A data intelligent association method and system for new energy monitoring

Abstract

The application discloses a kind of data intelligent association methods and systems for new energy monitoring, it is related to new energy monitoring technical field, including obtaining associated signal point information as sample data set, feature processing is carried out by natural language processing technology;Correlation signal point recommendation model is constructed, and correlation signal point recommendation model training is carried out;When signal point association configuration operation, automatically recommend to be associated signal point, and the selection result of configuration personnel is recorded into history base;Based on the updating strategy of quantitative trigger and sliding window, with historical recommendation matching record constitutes rolling data window, when the proportion of new sample reaches predetermined proportion, trigger model retraining;After completing the association of all signal points of single unit or device, select and export pick point template.The application can intelligently recommend signal point, shorten the working time of manual pick point, reduce the probability of error operation, improve configuration accuracy and configuration efficiency, and reusability is strong.

Description

Technical Field

[0001] This invention relates to the field of new energy monitoring technology, and in particular to a data intelligent association method and system for new energy monitoring. Background Technology

[0002] With the accelerated advancement of China's low-carbon energy transition, the installed capacity of wind and photovoltaic power generation has increased significantly. Simultaneously, due to the development and promotion of automation and information technologies in the new energy field, a single new energy power plant often deploys more than 20 systems with various functions and applications. In current practical applications, each system has its own hardware and software systems, resulting in numerous "information silos" within the power plant. Furthermore, many advanced application systems require data from the integrated monitoring system within the plant to achieve their own functions. For example, within a wind farm, wind power prediction systems, wind turbine energy efficiency analysis and fault early warning systems, active support control systems, auxiliary control equipment monitoring systems, wind farm operation and maintenance systems, and the information systems of each wind turbine manufacturer all need to interact with the wind farm's integrated monitoring system.

[0003] There is a need for information exchange between the various systems in a renewable energy power plant. Therefore, depending on the specific application, various engineering configuration tasks arise during the system deployment and commissioning phases, such as the selection of signal point associations between systems. For renewable energy power plants, the number of signal point accesses on the monitoring platform is enormous. For example, a 300MW photovoltaic power plant has no fewer than 200,000 access points, while a wind farm of the same capacity, although varying depending on the type of wind turbine, typically has 50,000 access points. Furthermore, when configuring signal point associations between systems, the same signal point is named differently in different systems. Currently, the renewable energy field lacks a standard similar to IEC 61850 in the power transmission and transformation field, which allows for unified SCD file parsing and configuration. The work of associating signal points at new energy power plants often relies on engineers to manually select and configure points, which involves a huge amount of work. Taking the energy efficiency analysis and fault early warning system of wind turbine units as an example, a single wind turbine often needs to connect 400 to 500 points from the integrated monitoring platform. Therefore, for a 300MW station, there will be 20,000 points that need to be manually configured and associated.

[0004] Currently, the primary method for configuring signal point associations in actual engineering projects is still manual configuration. According to literature and user manuals, the common format is to display the monitoring platform's signal points in a tree structure on the left and a table or similar format on the right for the signal points to be associated. Associating signal points is done manually by searching through the tree nodes layer by layer, dragging and dropping, or clicking, supplemented by functions such as keyword search, signal point insertion, and retaining empty signals. This traditional method is labor-intensive, inefficient, and prone to human error over long periods, leading to more error correction and rework.

[0005] Meanwhile, according to literature, with the new wave of development in technologies such as big data and artificial intelligence, natural language processing (NLP) technology has found wider applications, such as: comment opinion extraction, address recognition, text sentiment analysis, news tagging, word vector representation, etc. Published papers also show the gradual application of NLP technology in the construction of knowledge graphs and corpora in the power industry, but there are still many application scenarios to be explored. For example, in the association of signal points, where signal points in two systems have similar Chinese names despite not being exactly the same, introducing techniques such as word vectors and sentence similarity analysis from a NLP perspective, and using model prediction and automatic recommendation of association points, can significantly reduce engineering configuration time. Summary of the Invention

[0006] In view of the problems existing in current intelligent data association methods for new energy monitoring, this invention is proposed. Therefore, the problem this invention aims to solve is how to provide an intelligent data association method and system for new energy monitoring.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a data intelligent association method for new energy monitoring, which includes acquiring associated signal point information as a sample dataset and performing feature processing through natural language processing technology.

