A method, system, device and medium for identifying gross errors in power plant system measurement

Abstract

The present invention belongs to the technical field of power plant gross error identification and discloses a method, system, device, and medium for identifying gross errors in power plant system measurements. The method comprises the following steps: obtaining power plant system measurement data for which gross errors are to be identified; and performing predictions based on the obtained power plant system measurement data using a trained prediction model for the gross errors of the power plant system to obtain a gross error identification result. This invention specifically provides a method for intelligently identifying gross errors in power plant system measurements based on a multi-classification support vector machine (SVM). The method is capable of quickly and accurately identifying gross errors in the system and directly determining the location of fault points.

Description

Technical Field

[0001] The present invention belongs to the technical field of gross error identification in power plants, and in particular relates to a method, system, equipment and medium for identifying gross errors in power plant system measurements. Background Art

[0002] Reliable power plant system operation relies on precise measurement data. Accurate thermal parameters are the prerequisite and guarantee for the safe and economical operation of the entire power plant. These measurement data and thermal parameters are typically measured by sensor instruments within the power plant system. In practice, measurement data can exhibit a certain degree of deviation due to factors such as instrument accuracy, operator operational standards, and instrument failures. Factors such as instrument failures and sensor malfunctions can cause measured values ​​to deviate significantly from their true values, resulting in significant errors. The measured data cannot truly reflect the operating status of the power plant system, seriously impacting the safety and economic efficiency of the power plant system.

[0003] At present, the existing gross error identification methods can be roughly divided into three categories: the first category is the manual inspection method, which is to check the factors that may cause gross errors one by one; the second category is the hardware redundancy method, which is to use different measurement methods for the same process to make a comprehensive judgment on gross errors; the third category is the statistical hypothesis testing method, which is to conduct gross error analysis and testing based on the statistical characteristics of the measurement system, including the overall test method, the constraint equation test method, the measurement data test method, the sequential elimination method, etc.

[0004] To further explain, manual inspection methods are often only applicable to simple systems. For systems with strong coupling, it is difficult to identify factors that cause gross errors, making their applicability very poor. Hardware redundancy methods have high requirements for cost expenditure and on-site environment, and can only be used for a few important measurement points, but are difficult to apply to the entire system. Among statistical hypothesis testing methods, basic detection methods such as the overall test method and the constraint equation test method are less effective in identifying gross errors. They can usually only determine whether a system has gross errors but cannot accurately locate them. While combination methods such as the sequential elimination method can locate gross errors, their implementation process is relatively complex, and their identification efficiency and accuracy are low when used in complex systems. Furthermore, the third type of method is based on statistics and has high requirements for the number of measurements. When the measurement data is small, its reliability cannot be guaranteed, which will pose a significant challenge to the analysis and decision-making of operation and maintenance personnel. Summary of the Invention

[0005] The present invention aims to provide a method, system, device, and medium for identifying gross errors in power plant system measurements to address one or more of the aforementioned technical problems. Specifically, the present invention provides an intelligent method for identifying gross errors in power plant system measurements based on a multi-classification support vector machine (SVM). This method can quickly and accurately identify gross errors in the system and directly determine the location of the fault point.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A first aspect of the present invention provides a method for identifying gross errors in power plant system measurements, comprising the following steps:

[0008] Obtaining power plant system measurement data for gross error identification;

[0009] Based on the acquired measurement data of the power plant system, a prediction is performed using a trained measurement gross error prediction model of the power plant system to obtain a power plant system measurement gross error identification result;

[0010] The training steps of the trained measurement gross error prediction model of the power plant system include:

[0011] Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system;

[0012] The obtained training sample set is used to train the pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model includes multiple independent multi-classification support vector machines, the number of the multi-classification support vector machines is the same as the number of measurement points of the power plant system, and each multi-classification support vector machine is used to correspond to a measurement point of the power plant system; during training, the training is completed by constructing a dual problem and solving it for optimization.

[0013] A further improvement of the method of the present invention is that in the step of obtaining the power plant system measurement data to be measured for gross error identification,

[0014] The power plant system measurement data includes measurement values ​​of all to-be-identified measurement points in the power plant system.

