the statistical control of a single measuring point cannot cope with the joint drift under complex working conditions, especially in non-steady-state processes such as variable load and variable coal quality. The baseline mean and variance may not be applicable, resulting in a decrease in detection efficiency.

[0055] The correction method based on a mechanism model or a data-driven model calculates abnormal data in reverse by establishing a physical model of boiler combustion (for example, a heat balance equation, a combustion reaction kinetics model, etc.), or constructs a prediction relationship between measuring points by using neural networks, principal component regression, etc. Although this method improves the accuracy of correction to some extent, it is highly dependent on the model. If the model is not updated in time with the device state, the correction effect will deteriorate over time. Moreover, most models are not directly linked to energy efficiency evaluation targets, which may lead to "normal" data in numerical value after correction, but "distortion" in energy efficiency level. In addition, multivariate correction methods (for example, principal component analysis reconstruction) often lack consideration of working condition differentiation, making it difficult to maintain robustness in wide load and multiple coal types in actual scenarios.

[0056] It can be seen that the application of traditional data drift correction methods in the complex thermal system of the boiler always faces common problems such as single-point limitation, static baseline, model decoupling, and energy efficiency inaccuracy, resulting in a lack of multi-measuring-point collaborative analysis capability in boiler data drift correction, poor adaptability to dynamic changes in operating conditions, and disconnection between data drift correction and energy efficiency targets.

[0057] Therefore, the present disclosure provides a data drift correction method, system, device, medium and program product to solve the problems of a lack of multi-measuring-point collaborative analysis capability in existing boiler data drift correction, poor adaptability to dynamic changes in operating conditions, and disconnection between data drift correction and energy efficiency targets.

[0058] Embodiment 1

[0059] In one specific embodiment, a data drift correction method is provided, as shown in Figure 1 The correction method comprises:

[0060] S1, obtaining real-time operating data of a target device including different preset detection parameters;

[0061] S2, determining first running data with data drift and second running data without data drift based on real-time running data, a benchmark reference data set, and an energy efficiency weight corresponding to each preset detection parameter; the benchmark reference data set is pre-constructed based on normal historical running data of different preset detection parameters;

[0062] S3, correcting the first running data by using a preset correction method to obtain corrected running data, and merging the corrected running data and the second running data to obtain third running data;

[0063] S4, obtaining a first energy efficiency score of the benchmark reference data set and a second energy efficiency score of the third running data according to a preset energy efficiency evaluation model;

[0064] S5, in response to a difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, adjusting the energy efficiency weight corresponding to each preset detection parameter, returning to step S2 until the difference is less than or equal to the preset energy efficiency score deviation threshold or a maximum iteration number is reached, and taking the latest obtained third running data as target running data of different preset detection parameters in the target device.

[0065] Specifically, in order to realize data drift correction oriented by energy efficiency target, the anti-balance method can be used to establish an energy efficiency evaluation model M1 of the target device according to the running state of the target device. Taking a boiler combustion device as an example, based on the thermodynamic principle and the actual running data, considering m preset detection parameters such as fuel characteristics, air coefficient, and exhaust gas temperature, an energy efficiency evaluation model M1 is constructed to comprehensively evaluate the boiler combustion state at any time, as shown in formula (1):

[0066] (1)

[0067] wherein, is a feature vector composed of running data corresponding to m preset detection parameters, is running data corresponding to the mth preset detection parameter, and v represents an energy efficiency score.

[0068] The energy efficiency evaluation model M1 outputs an energy efficiency score of 0-100, and the higher the score, the better the efficiency of the boiler combustion.

[0069] In addition, n groups of normal historical running data of the target device in a normal running state are collected to construct a benchmark reference data set as a benchmark reference system for data drift correction. Wherein, is a feature vector composed of normal historical running data corresponding to m preset detection parameters.

[0070] In the correction of data drift, first, through step S1, a set of real-time running data of m preset detection parameters in the target device at a certain time point is collected , which is represented by vector .