[0009] Based on the characteristics of multi-label classification problems, an association signal point recommendation model is constructed and trained.

[0010] The associated signal point recommendation model is deployed in the configuration tool of the new energy monitoring system. During the signal point association configuration operation, the system automatically recommends signal points to be associated and records the selection results of the configuration personnel into the historical database.

[0011] The update strategy based on quantitative triggering and sliding window uses historical recommendation matching records to form a rolling data window, sets the window width, and triggers model retraining when the proportion of newly added samples reaches a predetermined proportion.

[0012] The generic pattern design parameterized template is introduced, and the selection template can be selected and exported after all signal points of a single unit or device are associated.

[0013] As a preferred embodiment of the intelligent data association method for new energy monitoring described in this invention, the step of obtaining the associated signal point information as a sample dataset includes: reading the signal name, signal type, and unit number from the wind turbine energy efficiency analysis and fault early warning system, and querying the signal point name of the monitoring platform according to the corresponding signal point ID in the monitoring platform database.

[0014] As a preferred embodiment of the data intelligent association method for new energy monitoring described in this invention, the feature processing using natural language processing technology includes:

[0015] The input signal names are preprocessed to obtain the word list WordList corresponding to the signal names;

[0016] For the word list, the word vector method is used, and the Skip-Gram model is selected and trained using an existing Chinese corpus containing a corpus of electrical engineering vocabulary.

[0017] Convert each Chinese word in the word list into an m-dimensional vector value for each input word;

[0018] For the numbers in the signal name, they are represented as vectors in binary form, also taking m dimensions, and together with the word vectors, they form the feature FvalList;

[0019] Signal types are encoded using one-hot encoding, and unit numbers are represented as vectors in binary form, with each vector having a dimension of m.

[0020] The input feature values ​​are composed of a word list consisting of word vectors, the unit number represented in binary form, and the encoded signal type. The output features are obtained by processing the output signal name in the same way, thus obtaining the sample dataset.

[0021] As a preferred embodiment of the intelligent data association method for new energy monitoring described in this invention, the associated signal point recommendation model includes an embedding layer, a hidden layer, a fully connected layer, and an output layer, and the model is trained using a cross-entropy loss function.

[0022] The hidden layer employs a deep neural network with 64 nodes. The transmission between units in the hidden layer is controlled by four gates: a forget gate, an input gate, a candidate gate, and an output gate. These gates control the degree of memory and forgetting of previous and current information, as shown below:

[0023]

[0024] In the formula, f t The output value of the forget gate, i t The output of the input gate, O t Let σ be the output value of the output gate, and s be the activation function. t-1 x is the output of the previous layer. t For the current input, Here is the forget gate weight matrix. Let b be the weight matrix between the input layer and the hidden layer of the forget gate. f This is the bias value for the forget gate. The input gate weight matrix, Let b be the weight matrix between the input gate, the input layer, and the hidden layer. t This is the bias value of the input gate. This is the output gate weight matrix. Let b be the weight matrix between the input layer and the hidden layer of the output gate. o This is the bias value for the output gate.

[0025] As a preferred embodiment of the intelligent data association method for new energy monitoring described in this invention, the step of training the association signal point recommendation model includes:

[0026] Embedding layers are used to map input features to the model's high-dimensional hidden space;

[0027] The LSTM hidden layer has 64 units, and the timing information is dynamically updated through forget gate, input gate, candidate gate and output gate;

[0028] The fully connected layer uses the sigmoid activation function to output the probability of each candidate signal point and calculates the cross-entropy loss for each label.

[0029] As a preferred embodiment of the intelligent data association method for new energy monitoring described in this invention, the update strategy based on quantitative triggering and sliding windows includes:

[0030] The accumulated recommendation matching records in the historical database are used to form a rolling data window;

[0031] A predetermined window width is set, and when the proportion of new record samples to the total number of samples in the window reaches a predetermined proportion, the associated signal point recommendation model is updated.

