Auxiliary prompt information generation method and system based on knowledge reasoning

Abstract

The invention relates to the technical field of information generation, and particularly discloses an auxiliary prompt information generation method and system based on knowledge reasoning, and the method comprises the steps: carrying out the semantic feature extraction and feature interaction fusion analysis of the operation state data of a current power grid and the behavior data of an operator through employing an artificial intelligence technology based on deep learning; according to the method, the hidden association between the operator behavior and the power grid operation state is mined, a rich context background is provided for generation of the auxiliary prompt information, and then the targeted auxiliary prompt information is intelligently generated in combination with an auxiliary prompt template problem. Therefore, by deeply analyzing the operation state of the power grid and the operation behavior of the personnel, the auxiliary prompt information conforming to the current situation can be dynamically generated, so that the intelligent level of power grid operation monitoring and management is improved, and the decision support capability of an operator in the face of complex situations is enhanced.

Description

Technical Field

[0001] The present application relates to the field of information generation, and more specifically, to a method and system for generating auxiliary prompt information based on knowledge reasoning. Background Art

[0002] In power systems, grid operation status monitoring and operational management are critical to ensuring a stable and secure power supply. With the continuous expansion of grid scale and the advancement of intelligent technology, the amount of real-time operational data generated by grid operations, such as voltage, current, power, and load distribution, as well as the behavioral data generated by operators when performing tasks such as scheduling, maintenance, and troubleshooting, has increased dramatically. This has made traditional grid operation monitoring and management methods, which rely on manual experience, increasingly unable to cope with the increasingly complex data processing requirements.

[0003] To improve the intelligence of power grid operations, new technologies are urgently needed to assist operators in decision-making. However, traditional methods for generating auxiliary prompts typically rely on pre-set template questions and corresponding fixed template answers. While these methods can provide some basic guidance, the fixed nature of these templates makes them inflexible in addressing the various complex situations that arise in power grid operations. This is especially true when faced with sudden failures or abnormal events, making it difficult to provide timely and effective prompts and suggestions.

[0004] Therefore, an optimized method and system for generating auxiliary prompt information based on knowledge reasoning is expected. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and system for generating auxiliary prompt information based on knowledge reasoning, which uses artificial intelligence technology based on deep learning to perform semantic feature extraction and feature interaction fusion analysis on the current power grid operation status data and operator behavior data, so as to dig out the implicit correlation between operator behavior and power grid operation status, provide a rich contextual background for the generation of auxiliary prompt information, and then combine the auxiliary prompt template problem to intelligently generate targeted auxiliary prompt information. In this way, by deeply analyzing the operation status of the power grid and the operation behavior of the personnel, auxiliary prompt information that conforms to the current situation can be dynamically generated, thereby improving the intelligence level of power grid operation monitoring and management, and enhancing the decision-making support capabilities of operators when facing complex situations.

[0006] According to one aspect of the present application, a method for generating auxiliary prompt information based on knowledge reasoning is provided, which includes:

[0007] Collect current grid operation status data through sensors and extract operator behavior data from interaction logs;

[0008] Performing vectorization processing on the current power grid operation state data and the operator behavior data respectively to obtain a power grid operation state embedded coding vector and an operator behavior data semantic coding vector;

[0009] Performing feature interaction coding on the operator behavior data semantic coding vector and the grid operation state embedded coding vector to obtain an operator behavior-grid operation state significant interaction fusion representation vector;

[0010] Get auxiliary prompt template questions;

[0011] Performing semantic encoding on the auxiliary prompt template question to obtain an auxiliary modal prompt question semantic encoding feature vector;

[0012] Auxiliary prompt information is generated based on the joint features of the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector.

[0013] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, feature interaction encoding is performed on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector, including: extracting the joint implicit features between the operator behavior data semantic coding vector and the power grid operation status embedded coding vector as a conditional feature vector; based on the semantic contribution of the operator behavior data semantic coding vector and the power grid operation status embedded coding vector relative to the conditional feature vector, feature modulation is performed on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector to obtain a modulated operator behavior data semantic coding vector and a modulated power grid operation status embedded coding vector; cross-domain interaction encoding is performed on the modulated operator behavior data semantic coding vector, the modulated power grid operation status embedded coding vector and the conditional feature vector based on a converter structure to obtain the operator behavior-power grid operation status significant interaction fusion representation vector.

[0014] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the current power grid operation status data and the operator behavior data are respectively vectorized to obtain a power grid operation status embedded coding vector and an operator behavior data semantic coding vector, including: using a power grid operation status embedded coding matrix to embed the current power grid operation status data to obtain the power grid operation status embedded coding vector; and semantically encoding the operator behavior data to obtain the operator behavior data semantic coding vector.

[0015] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the joint implicit features between the operator behavior data semantic coding vector and the grid operation status embedded coding vector are extracted as conditional feature vectors, including: inputting the operator behavior data semantic coding vector and the grid operation status embedded coding vector into a joint implicit feature capture network to obtain an operator behavior-grid operation status joint implicit feature vector; and performing feature activation based on a Sigmoid function on the operator behavior-grid operation status joint implicit feature vector to obtain the conditional feature vector.

[0016] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the operator behavior data semantic coding vector and the power grid operation status embedded coding vector are input into the joint implicit feature capture network to obtain the operator behavior-power grid operation status joint implicit feature vector, including: the operator behavior data semantic coding vector and the power grid operation status embedded coding vector are fused according to the position points and then passed through a neural network layer based on the tanh function to obtain the operator behavior-power grid operation status joint implicit feature vector.