[0071] Through step S2, data drift detection is performed on the real-time running data , according to the energy efficiency weight of each preset detection parameter , taking the reference data set D as the reference, the first running data with data drift and the second running data without data drift in the real-time running data are identified, wherein the energy efficiency weight of the i-th preset detection parameter is determined by sensitivity analysis or expert experience, and the energy efficiency weight of the important detection parameter is larger, and the energy efficiency weight of the secondary detection parameter is smaller. For example, the first running data is , represented by vector , and the second running data is , and the vector .

[0072] Step S3 corrects the first running data with data drift to obtain the corrected running data , and combines the corrected running data and the second running data to obtain the third running data corresponding to the m preset detection parameters .

[0073] Through step S4, each in the reference data set D is substituted into formula (1) to calculate the first energy efficiency score of the reference data set , the corrected third running data is substituted into formula (1) to calculate the second energy efficiency score after correction , and the correction effect is verified by formula (2):

[0074] (2)

[0075] Wherein, represents the preset energy efficiency score deviation threshold.

[0076] When the second energy efficiency score satisfies formula (2), it means that the corrected third running data satisfies the energy efficiency target requirement and can be used as the target running data, and the data drift correction is completed; when the second energy efficiency score does not satisfy formula (2), it means that the corrected third running data If the energy efficiency target requirement is not met, the energy efficiency weight corresponding to each preset detection parameter needs to be adjusted through step S5 , and the adjusted energy efficiency weight is used to re-screen the first running data with data drift and the second running data without data drift through step S2 The first running data with data drift is corrected until the third running data after correction satisfies the energy efficiency target requirement.

[0077] By constructing a closed-loop correction system with an energy efficiency evaluation model as a verification benchmark, the final acceptance standard of the entire correction process is that the deviation of the energy efficiency score from the working condition benchmark after correction must be within the allowed range. This "target-oriented" design ensures that the data correction work directly serves the fundamental purpose of improving the accuracy of boiler energy efficiency evaluation, avoids physical distortion caused by correction for correction, and enables the technical achievements to be directly translated into economic benefits of energy saving and consumption reduction.

[0078] The specific embodiment constructs a closed-loop correction system with an energy efficiency evaluation model as a verification benchmark. The final acceptance standard of the entire correction process is that the deviation of the energy efficiency score from the working condition benchmark after correction must be within the allowed range. This "target-oriented" design ensures that the data correction work directly serves the fundamental purpose of improving the accuracy of boiler energy efficiency evaluation, avoids physical distortion caused by correction for correction, and enables the technical achievements to be directly translated into economic benefits of energy saving and consumption reduction. By dynamically adjusting the weight of the preset detection parameter, the correction result is always optimized around the energy efficiency optimal target. This self-verification and continuous optimization mechanism based on the energy efficiency target ensures the effectiveness and reliability of the correction result in engineering practice, as well as the dual rationality in physical meaning and energy efficiency level, realizing the paradigm shift from "data-driven" to "target-driven".

[0079] In a specific embodiment, the reference data set includes a plurality of different working condition data subsets, and different working condition data subsets correspond to different operating conditions.

[0080] As shown in Figure 2 , step S2 includes:

[0081] S21, obtaining the target operating condition corresponding to the real-time running data, and determining the matching target working condition according to the target operating condition;

[0082] S22, determining the first running data and the second running data based on the real-time running data, the target working condition data subset corresponding to the target working condition, and the energy efficiency weight.

[0083] ​​Specifically, in order to enable the drift determination threshold to be adaptively adjusted according to the working condition, the reference data set D is divided into a plurality of working condition data subsets based on different operating conditions, each working condition data subset including a plurality of groups of normal historical operating data under the same or similar operating conditions . For example, for a boiler combustion device, the reference data set D can be divided into k working condition data subsets according to power generation, coal type, etc. , so as to establish an independent statistical feature reference for each typical operating state.

[0084] Therefore, when performing data drift detection on the real-time operating data in step S2, the target working condition data subset that matches the target operating condition of the real-time operating data is first determined according to the target operating condition of the real-time operating data , and then the first operating data in which data drift occurs and the second operating data in which data drift does not occur are identified from the real-time operating data based on the target working condition data subset as a reference according to the energy efficiency weight of each preset detection parameter.

