Regulation and control operation simulation training method, system and equipment based on collaborative filtering recommendation algorithm and medium

Abstract

The invention relates to the technical field of regulation and control operation simulation training, in particular to a regulation and control operation simulation training method, system and device based on a collaborative filtering recommendation algorithm and a medium. Collecting regulation and control operation data, and preprocessing the regulation and control operation data; a recommendation algorithm is adopted to analyze the regulation and control operation data and the project features, the similarity relation of the regulation and control tasks is mined, and a recommendation training scheme is generated according to the similarity relation; generating a task recommendation report according to the recommendation scheme, collecting user feedback information and constructing a closed-loop optimization mechanism; and fusing the feedback information into a model incremental learning process, so that the model is iteratively updated on new data. Continuous online learning of the model is realized through a feedback mechanism, and recommendation results are continuously optimized along with changes of tasks and user behaviors.

Description

Technical Field

[0001] This invention relates to the field of regulatory operation simulation training technology, and in particular to a regulatory operation simulation training method, system, device and medium based on collaborative filtering recommendation algorithm. Background Technology

[0002] With the continuous expansion of power system scale, the complexity of dispatching tasks is also increasing, posing challenges to traditional manual training and operational optimization methods. Dispatchers need to make quick decisions when faced with multiple operational points; however, manual decision-making is easily affected by fatigue and experience limitations, and traditional training methods are time-consuming and labor-intensive. Therefore, how to improve the training efficiency and operational accuracy of dispatchers has become an urgent problem to be solved in the field of power dispatching. This invention proposes a control operation simulation training method based on a collaborative filtering recommendation algorithm. By analyzing dispatcher operational behavior data and historical decisions, similarity is calculated and relevant training tasks or simulation scenarios are recommended. In addition, the system combines feedback and adaptive learning modules to continuously adjust the recommendation strategy based on dispatcher performance, ensuring continuous optimization of training content and methods. This data-driven recommendation mechanism can improve dispatcher learning efficiency and operational capabilities, optimize the dispatcher training process, and support the intelligent development of the power dispatching field. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] In a first aspect, the present invention provides a method for regulating and simulating training based on a collaborative filtering recommendation algorithm, comprising:

[0006] Collect and preprocess the control and regulation operation data;

[0007] A recommendation algorithm is used to analyze the operational data and project characteristics of the regulation, mine the similarity relationships of the regulation tasks, and generate recommended training schemes based on the similarity relationships;

[0008] Based on the recommendation scheme, a task recommendation report is generated, user feedback information is collected, and a closed-loop optimization mechanism is constructed.

[0009] By incorporating feedback into the incremental learning process of the model through online learning methods, the model can be iteratively updated on new data.

[0010] As a preferred embodiment of the regulation and operation simulation training method based on collaborative filtering recommendation algorithm described in this invention, the step of collecting regulation and operation data and preprocessing the regulation and operation data includes:

[0011] Based on the regulatory business logic and scheduling operation scenarios, key features are extracted from the raw data;

[0012] The processed data is stored uniformly.

[0013] As a preferred embodiment of the regulatory operation simulation training method based on collaborative filtering recommendation algorithm described in this invention, the method employs a recommendation algorithm to analyze regulatory operation data and project characteristics, and mines the similarity relationships of regulatory tasks, including:

[0014] Statistical algorithms are used to assign weights to each feature;

[0015] The similarity between all operation tasks is calculated using a vector similarity metric, and an adjacency matrix of the projects is established.

[0016] The preferred technical solutions in the embodiments of this application have the following beneficial effects:

[0017] Firstly, by using collaborative filtering algorithms to mine the matching relationship between schedulers and tasks, personalized training task lists are generated for schedulers with different ability levels or professional directions, avoiding a "one-size-fits-all" uniform training method, thereby significantly improving learning efficiency and training effect.

[0018] Secondly, a project feature vector space is established based on historical operation data and task characteristics. Training content is automatically recommended by calculating task similarity, replacing the tedious process of manually formulating training plans and reducing training organization costs and scheduling management manpower burden.