[0032] During the update, the oldest data is deleted, the latest data is added, and the associated signal point recommendation model is retrained.

[0033] Secondly, the present invention provides a data intelligent association system for monitoring new energy sources, comprising: a data processing unit for collecting historical data and performing feature processing and extraction on the historical data;

[0034] The model training unit is used to build a recommendation model for associated signal points and perform offline model training.

[0035] The prediction and recommendation unit is used to deploy the associated signal point recommendation model online. During the configuration operation of signal point association, it predicts the names of associated signal points and makes automatic recommendations.

[0036] The update unit is used to update the recommendation model of associated signal points online by combining sliding window and quantitative triggering.

[0037] The template processing unit is used to set the selection template and has the functions of XML format template file keyword parameterization, sequence number memory, and template export.

[0038] The beneficial effects of this invention are as follows: Feature engineering is achieved through NLP (Natural Language Processing), realizing the effective conversion of human text and words into computer- and algorithm-recognizable feature values; a predictive recommendation model is constructed using deep learning algorithms and cross-entropy, intelligently recommending signal points, shortening the time spent on manual point selection; and since model accuracy directly corresponds to the reduced configuration workload, its application significance is significant even when the initial model accuracy is low; the intelligent recommendation method also reduces the probability of errors during manual point selection, improving configuration accuracy; combined with the sliding window method, model updates are performed in a quantitative triggering mode, continuously improving model accuracy; and the use of a generic model enhances the reusability of point selection configuration, further improving configuration efficiency. Compared with existing technologies, this invention features an intelligent process, good configuration accuracy, higher configuration efficiency, and strong reusability. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is an overall flowchart of a data intelligent correlation method for new energy monitoring.

[0041] Figure 2 This is a controllable gate logic structure diagram of the LSTM network in a model of a data intelligent association method for new energy monitoring.

[0042] Figure 3 This is a flowchart of offline training for a data intelligent correlation method for new energy monitoring.

[0043] Figure 4 This is a system UML deployment diagram for a data intelligent association method for new energy monitoring.

[0044] Figure 5 This is a sliding window diagram illustrating a data intelligent correlation method for new energy monitoring. Detailed Implementation

[0045] To make the above-mentioned objects, features, and advantages of the present invention more readily understood, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0046] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0048] Reference Figures 1-5 This is the first embodiment of the present invention, which provides a data intelligent association method for new energy monitoring, including:

[0049] Taking the association of signal points between a wind turbine energy efficiency analysis and fault early warning system and a wind farm integrated monitoring platform as an example, the wind turbine energy efficiency analysis and fault early warning system needs to acquire historical and real-time data of each wind turbine on the wind farm integrated monitoring platform for functions such as turbine monitoring, energy efficiency analysis, and fault analysis. Therefore, in engineering configuration, it is necessary to associate and match the remote signaling points and telemetry points modeled by wind turbines in the wind turbine energy efficiency analysis and fault early warning system with the corresponding signal points stored in the remote signaling and telemetry tables on the wind farm integrated monitoring platform. Compared with the traditional process where the entire process is manually selected and associated by engineers, this invention proposes an intelligent association method and system for new energy monitoring data that combines natural language processing technology, deep learning, sliding window method, and parameterized templates. In scenarios where the database signal points of advanced new energy application systems need to be associated with the database signal points of the monitoring platform, this invention achieves efficient and accurate association of signal points.

[0050] S1: Obtain the associated signal point information as a sample dataset and perform feature processing using natural language processing techniques. Specific steps include:

[0051] First, obtain the database information of other completed engineering configuration sites, use the associated signal point information as a sample dataset, and read the signal name, signal type, and unit number from the database of the wind turbine energy efficiency analysis and fault early warning system as input (e.g., gearbox HS1 bearing temperature, signal type is telemetry, unit number is 5#). The ID number of the associated point stored in the database (the ID number in the monitoring platform database) is used to query and obtain the corresponding signal point name in the monitoring platform as the output (corresponding to the above example: the corresponding signal point name in the monitoring platform is: 5# wind turbine gearbox hub side bearing temperature).