[0015] A further improvement of the method of the present invention is that in the pre-constructed measurement gross error prediction model,

[0016] The multi-classification support vector machine uses the Gaussian RBF kernel for calculation, and the expression is,

[0017]

[0018] Where K(x,x') is the kernel function of the Gaussian RBF kernel, e is the natural logarithm base, γ is a parameter used to reflect the influence of a single sample on the entire classification hyperplane, x and x' represent two different input feature vectors, ‖x-x'‖ 2 Represents the Euclidean distance between two different input feature vectors; x=(x1,x2,…,x n ), represents the measurement data of all measuring points under different working conditions;

[0019] For the j-th multi-classification support vector machine, the optimization problem is expressed as:

[0020]

[0021] Where C is the penalty factor; φ(x i ) is the feature x i Data generated by kernel function mapping; ji is a parameter used to reflect the influence of soft margin; i is the label indicator value; ω is the coefficient of the decision boundary feature related term; b is the offset.

[0022] A further improvement of the method of the present invention is that, during the training, the step of constructing a dual problem and solving it for optimization to complete the training specifically includes:

[0023] For the j-th multi-class support vector machine, the Lagrange multiplier α is introduced into the optimization problem j ={α j1 ,α j2 ,…,α jn} and μ j ={μ j1 ,μ j2 ,…,μ jn} then form the dual problem, which can be expressed as,

[0024]

[0025] Among them, the Lagrange multiplier satisfies the KKT condition, which is expressed as,

[0026]

[0027] Where, α ji is the Lagrange multiplier α in the j-th SVM j The i-th component of ji represents the hinge loss function.

[0028] A further improvement of the method of the present invention is that, during the training, in the step of optimizing the training by constructing a dual problem and solving it,

[0029] The dual problem is solved using the numerical method of quadratic convex optimization problem.

[0030] A further improvement of the method of the present invention is that, during the training, in the step of optimizing the training by constructing a dual problem and solving it,

[0031] The sequential minimization method is used to optimize α; where,

[0032] First, initialize α with a random number and calculate the offset b, initializing the error term E i ,i=1,2,…,n;

[0033] Then, each time a pair of variables (α p ,α q ), the remaining parameters {α1,…,α p-1 ,α p+1 ,…,α q-1 ,α q+1 ,…,α n} is fixed and treated as a constant;

[0034] According to the constraints Get α q The update expression of is:

[0035]

[0036] Where, E' p =f(x p )-y p is α p The error between the corresponding predicted value and the true value; E' q =f(x q )-y q is α q The error between the corresponding predicted value and the true value;

[0037] η=K(x p ,x p )+K(x q ,x q )-2K(x p ,x q );

[0038] The expression is,

[0039]

[0040] The update expression of b is,

[0041]

[0042] In the iterative process, each parameter α in α is continuously updated according to the above update rules i , until all items in α meet the KKT conditions, completing the training and optimization of the model.

[0043] A further improvement of the method of the present invention is that in the step of obtaining the identification result of the gross error of the power plant system measurement,

[0044] The obtained power plant system measurement gross error identification result includes a label indication value corresponding to the measurement data of each measurement point, and the label indication value is used to indicate whether the measurement data of the measurement point is a gross error or a normal value.

[0045] A second aspect of the present invention provides a system for identifying gross errors in power plant system measurements, comprising:

[0046] A data acquisition module, used to acquire the power plant system measurement data to be measured for gross error identification;

[0047] an identification module, configured to perform prediction based on the acquired measurement data of the power plant system using a trained prediction model for gross measurement errors of the power plant system to obtain a gross measurement error identification result of the power plant system;

[0048] The training steps of the trained measurement gross error prediction model of the power plant system include:

[0049] Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system;

[0050] The obtained training sample set is used to train the pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model includes multiple independent multi-classification support vector machines, the number of the multi-classification support vector machines is the same as the number of measurement points of the power plant system, and each multi-classification support vector machine is used to correspond to a measurement point of the power plant system; during training, the training is completed by constructing a dual problem and solving it for optimization.

[0051] A third aspect of the present invention provides an electronic device, comprising:

[0052] at least one processor; and,

[0053] a memory communicatively connected to the at least one processor; wherein,

[0054] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method for identifying gross errors in power plant system measurements as described in any one of the first aspects of the present invention.