[0019] As a preferred embodiment of the control and simulation training method based on collaborative filtering recommendation algorithm described in this invention, wherein: the step of generating a recommendation training scheme based on similarity relationships includes,

[0020] Establish a scoring prediction model to calculate the predicted scores of the scheduler on untrained tasks;

[0021] A matrix factorization model is introduced to model the implicit preference relationship between the scheduler and the personnel.

[0022] The preferred technical solutions in the embodiments of this application have the following beneficial effects:

[0023] Firstly, by using matrix factorization, the interaction between the scheduler and the task is embedded into the same latent vector space, transforming explicit ratings or implicit behaviors into potential preference relationships. This allows for accurate prediction of the scheduler's tendencies on untrained tasks, enhancing the personalization and relevance of recommendations.

[0024] Secondly, by using a scoring prediction model to quantify the scheduler's potential interest in different tasks, the system can set personalized training priorities and dynamically adjust the order of training tasks, thereby achieving flexible arrangement and dynamic delivery of training content.

[0025] As a preferred embodiment of the control and simulation training method based on collaborative filtering recommendation algorithm described in this invention, wherein: a statistical algorithm is used to assign weights to each feature, expressed as follows:

[0026]

[0027] Among them, tf i,j df represents the frequency of the j-th feature in the i-th item; j This represents the number of items containing feature j; N is the total number of items.

[0028] As a preferred embodiment of the scheduling simulation training method based on collaborative filtering recommendation algorithm described in this invention, wherein: the establishment of a rating prediction model to calculate the scheduler's predicted rating on untrained tasks is expressed as follows:

[0029]

[0030] Where N(i; u) represents the set of tasks similar to task i that have been rated by scheduler u; r u,j sim(i,j) represents the rating of task j by the scheduler u; sim(i,j) represents the similarity between tasks i and j.

[0031] As a preferred embodiment of the collaborative filtering recommendation algorithm-based regulation and operation simulation training method described in this invention, the introduction of a matrix factorization model to model the implicit preference relationship between the scheduler and the user includes decomposing the rating matrix R into a user latent factor matrix U∈Rk×u and an item latent factor matrix V∈Rk×m, denoted as follows:

[0032]

[0033] The model parameters are updated using stochastic gradient descent, and the model is iteratively optimized, as shown below.

[0034]

[0035] Where: K represents the known set of ratings; λ represents the regularization parameter; U u and V i These are the latent factor vectors for user u and project i, respectively.

[0036] Secondly, the present invention provides a regulation and operation simulation training system based on a collaborative filtering recommendation algorithm, comprising:

[0037] The data acquisition module collects and preprocesses the control and regulation operation data.

[0038] The calculation module uses a recommendation algorithm to analyze the control operation data and project characteristics, mine the similarity relationships of control tasks, and generate recommended training schemes based on the similarity relationships.

[0039] The generation module generates a task recommendation report based on the recommendation scheme, collects user feedback information, and builds a closed-loop optimization mechanism.

[0040] The iterative module incorporates feedback information into the incremental learning process of the model through online learning methods, enabling the model to be iteratively updated on new data.

[0041] Thirdly, the present invention provides an electronic device, comprising:

[0042] Memory and processor;

[0043] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of a control operation simulation training method based on a collaborative filtering recommendation algorithm.

[0044] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the aforementioned method for regulating operation simulation training based on a collaborative filtering recommendation algorithm.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] First, adopting a project-based collaborative filtering method can more accurately capture the correlation between tasks and reduce the impact of the cold start problem.

[0047] Secondly, the model can continuously learn online through a feedback mechanism, constantly optimizing the recommendation results as the task and user behavior change.

[0048] Third, by employing matrix factorization and nearest neighbor search optimization, computation can be completed efficiently even on large-scale datasets.