[0052] Combination Figure 3 Feature processing is performed using natural language processing techniques.

[0053] Specifically, the signal name is first preprocessed by removing word segments, replacing special words, removing meaningless symbols, and removing stop words to obtain the word list WordList = [word1, word2, ..., wordN] corresponding to the signal name. Special words include: non-Chinese words, professional abbreviations, and professional synonyms. These special words are saved in a configuration file for dynamic addition and modification.

[0054] Then, further feature extraction is performed. For the WordList, the contextual relationship needs to be considered. The word2vec (word vector) method is used, and the Skip-Gram model is selected. The existing Chinese corpus containing power industry vocabulary is used for training.

[0055] Convert each Chinese word in the word list into an m-dimensional vector value for each input word;

[0056] For the numbers in the signal name, they are represented as vectors in binary form (also m-dimensional). These vectors, together with the word vectors, form the feature FvalList = [numVal1, numVal2, ..., numValN]. This method, because words with similar meanings will have their values ​​close together, reflects the distance and relationship between words, which meets the requirements of this invention.

[0057] Signal types are encoded using one-hot encoding (dimension set to m). Unit numbers are represented as vectors in binary form (dimension set to m).

[0058] Finally, the word list composed of word vectors, the unit number represented in binary form, and the signal type after one-hot encoding constitute the input feature value Xpoint = [Findex, Ftype, FvalList].

[0059] Similarly, for the output quantity (the signal point name corresponding to the monitoring platform), the same processing method as described above is used for data preprocessing and feature extraction to obtain the output feature value Y_sample. This yields the sample dataset (Xpoint, Y_sample).

[0060] S2: Based on the characteristics of multi-label classification problems, and combining techniques such as long short-term memory neural networks and cross-entropy, a recommendation model for associated signal points is constructed and trained.

[0061] The specific steps include: since the desired result is to recommend possible signal points on two platforms for each signal point to be associated, allowing configuration personnel to determine whether to select the recommended signal point, it is a multi-label classification problem, requiring modification of ordinary supervised learning models. The structure of the associated information point recommendation model designed in this invention consists of an embedding layer, a 64-node LSTM (Long Short-Term Memory) hidden layer, a fully connected layer with a sigmoid activation function, and an output layer.

[0062] For the LSTM part of the model, a deep neural network with 64 nodes is used. When the information is passed between units in the hidden layer, it is controlled by four controllable gates (forget gate, input gate, candidate gate, and output gate) to control the degree of memory and forgetting of previous and current information, as shown in the following formula:

[0063]

[0064] In the formula, f t The output value of the forget gate, i t The output of the input gate, O t σ is the output value of the output gate, s is the activation function. t-1 x is the output of the previous layer. t For the current input, Here is the forget gate weight matrix. Let b be the weight matrix between the input layer and the hidden layer of the forget gate. f This is the bias value for the forget gate. The input gate weight matrix, Let b be the weight matrix between the input gate, the input layer, and the hidden layer. t This is the bias value for the input gate. This is the output gate weight matrix. Let b be the weight matrix between the input layer and the hidden layer of the output gate.o This is the bias value for the output gate.

[0065] For the fully connected layer added after the LSTM hidden layer in the model, the probability is calculated using the sigmoid function, and the concept of cross-entropy is introduced. The loss function is designed as the cross-entropy function, thereby transforming the multi-label problem into multiple binary classification problems, so that the model meets the requirements of recommending associated signal points.

[0066] The associated signal point recommendation model constructed above is trained using the obtained sample training set (Xpoint, Y_sample).

[0067] S3: The model is deployed in the configuration tool; during the configuration operation of signal point association, the model automatically recommends signal points to be associated and stores the operation results of the configuration personnel in the history database.