[0055] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for identifying gross errors in power plant system measurements as described in any one of the first aspects of the present invention.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The method for identifying gross errors in power plant system measurements provided by the present invention performs predictions based on a trained gross error prediction model for the power plant system. The gross error prediction model includes multiple independent multi-classification support vector machines. The number of multi-classification support vector machines is the same as the number of measurement points in the power plant system. Each multi-classification support vector machine is used to correspond to a measurement point in the power plant system. The multi-classification support vector machine can accurately identify and detect the data at each measurement point, achieving accurate and reliable discrimination and location of gross errors, and achieving fast inference speed. Compared with statistical methods, multi-classification support vector machines can model and extract features of implicit relationships between measurement point variables, thereby exploring richer characteristics. Single-point inference does not require statistical analysis of data and has strong adaptability.

[0058] In the present invention, since the prediction model for each measuring point in the multi-classification support vector machine model is trained independently, the diagnostic method constructed by the present invention will not cause errors to spread to other measuring points due to failures of individual measuring points, unlike statistical methods, thereby causing misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art; obviously, the drawings described below are some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0060] Figure 1 This is a flow chart of a method for identifying gross errors in power plant system measurements provided by an embodiment of the present invention;

[0061] Figure 2 1 is a schematic diagram of a training process of a measurement gross error prediction model according to an embodiment of the present invention;

[0062] Figure 3 Schematic diagram of a system for identifying gross errors in power plant system measurements provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0064] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0065] The present invention is described in further detail below with reference to the accompanying drawings:

[0066] See also Figure 1 The embodiment of the present invention provides a method for identifying gross errors in power plant system measurements, which specifically includes the following steps:

[0067] Step 1: obtaining power plant system measurement data for gross error identification; explanatory, the power plant system measurement data includes measurement values ​​of all measurement points in the power plant system;

[0068] Step 2: Based on the power plant system measurement data obtained in step 1, a trained measurement gross error prediction model of the power plant system is used to perform prediction and identification to obtain a power plant system measurement gross error identification result;

[0069] The training steps of the trained measurement gross error prediction model of the power plant system include:

[0070] Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system;

[0071] The obtained training sample set is used to train a pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model is a multi-classification support vector machine model; the multi-classification support vector machine model includes the same number of support vector machines as the measuring points of the power plant system, and each support vector machine is used to correspond to a measuring point of the power plant system; during training, the training is completed by constructing a dual problem and using a sequential minimization method for iterative calculation.

[0072] See also Figure 2 In an embodiment of the present invention, the training and acquisition steps of the intelligent prediction model for identifying and detecting gross errors in power plant system measurements based on the SVM algorithm include:

[0073] Step 1: Based on the data set containing gross measurement errors in the real power plant system, a sample set is constructed;

[0074] An exemplary embodiment may include the following steps:

[0075] From the measurement data set of the power plant system, data of n measurement conditions are selected as the data set T of the multi-classification support vector machine, which is expressed as follows:

[0076] T=(T1,T2,...,T i ,...,T n );

[0077] T i ={(x1,y1),(x2,y2),...,(x j ,y j ),...,(x m ,y m )}, i=1,2,...n; y j ∈{+1,-1}, j=1,2,...m;

[0078] Where x j is the jth eigenvector, and its physical meaning is a series of measurement data of the jth measurement point in the power plant system; j is the gross error criterion, the exemplary y j =1 means the data is a gross error, y j =-1 means the data is not a gross error; T i is the measurement data of the i-th measurement condition.

[0079] In the preferred technical solution of the embodiment of the present invention, when selecting the data set, it should be ensured that the gross error ratio of each measuring point is basically the same and not less than 30% of the total data of the measuring point; in addition, the data set T is divided into training samples T according to the ratio of 10% to 20% of the test samples. train =(T1,T2,…,T n' ), n'<n and test sample T test =(T n'+1 ,T n'+2 ,…,T n ) two parts; test sample T train Used to train the measurement gross error prediction model, validation set T test Used to test the training effect of the model.

[0080] Step 2: There are m measurement points in the power plant system, so m SVMs are initially constructed; among them,

[0081] The specific exemplary steps include:

[0082] In order to clearly identify which measurement points among the m measurement points are gross errors, it is necessary to construct m independent SVMs to achieve one-to-many multi-classification; among them, the j-th SVM is used to determine whether the input data belongs to the measurement point with ID number j. Its input data is x=(x1,x2,…,x n ), which represents the measurement data of all measuring points under different working conditions; its output data is y=(y1,y2,…,y n ), where y i ∈{-1,1}, i=1,2,…,n, if the measured data is within the normal value range of the measuring point, the output is -1; if the measured data is not within the normal range of the measuring point, that is, the data is an error or the value of another measuring point, the data is considered to be a gross error and the output is 1.