[0049] Fourth, by using task scoring prediction and scheduler preference modeling mechanisms, a closed-loop process is achieved, from historical operation data analysis and similar task recommendation to task feedback and retraining of the model, thereby enhancing the efficiency of knowledge transfer. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the 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.

[0051] Figure 1 This is a visual diagram illustrating obstacle avoidance trajectories considering and not considering terrain error in a collaborative filtering recommendation algorithm-based control simulation training method according to an embodiment of the present invention. Detailed Implementation

[0052] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, 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.

[0053] Example 1, an embodiment of the present invention, provides a method for regulating and simulating training based on a collaborative filtering recommendation algorithm, comprising:

[0054] S1: Collect and preprocess the control operation data;

[0055] S2: Using a recommendation algorithm, analyze the operational data and project characteristics of the regulation, mine the similarity relationships of the regulation tasks, and generate recommended training schemes based on the similarity relationships;

[0056] S3: Generate a task recommendation report based on the recommendation scheme, collect user feedback information, and build a closed-loop optimization mechanism;

[0057] S4: By using online learning methods, feedback information is integrated into the incremental learning process of the model, enabling the model to be iteratively updated on new data.

[0058] It should be noted that, addressing the challenges of complex power system dispatching tasks, limitations of manual decision-making, and the inefficiency of traditional training, a simulation training method for control operation based on a collaborative filtering recommendation algorithm is proposed. By collecting and preprocessing control operation data, a data foundation is laid for understanding task complexity. The recommendation algorithm analyzes the data, mines task similarities, and generates personalized training plans, accurately improving dispatchers' ability to handle complex scenarios. The solution also generates recommendation reports and collects user feedback, forming a closed-loop optimization mechanism to ensure the effectiveness and continuous improvement of training content. An online learning method is employed, integrating feedback into model iteration updates, allowing the recommendation strategy to be continuously optimized through practice. This closed-loop training system based on data, intelligent recommendation, and continuous learning effectively compensates for the limitations of manual decision-making, significantly improves dispatcher training efficiency and operational accuracy, and provides strong support for the intelligent development of power dispatching.

[0059] Example 2, refer to Figure 1 This is one embodiment of the present invention. Based on the above embodiment, a method for regulating operation simulation training based on collaborative filtering recommendation algorithm is provided.

[0060] The embodiments of this application can combine the business characteristics of control and operation with the personalized needs of dispatchers, and use collaborative filtering recommendation algorithms to provide an efficient simulation training method.

[0061] In this embodiment of the application, step S1, which involves collecting and preprocessing the control operation data, includes steps A1-A2:

[0062] A1: Extract key features from the raw data based on the control business logic and scheduling operation scenarios.

[0063] A2: Store the processed data in a unified manner.

[0064] It should be noted that control operation data, such as historical operation records, scheduler operation logs, simulation training data, and feedback data, are collected through API interfaces, data file uploads, and timed data synchronization.

[0065] Furthermore, in order to improve data quality, the control and operation data is cleaned and processed. The specific implementation method is as follows:

[0066] Outlier identification in numerical data using the Z-Score standard deviation method:

[0067]

[0068] Where xi is the i-th data point, μ is the sample mean, and σ is the sample standard deviation. When |Zi| > 3, it is considered an outlier and is deleted or corrected by interpolation.

[0069] For time series data, use linear interpolation to fill in missing values:

[0070]

[0071] Here, x(t0) and x(t1) are the valid data at the two time points before and after the missing value.

[0072] For categorical variables, use the mode to fill in or remove missing records.

[0073] By constructing a hash function H(x), each record is uniquely identified.

[0074] H(x) = hash(operation type + timestamp + device ID)

[0075] Numerical data is normalized to ensure consistent scale across different features, facilitating processing by collaborative filtering algorithms. The Min-Max normalization method is employed.

[0076]

[0077] Mapping all feature values ​​to the [0,1] interval improves the accuracy of similarity calculation in collaborative filtering algorithms.