[0085] The specific embodiment establishes a multi-working condition dynamic reference library based on the working condition division strategy of operating constraints, establishes an independent statistical feature reference for each typical operating state through multi-dimensional condition division such as load-coal quality, and enables the reference and dynamic threshold under the current working condition to be automatically matched when performing drift determination and correction, so that the detection and correction standard is always synchronized with the actual operating state of the device, ensures that the drift determination threshold can be adaptively adjusted according to the working condition, and greatly improves the adaptability and robustness in complex industrial scenarios such as wide load operation and frequent fluctuations in coal quality.

[0086] In one specific embodiment, step S22 includes:

[0087] S221, calculating the Mahalanobis distance of the real-time operating data and the normal historical operating data in the target working condition data subset;

[0088] S222, in response to the Mahalanobis distance being greater than a second preset threshold, determining the first operating data and the second operating data based on the real-time operating data, the normal historical operating data in the target working condition data subset, and the energy efficiency weight corresponding to each preset detection parameter.

[0089] S223, in response to the Mahalanobis distance being less than or equal to the second preset threshold, determining that the real-time operating data does not have data drift.

[0090] Specifically, in order to realize the cooperative correction based on system correlation, when the data drift is detected, the Mahalanobis distance dis between the real-time running data and the target working condition data subset is calculated, as shown in formula (3): and the target working condition data subset , as shown in formula (3):

[0091] (3)

[0092] wherein, represents the mean vector of the target working condition data subset , and represents the covariance matrix of the target working condition data subset .

[0093] When dis>dis_th, it is determined that data drift occurs. Wherein, dis_th is determined according to the specific running condition characteristics and historical statistical characteristics. When data drift occurs, the first running data in which the data drift occurs needs to be further determined, and the first running data satisfies formula (4):

[0094] (4) wherein,

[0095] represents the real-time running data corresponding to the jth preset detection parameter, represents the mean value of the normal historical running data corresponding to the jth preset detection parameter in the target working condition data subset , and

[0096] represents the energy efficiency weight corresponding to the jth preset detection parameter, dis_th represents the preset drift threshold. Thus, the first running data in which the data drift occurs and the second running data in which the data drift does not occur are determined. The specific embodiment adopts the Mahalanobis distance to perform multivariate joint drift detection, can capture the abnormal situation of the whole group of measuring points, breaks through the limitation of the traditional single measuring point correction, introduces the weight coefficient based on the energy efficiency sensitivity when identifying the first running data in which the data drift occurs, establishes the differentiated processing mechanism of important detection parameters and secondary detection parameters, and ensures accurate identification of the first running data in which the data drift occurs. In one specific embodiment, as shown in FIG. 3, step S3 includes:

[0097]

[0098] In one specific embodiment, as shown in FIG. 3, step S3 includes: Figure 3

[0099] ​​​​S31, acquire first detection parameters corresponding to the first running data, and second detection parameters corresponding to the second running data;

[0100] S32, based on the reference data set, acquire first normal historical running data corresponding to the first detection parameters, and second normal historical running data corresponding to the second detection parameters, and construct a data drift correction model;

[0101] Step S32 includes: taking the second normal historical running data as input, the first normal historical running data as output, training the preset model to obtain the data drift correction model;

[0102] Wherein, the preset model includes any one of a multiple linear regression model, a support vector regression model, and a neural network learning model;

[0103] S33, input the second running data into the data drift correction model to obtain the corrected running data.

[0104] Specifically, after determining the first running data occurring data drift and the second running data not occurring data drift, the first normal historical running data corresponding to the same detection parameters as the first running data can be selected from the target working condition data subset as label data, the second normal historical running data corresponding to the same detection parameters as the second running data is selected from the target working condition data subset as feature data, the feature data is taken as the input of the preset model, and the label data is taken as the output of the preset model, to construct the data drift correction model M2, as shown in formula (5):

[0105] (5)

[0106] Wherein, is a model error term, and the preset model can be any one of a multiple linear regression model, a support vector regression model, and a neural network learning model.