[0068] The specific steps include: deploying the model in the configuration tool of the wind turbine energy efficiency analysis and fault early warning system;

[0069] In the configuration operation of signal point association, each time the engineering configuration personnel click on the space corresponding to the information point to be associated on the configuration tool interface, the recommended signal point name on the monitoring platform is obtained through model prediction and displayed in the drop-down bar.

[0070] The engineering configuration personnel make a selection. If the recommended point matches, they click to select it. If the recommended point does not match, they continue to use the original traditional manual point selection method, selecting a matching signal point from the signal points of the monitoring platform displayed in a tree structure for association.

[0071] The tool stores the operation results of whether the engineering configuration personnel selected the recommended associated points in the history database. The scenario of this invention is an asymmetric scenario mode where significant application effects can be achieved even with low model accuracy requirements. In the example, the initial prediction accuracy of the model is about 65%. Therefore, for every 100 signal points configured, the engineering configuration personnel only need to manually search and select 35 points, saving 65% of the working time. Typically, a wind turbine has about 1000 access points on the monitoring platform. Taking 500 signal points associated with the wind turbine energy efficiency analysis and fault early warning system as an example, a wind turbine can save the time of manually selecting and associating 65*5=325 signal points from tree nodes.

[0072] S4: The update strategy of quantitative triggering + sliding window method is adopted. The accumulated operation results are used as the dataset, and the sample number threshold and window width are set for model update.

[0073] Specifically, as the number of engineering configuration operations increases, the historical records of recommended associated point matching will continue to increase.

[0074] Considering the number of monitoring platform access points for wind farms with a typical installed capacity, and the sample requirements for model training, combined with, for example... Figure 5 The sliding window method shown sets the width N of the sliding window to the number of samples (10,000), and triggers an update when the proportion of new samples reaches 30%. When the threshold is reached, a model update is triggered.

[0075] S5: Introduces a "generic" pattern to design parameterized templates, completes the association of all signal points of a single device, and allows users to save and export the point selection template. Point selection work for the same device can be reused.

[0076] Specifically, the signal point association configuration (i.e., selecting and assigning points) process includes various business logics, such as point table sorting, placeholders for empty points, and verification before adding nodes. The parameterized template function is used to set the following features during the use of the point assignment template: keyword parameterization, which identifies the insertion position of keywords by adding wildcards to the required locations in the template name; and sequence number recording, which allows newly created point assignment templates to record the point assignment sequence number and order. After completing the signal point association for a single device, users can choose whether to save and export the point assignment template based on the actual project situation. If the point assignment template is exported, it is exported in XML format.

[0077] Furthermore, this embodiment also provides a data intelligent association system for new energy monitoring, including: a data processing unit that collects relevant historical data and performs feature processing and extraction on the historical data;

[0078] The model training unit constructs a recommendation model for associated signal points and performs offline model training.

[0079] The prediction and recommendation unit is deployed online. When configuring the association of signal points, the model predicts the names of the associated signal points and makes automatic recommendations.

[0080] The update unit uses a combination of sliding window and quantitative triggering to update the recommendation model of associated signal points online;

[0081] The template processing unit sets up selection templates and features XML format template file keyword parameterization, sequence number memory, and template export functions.