[0078] It should be noted that the specific implementation methods for steps A1-A2 are as follows:

[0079] Based on the regulatory business logic and scheduling operation scenarios, the following key features are extracted from the raw data and used as inputs for the recommendation algorithm and training module:

[0080] (1) Types of dispatching operations: including ticket drafting, maintenance management, network command issuance, and accident handling;

[0081] (2) Operation execution time: Record the start and end times of the operation;

[0082] (3) Equipment information involved: Identify substations, switching equipment, lines, etc. involved in the dispatching process;

[0083] (4) Operation frequency: The frequency of use of specific operation types or equipment by the dispatcher;

[0084] (5) Task result feedback: including whether the operation was correct, the completion time, and the score.

[0085] Finally, the processed data is stored uniformly in a relational database or distributed storage system for use by the collaborative filtering recommendation model's calling and training modules.

[0086] In this embodiment of the application, step S2 uses a recommendation algorithm to analyze the regulation operation data and project characteristics, mine the similarity relationship of regulation tasks, and generate a recommended training scheme based on the similarity relationship;

[0087] In this embodiment, the recommendation algorithm in step S2 employs an item-based collaborative filtering algorithm. This algorithm analyzes historical data and item characteristics of control operations, mines similarity relationships between control tasks, and then predicts the scheduler's performance in new tasks based on past behavior, generating a personalized recommendation training scheme. This includes steps B1-B4:

[0088] B1: Statistical algorithms are used to assign weights to each feature.

[0089] B2: Calculate the similarity between all operation tasks using the vector similarity metric and establish the adjacency matrix of the projects.

[0090] B3: Establish a scoring prediction model to calculate the predicted scores of the scheduler on untrained tasks.

[0091] B4: Introduce a matrix factorization model to model the implicit preference relationship between the dispatcher and the characters.

[0092] In an optional embodiment, the recommendation algorithm in step S2 is the same as that used in this embodiment, which analyzes the regulation operation data and project characteristics, mines the similarity relationship of regulation tasks, and generates a recommended training scheme based on the similarity relationship.

[0093] In this embodiment, the recommendation algorithm in step S2 employs an item-based collaborative filtering algorithm. This algorithm analyzes historical data and item characteristics of control operations, mines similarity relationships between control tasks, and then predicts the scheduler's performance in new tasks based on past behavior, generating a personalized recommendation training scheme. This includes steps B1-B4:

[0094] B1: Statistical algorithms are used to assign weights to each feature.

[0095] B2: Calculate the similarity between all operation tasks using the vector similarity metric and establish the adjacency matrix of the projects.

[0096] B3: Establish a scoring prediction model to calculate the scheduler's predicted scores on untrained tasks.

[0097] B4: Introduce a matrix factorization model to model the implicit preference relationship between the dispatcher and the characters.

[0098] In an optional embodiment, the recommendation algorithm in step S2 is a content-based recommendation algorithm. When the power dispatch center has accumulated enough structured dispatcher training records and feedback data, if it is necessary to quickly provide preliminary training suggestions based on task attributes for newly hired or inexperienced dispatchers, or to provide basic recommendations for specific types of tasks (such as new equipment operation), content-based recommendation can be used as an auxiliary or transitional solution.

[0099] In another optional embodiment, the recommendation algorithm in step S2 employs a hybrid recommendation algorithm. When the system has run for a period of time, accumulating sufficient scheduler interaction data and feedback, and it is found that a single collaborative filtering algorithm has significant shortcomings in certain aspects (such as new task recommendation, cold start problem, and recommendation interpretability), a hybrid recommendation algorithm can be introduced. For example, when it is desired that the system can recommend not only tasks that "others also like," but also tasks "related to your existing skills," hybrid recommendation is an ideal choice. When stronger recommendation interpretability is required to meet the scheduler's requirements for recommendation rationality, hybrid recommendation can also better meet this need.