[0107] After obtaining the data drift correction model M2, the second running data not occurring data drift in the second running data is input into the data drift correction model M2, so that the corrected running data is obtained, and the corrected running data and the second running data are combined to obtain the third running data corresponding to the m preset detection parameters .

[0108] This specific implementation method, based on an anomaly identification mechanism and regression correction model using energy efficiency weights, intelligently calculates the reasonable value of the first operating data that has experienced data drift based on the second operating data that has not drifted. This achieves collaborative correction based on system correlation, fundamentally solving the technical problem of "correcting one, disturbing others." This multi-source data fusion collaborative correction method not only accurately locates specific fault measurement points but also ensures the systematicity and coordination of the entire dataset after correction, thus providing a highly consistent and reliable data foundation for subsequent energy efficiency assessment and optimization control. Furthermore, it organically combines the data drift correction model with the energy efficiency evaluation model of the inverse equilibrium method. The data drift detection model provides the ability to learn complex nonlinear relationships, while the energy efficiency evaluation model built on the first law of thermodynamics ensures the physical rationality of the correction process. This "physical constraint + data learning" architecture avoids the rigidity of pure mechanistic models and overcomes the physical distortion that may occur in pure data models, forming a unique hybrid driving paradigm.

[0109] In a specific example, during the actual operation of a 660MW supercritical coal-fired boiler, factors such as high-temperature and high-pressure environment, long-term sensor aging, and coal quality fluctuations caused systematic drift in key measurement data. These drifts severely affected the accuracy of boiler energy efficiency assessment, thereby impacting operation optimization and energy conservation and emission reduction effects. Figure 4 As shown, this specific example demonstrates data drift correction for this 660MW boiler.

[0110] The 660MW boiler is configured with 32 preset detection parameters as key measuring points, including 8 temperature measuring points (such as flue gas temperature, hot air temperature, inlet and outlet temperatures of each heating surface, etc.), 6 pressure measuring points (such as furnace pressure, pressure of each flue section, etc.), 5 flow measuring points (such as fuel quantity, air volume, steam flow, etc.), 7 component analysis measuring points (such as flue gas oxygen content, CO content, NOx concentration, etc.), and 6 other operating parameters (such as load rate, excess air coefficient, etc.).

[0111] Based on the power generation P and coal quality characteristics, the operating conditions are divided into four main ranges in step S101:

[0112]

[0113] Based on the first law of thermodynamics and the principle of boiler heat balance, an energy efficiency evaluation model M1 is established through step S102:

[0114]

[0115] in, represents the total heat entering the boiler per unit of time, represents the total heat lost per unit of time that fails to be effectively utilized by the boiler.

[0116] After the preset detection parameters, working condition division, and construction of the energy efficiency evaluation model are configured, real-time running data vectors are collected through step S103 , based on the criterion (Rao criterion) to eliminate abnormal values, and data standardization processing is performed on .

[0117] For the target running condition of the real-time running data vector , the target working condition data set is determined as , and the Mahalanobis distance between is calculated , wherein is a 32-dimensional mean vector of , and is a 32x32 covariance matrix of .

[0118] Based on historical statistical analysis, a second preset threshold is set, and when , it is determined that data drift occurs, thereby completing the data drift determination of the real-time running data.

[0119] When data drift occurs, the real-time running data corresponding to the 32 preset detection parameters is analyzed one by one through step S104 according to formula (4), wherein the energy efficiency weight is determined through sensitivity analysis, important detection parameters (such as flue gas temperature and oxygen content) are given a larger weight ( > 1), and secondary detection parameters are given a smaller weight ( < 1). A preset drift threshold = 2.0 is set, and the first running data that has data drift is identified.

[0120] A support vector regression model (SVR) is used to construct a data drift correction model M2 through step S105.

[0121] The second running data is input into M2 through step S106, and the corrected running data is obtained, and the third running data is obtained by combination;

[0122] The correction effect is verified by formula (2) in step S107 to determine the third operation data whether the energy efficiency target requirement is met, if yes, the correction is completed, if not, the energy efficiency weight is adjusted , return to step S104 to re-perform data drift detection and correction, wherein, = 1.5.