[0083] Step 3: Select an appropriate kernel function and define each SVM in detail;

[0084] The specific exemplary steps include:

[0085] The principle of SVM is to use hyperplane f(x)=ω T x+b is used to classify the data into two categories. For the problem of predicting gross errors in power plant system measurement data, it is usually difficult to use a linear classifier due to the complexity of the data set. Therefore, it is necessary to use a kernel method to convert the nonlinear problem into a linear classification problem. For the nonlinear SVM algorithm, the technical solution of the embodiment of the present invention uses a Gaussian RBF kernel for calculation, and its mathematical expression is:

[0086]

[0087] Among them, K(x,x') is the kernel function of the Gaussian RBF kernel, e is the natural logarithm base, the parameter γ mainly reflects the influence of a single sample on the entire classification hyperplane, x and x' represent two different input feature vectors, ‖x-x'‖ 2 represents the Euclidean distance between these two vectors.

[0088] For the j-th SVM, the optimization problem is:

[0089]

[0090] Among them, C is the penalty factor, φ(x i ) is the feature x i The data generated by kernel function mapping, ξ ji It reflects the influence of the soft margin, that is, the degree of error allowed in the classification of each data point, ω is the coefficient of the decision boundary feature-related term, and b is the offset.

[0091] Step 4: Optimize the hyperparameters of SVM;

[0092] The specific exemplary steps include:

[0093] Before using the RBF kernel to classify nonlinear data, it is necessary to select the values ​​of the two parameters γ and C in step 3. These two parameters are called hyperparameters, and their values ​​directly affect the final classification effect;

[0094] For the parameter γ, when γ is relatively small, the influence of a single sample on the entire classification hyperplane is relatively far, and it is easy to be selected as a support vector. Conversely, when γ is relatively large, the influence of a single sample on the entire classification hyperplane is relatively close, and it is not easy to be selected as a support vector. The support vectors of the entire model will also be fewer, and the model will become more complex; for the parameter C, when C is relatively large, the loss function will also be larger, the influence of farther outliers on the classifier will be greater, the model of support vectors and hyperplanes will become more complex, and it will also be easy to overfit. Conversely, when C is relatively small, the influence of farther outliers on the classifier will be smaller, and the model of support vectors and hyperplanes will also be simpler.

[0095] Hyperparameter optimization refers to finding the hyperparameter value combination that maximizes model performance. The optimization method for the hyperparameters of the SVM model is as follows: First, the values ​​of C and γ are selected to determine the parameter combination for grid search. For example, the value range of C is selected as: {0.1, 1, 10, 100, 1000}, and the value range of γ is selected as: {0.1, 0.2, 0.3, 0.4, 0.5}. In this way, 25 parameter combinations are determined and grid search is performed; then the cross-validation fold k is defined, that is, the training data is divided into k blocks, of which k-1 blocks are used for training and 1 block is used to evaluate the performance of the test model. Each evaluation can obtain an evaluation index E i , and finally take the average value of the model evaluation index According to the evaluation index of each parameter, the final parameter combination is selected as the result of hyperparameter optimization.

[0096] Step 5: Solve the optimization problem and train the model; specifically, the following steps are included:

[0097] Solving the soft margin nonlinear SVM usually utilizes the duality of its optimization problem. For the j-th SVM, the Lagrange multiplier α={α1,α2,…,α n} and μ={μ1,μ2,…,μ n}Then there is the following dual problem:

[0098]

[0099] The constraints of the dual problem contain inequality relations, and the condition for the existence of a local optimal solution is that the Lagrange multiplier satisfies the KKT condition, that is:

[0100]

[0101] Among them, L ji represents the hinge loss function.

[0102] The dual problem can be solved by using the numerical method of quadratic convex optimization problem. The present invention uses the sequential minimization method (SMO) to optimize α. The basic idea is:

[0103] First, initialize α with a random number and calculate the offset b, initializing the error term E i ,i=1,2,…,n;

[0104] Then, each time a pair of variables (α p ,α q ), the remaining parameters {α1,…,α p-1 ,α p+1 ,…,α q-1 ,α q+1 ,…,α n}Fixed and treated as a constant.