[0100] In this embodiment of the application, the statistical algorithm in step B1 adopts the TF-IDF method, and the specific implementation is as follows:

[0101] Each operation item is represented as a high-dimensional vector using a project-feature matrix. Each row represents an operation item, and each column represents a feature, such as equipment type, control type, task achievement rate, etc. The TF-IDF method is used to assign weights to each feature, reducing the weights of high-frequency but less discriminative features. The TF-IDF weight calculation formula is as follows:

[0102]

[0103] Among them, tf i,j df represents the frequency of the j-th feature in the i-th item; j This represents the number of items containing feature j; N is the total number of items.

[0104] In an optional embodiment, the statistical algorithm in step B1 can also employ a chi-square test, when the characteristics of the control task are primarily discrete or can be easily discretized. For example, the task achievement rate can be divided into several levels: "high," "medium," and "low." Similarly, the task needs to have some quantifiable classification result or attribute in order to perform a chi-square test to assess the independence between the characteristics and the result.

[0105] In another optional embodiment, the statistical algorithm in step B1 can also use information gain. When there is a control task scenario with clear classification labels, information gain can be selected to assign weights to each feature.

[0106] In this embodiment, the vector similarity measurement method in step B2 uses cosine similarity to calculate the similarity between items, calculates the similarity between all operation tasks, and establishes an adjacency matrix for the items. The specific implementation is as follows:

[0107]

[0108] The similarity calculation results between all pairs of items are constructed into an item similarity matrix S∈Rm×m, where Si,j=sim(i,j).

[0109] In an optional embodiment, the vector similarity metric in step B2 can also employ Euclidean distance when the item features represent specific, measurable physical quantities, and the absolute differences between these quantities are significant for task similarity. Both control tasks involve adjusting the same type of equipment, but the adjustment magnitudes differ greatly; Euclidean distance can effectively reflect this difference.

[0110] In another alternative embodiment, the vector similarity measurement method in step B2 can also employ the Pearson correlation coefficient, which is useful when there is a latent linear relationship between item features and when the centralization of features is not critical. For example, if the features of a regulatory task tend to increase or decrease together, the Pearson coefficient can effectively capture this pattern of coordinated change.

[0111] In this embodiment of the application, step B3 is specifically implemented as follows:

[0112] After obtaining the similarity matrix between projects, the scoring prediction stage begins. The goal is to predict the scheduler's likely performance on incomplete tasks based on their scoring records of completed tasks, thereby generating a recommended task list.

[0113] The prediction score for scheduler u on untrained task i is:

[0114]

[0115] Where N(i; u) represents the set of tasks similar to task i that have been rated by scheduler u; r u,j sim(i,j) represents the rating of task j by the scheduler u; sim(i,j) represents the similarity between tasks i and j.

[0116] The specific implementation method of step B4 is as follows:

[0117] To further model the implicit preference relationship between schedulers and tasks, a matrix factorization model is introduced. The rating matrix R is decomposed into a user latent factor matrix U∈Rk×u and an item latent factor matrix V∈Rk×m, i.e.:

[0118]

[0119] The training objective is to minimize the predicted score. Compared with the actual score r u,i The differences between them.

[0120] The model parameters are updated using stochastic gradient descent (SGD), and the model is iteratively optimized.

[0121]

[0122] Where K represents the known set of ratings; λ represents the regularization parameter; U u and V i These are the latent factor vectors for user u and project i, respectively.

[0123] The model training is completed by iteratively updating U and V until the loss function converges.

[0124] It should be noted that MAE (mean absolute error) and RMSE (root mean square error) are used to evaluate the error of the recommendation results to ensure the accuracy of the recommended path.

[0125] The specific algorithm for MAE (Mean Absolute Error) is as follows:

[0126]

[0127] The specific algorithm for RMSE (Root Mean Square Error) is as follows:

[0128]

[0129] In this embodiment of the application, step S3, which involves generating a task recommendation report based on the recommendation scheme, collecting user feedback information, and constructing a closed-loop optimization mechanism, is specifically implemented as follows:

[0130] Furthermore, for each recommended task i, the system displays the following:

[0131] Match score: i.e., predicted score

[0132] Similar task sources: Showing representative tasks from the task nearest neighbor set N(i; u) used for score prediction, and their actual scores r. u,i ;

[0133] Operational suggestions: Based on each dimension of the task feature vector (such as device type, task difficulty, historical achievement rate, etc.), we provide operational precautions and recommended strategies.