[0123] After the correction of the 660 MW supercritical coal-fired boiler, the key indicators are significantly improved, as shown in Table 1, the error of energy efficiency score and the measurement deviation of oxygen content are greatly reduced by about 50%, the error of boiler efficiency evaluation is reduced by 36.4%, and the error of flue gas temperature is also reduced. It can be seen that the data drift correction method provided by the present disclosure effectively improves the accuracy of the measurement data and the reliability of the boiler energy efficiency evaluation, and provides a solid data foundation for operation optimization.

[0124] Table 1

[0125]

[0126] The embodiment constructs a closed-loop correction system with the energy efficiency evaluation model as the core, verifies the energy efficiency score deviation before and after correction through iteration, and dynamically adjusts the weight of the preset detection parameter, so that the correction result is always optimized around the optimal energy efficiency target, ensuring the dual rationality of the correction result in the physical sense and the energy efficiency level, and realizing the paradigm shift from "data-driven" to "target-driven"; through the Mahalanobis distance for multivariate joint drift detection, the abnormal situation of the whole group of measurement points can be captured, realizing the cooperative correction based on system correlation, and fundamentally overcoming the drawbacks of the traditional method of "correcting one and disturbing others"; and based on the working condition division strategy of the operation constraint, a dynamic reference system is established for each typical operation state, so that the drift determination threshold can be adaptively adjusted with the working condition, significantly improving the adaptability and robustness of the data drift correction in complex industrial scenarios.

[0127] Embodiment 2

[0128] In one specific embodiment, a data drift correction system is provided, as shown in Figure 5 The correction system comprises:

[0129] The acquisition module 100 is configured to acquire real-time operation data including different preset detection parameters in the target device;

[0130] The detection module 200 is configured to determine first operation data with data drift and second operation data without data drift based on the real-time operation data, the reference data set, and the energy efficiency weight corresponding to each preset detection parameter; the reference data set is pre-constructed based on normal historical operation data of different preset detection parameters;

[0131] The correction module 300 is configured to correct the first operation data in a preset correction manner to obtain corrected operation data, and combine the corrected operation data and the second operation data to obtain third operation data.

[0132] The verification module 400 is configured to obtain a first energy efficiency score of the benchmark reference data set and a second energy efficiency score of the third operation data according to a preset energy efficiency evaluation model.

[0133] The determination module 500 is configured to, in response to a difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, adjust an energy efficiency weight corresponding to each preset detection parameter, and call the detection module 200 until the difference is less than or equal to the preset energy efficiency score deviation threshold or a maximum iteration number is reached, and take the latest obtained third operation data as target operation data of different preset detection parameters in the target device.

[0134] In a specific embodiment, the correction module 300 is specifically configured to obtain a first detection parameter corresponding to the first operation data and a second detection parameter corresponding to the second operation data, obtain first normal historical operation data corresponding to the first detection parameter and second normal historical operation data corresponding to the second detection parameter based on the benchmark reference data set, construct a data drift correction model, and input the second operation data into the data drift correction model to obtain the corrected operation data.

[0135] In a specific embodiment, the correction module 300 is specifically configured to train a preset model by taking the second normal historical operation data as input and the first normal historical operation data as output to obtain the data drift correction model.

[0136] The preset model includes any one of a multiple linear regression model, a support vector regression model, and a neural network learning model.

[0137] In a specific embodiment, the benchmark reference data set includes a plurality of different working condition data subsets, and the different working condition data subsets correspond to different operation conditions.

[0138] The detection module 200 is specifically configured to obtain a target operation condition corresponding to real-time operation data, and determine a matched target working condition according to the target operation condition, and determine the first operation data and the second operation data based on the real-time operation data, a target working condition data subset corresponding to the target working condition, and an energy efficiency weight.

[0139] In a specific implementation, the detection module 200 is specifically configured to calculate a Mahalanobis distance of the real-time operation data and the normal historical operation data in the target working condition data subset; in response to the Mahalanobis distance being greater than a second preset threshold, determine the first operation data and the second operation data based on the real-time operation data, the normal historical operation data in the target working condition data subset, and the energy efficiency weight corresponding to each preset detection parameter; or, in response to the Mahalanobis distance being less than or equal to the second preset threshold, determine that none of the real-time operation data has data drift.