[0105] According to the constraints We can deduce α q Get the updated expression:

[0106]

[0107] Where, E' p =f(x p )-y p is α p The error between the corresponding predicted value and the true value;

[0108] E' q =f(x q )-y q is α q The error between the corresponding predicted value and the true value;

[0109] η=K(x p ,x p )+K(x q ,x q )-2K(x p ,x q ).

[0110] The expression can be expressed by We get:

[0111]

[0112] For the update of b, we have:

[0113]

[0114] During the iteration process, each parameter α in α is continuously updated according to the above rules. i , until all items in α meet the KKT conditions, the model training and optimization can be completed.

[0115] Step 6: Use the test sample to test the trained model;

[0116] A specific example involves training a model for predicting gross errors in power plant system measurements. The model's effectiveness is then tested using test samples. If the predicted accuracy is above 95%, the training parameters are saved, completing the model. Otherwise, some parameters are adjusted and the model is retrained.

[0117] In an embodiment of the present invention, the steps of using an intelligent prediction model to predict gross errors include: completing the construction and training of an intelligent prediction model for gross errors in power plant system measurements through the above steps; using this model to predict the measurement data of the power plant system; if the output value of a measurement point is 1, the measurement point is a fault point and the measurement data has a gross error; if the output value of the measurement point is -1, the point is a normal point and has no gross error. Thus, the construction of an intelligent prediction model for gross errors in power plant system measurements based on a multi-classification SVM algorithm is completed, and this model is used to analyze and predict the power plant system measurement data, classifying fault points and non-fault points.

[0118] The multi-classification SVM-based power plant system measurement gross error prediction model proposed in the embodiment of the present invention is based on a gross error test dataset from a real power plant system. It can be applied to various complex power plant systems, accurately identifying and detecting data at each measurement point, achieving the discrimination and location of gross errors with accuracy, reliability, and fast reasoning speed. Compared with statistical methods, the multi-classification SVM model can model the implicit relationships between measurement point variables and extract features, thereby mining richer characteristics; single-point reasoning does not require data statistics and has strong adaptability. In addition, because the prediction model for each measurement point in the multi-classification SVM model is trained independently, the diagnostic method constructed by the present invention does not cause errors to spread to other measurement points due to failures at individual measurement points, thereby causing misjudgments, as in statistical methods. In summary, the method proposed by the present invention is based on the measurement data of a real power plant system, standardizes the data samples for model construction, establishes multiple SVMs based on the Gaussian RBF kernel method based on the number of measurement points, and uses the SMO training model to ultimately transform the complex problem of identifying and locating gross errors in power plant system measurements into a simple SVM multi-classification problem.

[0119] The following are device embodiments of the present invention, which can be used to perform the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0120] See also Figure 3 , an embodiment of the present invention provides a system for identifying gross errors in power plant system measurements, comprising:

[0121] A data acquisition module, used to acquire the power plant system measurement data to be measured for gross error identification;

[0122] an identification module, configured to perform prediction based on the acquired measurement data of the power plant system using a trained prediction model for gross measurement errors of the power plant system to obtain a gross measurement error identification result of the power plant system;

[0123] The training steps of the trained measurement gross error prediction model of the power plant system include:

[0124] Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system;

[0125] The obtained training sample set is used to train the pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model includes multiple independent multi-classification support vector machines, the number of the multi-classification support vector machines is the same as the number of measurement points of the power plant system, and each multi-classification support vector machine is used to correspond to a measurement point of the power plant system; during training, the training is completed by constructing a dual problem and solving it for optimization.

[0126] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the method for identifying gross errors in power plant system measurement.

[0127] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device, used to store programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides storage space, which stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by a processor. These instructions may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium herein may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor may load and execute the one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the method for identifying gross errors in power plant system measurements in the above-mentioned embodiment.