[0134] Furthermore, collect user feedback and build a closed-loop optimization mechanism:

[0135] After the recommendation results are displayed, the system allows schedulers to provide feedback after executing or browsing the recommended tasks. If a scheduler confirms that a recommended task is "useful" or "matches" the task, this action can be considered a positive implicit rating. If the scheduler skips the task or explicitly marks it as "not applicable," it is considered negative feedback. This feedback data can be converted into a new rating record.

[0136] Positive feedback: setting r u,i =1;

[0137] Negative feedback: setting r u,i =0;

[0138] The new data will be used as supplementary training samples, re-added to the rating matrix R, and re-participated in the loss function optimization during model training. As feedback samples accumulate, the model will gradually correct the scheduler's preference vector U. u This enables the personalized model to continuously self-optimize.

[0139] In this embodiment of the application, step S4, which integrates feedback information into the incremental learning process of the model through an online learning method, enables the model to iteratively update on new data. The specific implementation method is as follows:

[0140] It should be noted that the feedback mechanism enables the recommendation system to "self-improve" and improve the accuracy of recommendations.

[0141] Feedback data collection: Collect information such as the scheduler's click behavior on recommended tasks, completion status, and feedback on recommendation results.

[0142] This feedback information is uniformly coded into a comprehensive feedback score r. ui i (t) This indicates the degree of feedback from user u to task i at time t:

[0143] r ui i (t) =f(c ui i (t) ,f ui i (t) ,s ui i (t) )

[0144] Where: c ui i (t) Indicates a click action (1 indicates a click, 0 indicates no click); fui i (t) : Indicates whether the task is completed (1 indicates completion, 0 indicates incomplete); s ui i (t) : Indicates an explicit rating (1–5 points).

[0145] Feedback data processing: Transform the feedback data into new training samples and update the task's preference matrix and the item similarity matrix.

[0146] 3. Adaptive Model Update: By using online learning methods, the feedback data from the scheduler is incorporated into the incremental learning process of the model, enabling the model to be iteratively updated based on new data.

[0147] The predictive expression for the recommendation rating is:

[0148]

[0149] An online optimization method based on stochastic gradient descent (SGD) is introduced to perform local model updates on newly added feedback samples:

[0150]

[0151] in: η is the prediction error; η is the learning rate; λ is the regularization coefficient.

[0152] The new prediction score is:

[0153]

[0154] This rating is used to update the recommended ranking list in real time, forming a closed loop of "feedback-learning-recommendation".

[0155] Example 3 illustrates a schematic scheme for a regulation and operation simulation training method based on a collaborative filtering recommendation algorithm. It should be noted that the technical solution of this system for regulation and operation simulation training based on a collaborative filtering recommendation algorithm is based on the same concept as the technical solution of the regulation and operation simulation training method based on a collaborative filtering recommendation algorithm described above. Details not described in detail in this embodiment can be found in the description of the technical solution of the regulation and operation simulation training method based on a collaborative filtering recommendation algorithm described above.

[0156] This embodiment also provides a regulation and operation simulation training system based on a collaborative filtering recommendation algorithm, including:

[0157] The data acquisition module collects and preprocesses the control and regulation operation data.

[0158] The calculation module uses a recommendation algorithm to analyze the control operation data and project characteristics, mine the similarity relationships of control tasks, and generate recommended training schemes based on the similarity relationships.

[0159] The generation module generates a task recommendation report based on the recommendation scheme, collects user feedback information, and builds a closed-loop optimization mechanism.

[0160] The iterative module incorporates feedback information into the incremental learning process of the model through online learning methods, enabling the model to be iteratively updated on new data.