[0082] In summary, feature engineering was achieved through NLP (Natural Language Processing) technology, effectively converting human text and words into computer- and algorithm-recognizable feature values. A predictive recommendation model was constructed using deep learning algorithms and cross-entropy to intelligently recommend signal points, reducing the time spent on manual point selection. Furthermore, since model accuracy directly corresponds to the reduced configuration workload, its application is significant even when the initial model accuracy is low. The intelligent recommendation method also reduces the probability of errors during manual point selection, improving configuration accuracy. Combining the sliding window method with a quantitative triggering mode for model updates allows for continuous improvement in model accuracy. The adoption of a generic model enhances the reusability of point selection configuration, further improving configuration efficiency. Compared with existing technologies, this invention features an intelligent process, good configuration accuracy, higher configuration efficiency, and strong reusability.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data intelligent association method for new energy monitoring, characterized in that: include, The associated signal point information is obtained as a sample dataset, and feature processing is performed using natural language processing techniques. The process of obtaining associated signal point information as a sample dataset includes: reading the signal name, signal type, and unit number from the wind turbine energy efficiency analysis and fault early warning system, and querying the monitoring platform signal point name based on the corresponding signal point ID in the monitoring platform database; The feature processing using natural language processing technology includes: The input signal names are preprocessed to obtain the word list WordList corresponding to the signal names; For the word list, the word vector method is used, and the Skip-Gram model is selected and trained using an existing Chinese corpus containing a corpus of electrical engineering vocabulary. Convert each Chinese word in the word list into an m-dimensional vector value for each input word; For the numbers in the signal name, they are represented as vectors in binary form, also taking m dimensions, and together with the word vectors, they form the feature FvalList; Signal types are encoded using one-hot encoding, and unit numbers are represented as vectors in binary form, with each vector having a dimension of m. The input feature values ​​are composed of a word list consisting of word vectors, the unit number represented in binary form, and the encoded signal type. The output features are obtained by processing the output signal name in the same way, thus obtaining the sample dataset. Based on the characteristics of multi-label classification problems, an association signal point recommendation model is constructed and trained. The associated signal point recommendation model is deployed in the configuration tool of the new energy monitoring system. During the signal point association configuration operation, the system automatically recommends signal points to be associated and records the selection results of the configuration personnel into the historical database. The associated signal point recommendation model includes an embedding layer, a hidden layer, a fully connected layer, and an output layer, and is trained using a cross-entropy loss function. The hidden layer employs a deep neural network with 64 nodes. The transmission between units in the hidden layer is controlled by four gates: a forget gate, an input gate, a candidate gate, and an output gate. These gates control the degree of memory and forgetting of previous and current information, as shown below: In the formula, The output value of the forget gate. The output of the input gate. The output value of the output gate. For activation function, This is the output of the previous layer. For the current input, Here is the forget gate weight matrix. This is the weight matrix between the input layer and the hidden layer, which is the forget gate. This is the bias value for the forget gate. The input gate weight matrix, This is the weight matrix between the input gate, the input layer, and the hidden layer. This is the bias value of the input gate. This is the output gate weight matrix. This is the weight matrix between the input layer and the hidden layer of the output gate. This is the bias value for the output gate; The update strategy based on quantitative triggering and sliding window uses historical recommendation matching records to form a rolling data window, sets the window width, and triggers model retraining when the proportion of newly added samples reaches a predetermined proportion. The generic pattern design parameterized template is introduced, and the selection template can be selected and exported after all signal points of a single unit or device are associated.

2. The intelligent data association method for new energy monitoring as described in claim 1, characterized in that: The training of the associated signal point recommendation model includes: Embedding layers are used to map input features to the model's high-dimensional hidden space; The LSTM hidden layer has 64 units, and the timing information is dynamically updated through forget gate, input gate, candidate gate and output gate; The fully connected layer uses the sigmoid activation function to output the probability of each candidate signal point and calculates the cross-entropy loss for each label.

3. The intelligent data association method for new energy monitoring as described in claim 2, characterized in that: The update strategy based on quantitative triggering and sliding window includes: The accumulated recommendation matching records in the historical database are used to form a rolling data window; A predetermined window width is set, and when the proportion of new record samples to the total number of samples in the window reaches a predetermined proportion, the associated signal point recommendation model is updated. During the update, the oldest data is deleted, the latest data is added, and the associated signal point recommendation model is retrained.

4. A data intelligent association system for new energy monitoring, based on the data intelligent association method for new energy monitoring as described in any one of claims 1 to 3, characterized in that: include, The data processing unit is used to collect historical data and perform feature processing and extraction on the historical data; The model training unit is used to build a recommendation model for associated signal points and perform offline model training. The prediction and recommendation unit is used to deploy the associated signal point recommendation model online. During the configuration operation of signal point association, it predicts the names of associated signal points and makes automatic recommendations. The update unit is used to update the recommendation model of associated signal points online by combining sliding window and quantitative triggering. The template processing unit is used to set the selection template and has XML format keyword parameterization function, serial number memory function, and template export function.

Images