[0140] In a specific implementation, the first operation data satisfies the following formula:

[0141] ;

[0142] wherein, denotes the real-time operation data corresponding to the jth preset detection parameter, denotes the mean value of the normal historical operation data corresponding to the jth preset detection parameter in the target working condition data subset, denotes the energy efficiency weight corresponding to the jth preset detection parameter, denotes a preset drift threshold.

[0143] For the system embodiment, since it basically corresponds to the method embodiment, the relevant part can be referred to the part of the method embodiment. The system embodiment described above is only illustrative, wherein the units described as separate components can or can not be physically separated, and the components of the unit can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present disclosure according to actual needs.

[0144] The embodiment constructs a closed-loop correction system with the energy efficiency evaluation model as the core, verifies the energy efficiency score deviation before and after correction through iteration, and dynamically adjusts the weight of the preset detection parameter, so that the correction result is always optimized around the energy efficiency optimal target, ensuring the dual rationality of the correction result in the physical sense and the energy efficiency level, and realizing the paradigm shift from "data-driven" to "target-driven"; the Mahalanobis distance is used for multivariate joint drift detection, which can capture the abnormal situation of the whole group of measuring points, realize the collaborative correction based on system correlation, and fundamentally overcome the disadvantages of the traditional method of "correcting one and disturbing others"; and based on the working condition division strategy of operation constraints, a dynamic reference system is established for each typical operation state, so that the drift judgment threshold can be adaptively adjusted with the working condition, and the adaptability and robustness of the data drift correction in complex industrial scenes are significantly improved.

[0145] Embodiment 3

[0146] Figure 6 This is a schematic diagram of the structure of an electronic device according to an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the data drift correction method described in any of the above embodiments. Figure 6 The electronic device 30 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0147] like Figure 6 As shown, the electronic device 30 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0148] Bus 33 includes a data bus, an address bus, and a control bus.

[0149] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0150] The memory 32 may also include a program tool 325 (or utility) having a set (at least one) program module 324, such program module 324 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0151] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the data drift correction method provided in any of the above embodiments.

[0152] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed through input / output (I / O) interface 35. Furthermore, electronic device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. As shown, network adapter 36 communicates with other modules of electronic device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0153] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, such division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into units / modules embodied by multiple units / modules.

[0154] Embodiment 4

[0155] The embodiments of the present disclosure further provide a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the data drift correction method of any of the above embodiments.

[0156] More specifically, the readable storage medium can include, but is not limited to, a portable disc, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0157] Embodiment 5

[0158] The embodiments of the present disclosure further provide a computer program product, comprising a computer program, which, when executed by a processor, implements the data drift correction method of any of the above embodiments.

[0159] The program code for carrying out the computer program product of the present disclosure can be written in any combination of one or more programming languages, and can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0160] Although the specific embodiments of the present disclosure are described above, those skilled in the art should understand that this is only an illustration, and the protection scope of the present disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications all fall within the protection scope of the present disclosure.

Claims

1. A method for correcting data drift, characterized in that, The correction method includes: Acquire real-time operating data from the target device, including different preset detection parameters; Based on the real-time operating data, the benchmark reference dataset, and the energy efficiency weight corresponding to each preset detection parameter, the first operating data in which data drift occurred and the second operating data in which data drift did not occur are determined; the benchmark reference dataset is pre-constructed based on normal historical operating data with different preset detection parameters; The first running data is corrected using a preset correction method to obtain corrected running data, and the corrected running data and the second running data are merged to obtain the third running data; The first energy efficiency score of the benchmark reference dataset and the second energy efficiency score of the third operating data are obtained according to the preset energy efficiency evaluation model. In response to the difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, the energy efficiency weight corresponding to each preset detection parameter is adjusted, and the step of determining the first operating data with data drift and the second operating data without data drift based on the real-time operating data, the benchmark reference dataset, and the energy efficiency weight corresponding to each preset detection parameter is returned to be executed until the difference is less than or equal to the preset energy efficiency score deviation threshold or the maximum number of iterations is reached, and the latest obtained third operating data is used as the target operating data for different preset detection parameters in the target device.