[0161] This embodiment also provides an electronic device applicable to a situation of regulation and operation simulation training based on a collaborative filtering recommendation algorithm, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the regulation and operation simulation training method based on a collaborative filtering recommendation algorithm as proposed in the above embodiment.

[0162] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a simulation training method for regulating operation based on a collaborative filtering recommendation algorithm as proposed in the above embodiments.

[0163] The storage medium proposed in this embodiment and the implementation of a control operation simulation training method based on collaborative filtering recommendation algorithm proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0164] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0165] 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 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 method for regulating and simulating training based on a collaborative filtering recommendation algorithm, characterized in that, include: Collect and preprocess the control and regulation operation data; A recommendation algorithm is used to analyze the operational data and project characteristics of the regulation, mine the similarity relationships of the regulation tasks, and generate recommended training schemes based on the similarity relationships; Based on the recommendation scheme, a task recommendation report is generated, user feedback information is collected, and a closed-loop optimization mechanism is constructed. Feedback information is incorporated into the incremental learning process of the model, enabling the model to be iteratively updated on new data.

2. The method for regulating operation simulation training based on collaborative filtering recommendation algorithm as described in claim 1, characterized in that, The process involves collecting and preprocessing the control and regulation operation data. include, Based on the regulatory business logic and scheduling operation scenarios, key features are extracted from the raw data; The processed data is stored uniformly.

3. The method for regulating and simulating training based on collaborative filtering recommendation algorithm as described in claim 2, characterized in that, The aforementioned method employs a recommendation algorithm to analyze operational data and project characteristics, thereby uncovering similarities between operational tasks. include, Statistical algorithms are used to assign weights to each feature; The similarity between all operation tasks is calculated using a vector similarity metric, and an adjacency matrix of the projects is established.

4. The method for regulating operation simulation training based on collaborative filtering recommendation algorithm as described in claim 3, characterized in that, The step of generating a recommended training scheme based on similarity relationships includes: Establish a scoring prediction model to calculate the predicted scores of the scheduler on untrained tasks; A matrix factorization model is introduced to model the implicit preference relationship between the scheduler and the personnel.

5. The method for regulating operation simulation training based on collaborative filtering recommendation algorithm as described in claim 4, characterized in that, The statistical algorithm is used to assign weights to each feature, as shown below. Among them, tf i,j df represents the frequency of the j-th feature in the i-th item; j This represents the number of items containing feature j; N is the total number of items.

6. The method for regulating operation simulation training based on collaborative filtering recommendation algorithm as described in claim 5, characterized in that, The established scoring prediction model calculates the scheduler's predicted score on untrained tasks, denoted as follows: Where N(i; u) represents the set of tasks similar to task i that have been rated by scheduler u; r u,j sim(i,j) represents the rating of task j by the scheduler u; sim(i,j) represents the similarity between tasks i and j.

7. The method for regulating operation simulation training based on collaborative filtering recommendation algorithm as described in claim 6, characterized in that, The introduction of a matrix factorization model to model the implicit preference relationship between schedulers and personnel includes decomposing the rating matrix R into a user latent factor matrix U∈Rk×u and an item latent factor matrix V∈Rk×m, denoted as follows: The model parameters are updated using stochastic gradient descent, and the model is iteratively optimized, as shown below. Where: K represents the known set of ratings; λ represents the regularization parameter; U u and V i These are the latent factor vectors for user u and project i, respectively.

8. A control operation simulation training system based on a collaborative filtering recommendation algorithm, using the method described in any one of claims 1-7, characterized in that, include: The data acquisition module collects and preprocesses the control and regulation operation data. The calculation module uses a recommendation algorithm to analyze the control operation data and project characteristics, mine the similarity relationships of control tasks, and generate recommended training schemes based on the similarity relationships. The generation module generates a task recommendation report based on the recommendation scheme, collects user feedback information, and builds a closed-loop optimization mechanism. The iterative module incorporates feedback information into the incremental learning process of the model through online learning methods, enabling the model to be iteratively updated on new data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. 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 steps of the method according to any one of claims 1 to 7.

Images