[0128] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for identifying gross errors in power plant system measurements, characterized in that: The following steps are involved: Obtaining power plant system measurement data for gross error identification; Based on the acquired measurement data of the power plant system, a prediction is performed using a trained measurement gross error prediction model of the power plant system to obtain a power plant system measurement gross error identification result; The training steps of the trained measurement gross error prediction model of the power plant system include: Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system; The obtained training sample set is used to train the pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model includes multiple independent multi-classification support vector machines, the number of the multi-classification support vector machines is the same as the number of measurement points of the power plant system, and each multi-classification support vector machine is used to correspond to a measurement point of the power plant system; during training, the training is completed by constructing a dual problem and solving it for optimization.

2. A method for identifying gross errors in power plant system measurements according to claim 1, characterized in that: In the step of obtaining the power plant system measurement data to be measured for gross error identification, The power plant system measurement data includes measurement values ​​of all to-be-identified measurement points in the power plant system.

3. The method for identifying gross errors in power plant system measurements according to claim 1, characterized in that: In the pre-built measurement gross error prediction model, The multi-classification support vector machine uses the Gaussian RBF kernel for calculation, and the expression is, ； Where, is the kernel function of the Gaussian RBF kernel, is the base of natural logarithm, is a parameter used to reflect the influence of a single sample on the entire classification hyperplane. 、 represents two different input feature vectors, Represents the Euclidean distance between two different input feature vectors; , represents the measurement data of all measuring points under different working conditions; For the The optimization problem of a multi-class support vector machine is expressed as: ； Where, is the penalty factor; It is a feature Data generated by kernel mapping; is a parameter used to reflect the influence of soft margin; Indicates the value for the label; is the coefficient of the decision boundary feature related term; is the offset.

4. A method for identifying gross errors in power plant system measurements according to claim 3, characterized in that: During the training, the steps of constructing a dual problem and solving it for optimization to complete the training specifically include: For the first Multi-class support vector machines, introducing Lagrange multipliers in the optimization problem and Then the dual problem is formed, which is expressed as, ； Among them, the Lagrange multiplier satisfies the KKT condition, which is expressed as, ； Where, For the j Lagrange multipliers in SVM No. i Quantity represents the hinge loss function.

5. A method for identifying gross errors in power plant system measurements according to claim 4, characterized in that: During the training, the training is completed by constructing a dual problem and solving it for optimization. The dual problem is solved using the numerical method of quadratic convex optimization problem.

6. A method for identifying gross errors in power plant system measurements according to claim 4, characterized in that: During the training, the training is completed by constructing a dual problem and solving it for optimization. Optimize using sequential minimization ;in, First, the random number is initialized And calculate the offset , initialize the error term ; Next, we select a pair of variables at a time , the remaining parameters Fixed and treated as a constant; According to the constraints , get The update expression of is: ； Where, for The error between the corresponding predicted value and the true value; for The error between the corresponding predicted value and the true value; ； The expression is, ； The update expression of is: ； ； During the iteration process, the update expression is continuously updated according to the update expression Each parameter in ,until All items in satisfy the KKT conditions, completing the training and optimization of the model.

7. The method for identifying gross errors in power plant system measurements according to claim 1, characterized in that: In the step of obtaining the identification result of the gross error of the power plant system measurement, The obtained power plant system measurement gross error identification result includes a label indication value corresponding to the measurement data of each measurement point, and the label indication value is used to indicate whether the measurement data of the measurement point is a gross error or a normal value.

8. A system for identifying gross errors in power plant system measurements, characterized in that: include: A data acquisition module, used to acquire the power plant system measurement data to be measured for gross error identification; an identification module, configured to perform prediction based on the acquired measurement data of the power plant system using a trained prediction model for gross measurement errors of the power plant system to obtain a gross measurement error identification result of the power plant system; The training steps of the trained measurement gross error prediction model of the power plant system include: Obtaining a training sample set; wherein each training sample in the training sample set includes an ID, measurement data, and a gross error label of each measurement point in the power plant system; The obtained training sample set is used to train the pre-constructed measurement gross error prediction model to obtain a trained measurement gross error prediction model of the power plant system; wherein the pre-constructed measurement gross error prediction model includes multiple independent multi-classification support vector machines, the number of the multi-classification support vector machines is the same as the number of measurement points of the power plant system, and each multi-classification support vector machine is used to correspond to a measurement point of the power plant system; during training, the training is completed by constructing a dual problem and solving it for optimization.

9. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for identifying gross errors in power plant system measurement according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying gross errors in power plant system measurements according to any one of claims 1 to 7 is implemented.

Images