2. The correction method according to claim 1, characterized in that, The step of correcting the first operating data using a preset correction method to obtain corrected operating data includes: Obtain the first detection parameter corresponding to the first running data, and the second detection parameter corresponding to the second running data; Based on the benchmark reference dataset, first normal historical operation data corresponding to the first detection parameter and second normal historical operation data corresponding to the second detection parameter are obtained to construct a data drift correction model. The second running data is input into the data drift correction model to obtain the corrected running data.

3. The correction method according to claim 2, characterized in that, The step of acquiring first normal historical operating data corresponding to the first detection parameter and second normal historical operating data corresponding to the second detection parameter based on the benchmark reference dataset, and constructing a data drift correction model includes: Using the second normal historical operating data as input and the first normal historical operating data as output, the preset model is trained to obtain the data drift correction model; The preset model includes any one of the following: multiple linear regression model, support vector regression model, and neural network learning model.

4. The correction method according to any one of claims 1 to 3, characterized in that, The benchmark dataset includes several different subsets of operating condition data, with each subset corresponding to different operating conditions. The step of determining the first operating data that experienced data drift and the second operating data that did not experience data drift based on the real-time operating data, the benchmark reference dataset, and the energy efficiency weight corresponding to each preset detection parameter includes: Obtain the target operating conditions corresponding to the real-time operating data, and determine the matching target operating conditions based on the target operating conditions; Based on the real-time operating data, the target operating condition data subset corresponding to the target operating condition, and the energy efficiency weight, the first operating data and the second operating data are determined.

5. The correction method according to claim 4, characterized in that, The step of determining the first operating data and the second operating data based on the real-time operating data, the subset of target operating condition data corresponding to the target operating condition, and the energy efficiency weight includes: Calculate the Mahalanobis distance between the real-time operating data and the normal historical operating data in the target operating condition data subset; In response to the Mahalanobis distance being greater than a second preset threshold, the first operating data and the second operating data are determined based on the real-time operating data, the normal historical operating data in the target operating condition data subset, and the energy efficiency weight corresponding to each preset detection parameter; or, In response to the Mahalanobis distance being less than or equal to a second preset threshold, it is determined that no data drift has occurred in the real-time running data.

6. The correction method according to claim 5, characterized in that, The first running data satisfies the following formula: ； in, This represents the real-time running data corresponding to the j-th preset detection parameter. This represents the mean of the normal historical operating data corresponding to the j-th preset detection parameter in the target operating condition data subset. This represents the energy efficiency weight corresponding to the j-th preset detection parameter. This indicates the preset drift threshold.

7. A data drift correction system, characterized in that, The correction system includes: The acquisition module is used to acquire real-time operating data from the target device, including different preset detection parameters; The detection module is used to determine the first operating data that has experienced data drift and the second operating data that has not experienced data drift based on the real-time operating data, the benchmark reference dataset, and the energy efficiency weight corresponding to each preset detection parameter; the benchmark reference dataset is pre-constructed based on normal historical operating data with different preset detection parameters; The calibration module is used to calibrate the first running data using a preset calibration method to obtain calibrated running data, and to merge the calibrated running data and the second running data to obtain third running data; The verification module is used to obtain the first energy efficiency score of the benchmark reference dataset and the second energy efficiency score of the third running data according to the preset energy efficiency evaluation model. The determination module is used to adjust the energy efficiency weight corresponding to each preset detection parameter in response to the difference between the first energy efficiency score and the second energy efficiency score being greater than a preset energy efficiency score deviation threshold, and return to call the detection module until the difference is less than or equal to the preset energy efficiency score deviation threshold or the maximum number of iterations is reached, and the latest obtained third operating data is used as the target operating data for different preset detection parameters in the target device.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the data drift correction method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the data drift correction method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the data drift correction method according to any one of claims 1 to 6.

Images