[0017] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, based on the semantic contribution of the operator behavior data semantic coding vector and the power grid operation status embedded coding vector relative to the conditional feature vector, the operator behavior data semantic coding vector and the power grid operation status embedded coding vector are feature modulated to obtain the modulated operator behavior data semantic coding vector and the modulated power grid operation status embedded coding vector, including: calculating the first semantic contribution of the operator behavior data semantic coding vector relative to the conditional feature vector; calculating the second semantic contribution of the power grid operation status embedded coding vector to the conditional feature vector; normalizing the first semantic contribution and the second semantic contribution, and using the normalized first semantic contribution and second semantic contribution to perform weighted modulation on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector to obtain the modulated operator behavior data semantic coding vector and the modulated power grid operation status embedded coding vector.

[0018] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the first semantic contribution of the operator behavior data semantic encoding vector relative to the conditional feature vector is calculated, including: calculating the dot division vector between the operator behavior data semantic encoding vector and the conditional feature vector, and calculating the base-two logarithm of the absolute value of each eigenvalue in the dot division vector to obtain a semantic correlation representation vector; calculating the dot product vector between the semantic correlation representation vector and the operator behavior data semantic encoding vector, and calculating the exponential function value with base e and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the first semantic contribution.

[0019] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the semantic coding vector of the modulated operator behavior data, the embedded coding vector of the modulated power grid operation status and the conditional feature vector are subjected to cross-domain interactive coding based on the converter structure to obtain the operator behavior-power grid operation status significant interaction fusion representation vector, including: using the semantic coding vector of the modulated operator behavior data as the query vector, the embedded coding vector of the modulated power grid operation status as the key vector and the conditional feature vector as the value vector, and inputting the semantic coding vector of the modulated operator behavior data, the embedded coding vector of the modulated power grid operation status and the conditional feature vector into the feature-to-feature significant guided interaction module based on the converter structure to obtain the operator behavior-power grid operation status significant interaction fusion representation vector.

[0020] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, auxiliary prompt information is generated based on the joint features of the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector, including: feature splicing of the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector and then inputting the feature into the auxiliary prompt information decoding generator based on the RNN model to generate the auxiliary prompt information.

[0021] According to another aspect of the present application, a system for generating auxiliary prompt information based on knowledge reasoning is provided, which includes:

[0022] The data acquisition module is used to collect current grid operation status data through sensors and extract operator behavior data from interaction logs;

[0023] a data vectorization module, configured to perform vectorization processing on the current power grid operation state data and the operator behavior data respectively to obtain a power grid operation state embedded coding vector and an operator behavior data semantic coding vector;

[0024] a feature interaction coding module, configured to perform feature interaction coding on the operator behavior data semantic coding vector and the grid operation state embedded coding vector to obtain an operator behavior-grid operation state significant interaction fusion representation vector;

[0025] Prompt template acquisition module, used to obtain auxiliary prompt template questions;

[0026] A semantic encoding module, configured to perform semantic encoding on the auxiliary prompt template question to obtain a semantic encoding feature vector of the auxiliary modal prompt question;

[0027] An information generation module is used to generate auxiliary prompt information based on the joint features of the operator behavior-grid operation state significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector.

[0028] In the above-mentioned knowledge reasoning-based auxiliary prompt information generation system, the feature interaction coding module includes: an implicit feature extraction unit, used to extract the joint implicit features between the operator behavior data semantic coding vector and the power grid operation status embedded coding vector as a conditional feature vector; a feature modulation unit, used to perform feature modulation on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector based on the semantic contribution of the operator behavior data semantic coding vector and the power grid operation status embedded coding vector relative to the conditional feature vector to obtain the modulated operator behavior data semantic coding vector and the modulated power grid operation status embedded coding vector; a cross-domain interaction coding unit, used to perform cross-domain interaction coding based on a converter structure on the modulated operator behavior data semantic coding vector, the modulated power grid operation status embedded coding vector and the conditional feature vector to obtain the operator behavior-power grid operation status significant interaction fusion representation vector.

[0029] Compared with the existing technology, the present application provides a method and system for generating auxiliary prompt information based on knowledge reasoning. It uses artificial intelligence technology based on deep learning to extract semantic features and perform feature interaction fusion analysis on the current power grid operation status data and operator behavior data to dig out the implicit association between operator behavior and power grid operation status, provide a rich contextual background for the generation of auxiliary prompt information, and then combine the auxiliary prompt template problem to intelligently generate targeted auxiliary prompt information. In this way, by deeply analyzing the operation status of the power grid and the operation behavior of the personnel, auxiliary prompt information that conforms to the current situation can be dynamically generated, thereby improving the intelligence level of power grid operation monitoring and management, and enhancing the decision-making support ability of operators when facing complex situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0031] Figure 1 This is a flowchart of a method for generating auxiliary prompt information based on knowledge reasoning according to an embodiment of the present application.

[0032] Figure 2 Schematic diagram of data flow of the auxiliary prompt information generation method based on knowledge reasoning according to an embodiment of the present application.

[0033] Figure 3 This is a flowchart of sub-step S2 of the method for generating auxiliary prompt information based on knowledge reasoning according to an embodiment of the present application.

[0034] Figure 4 This is a flowchart of sub-step S3 of the method for generating auxiliary prompt information based on knowledge reasoning according to an embodiment of the present application.

[0035] Figure 5 This is a flowchart of sub-step S31 of the auxiliary prompt information generation method based on knowledge reasoning according to an embodiment of the present application.

[0036] Figure 6 This is a block diagram of an auxiliary prompt information generation system based on knowledge reasoning according to an embodiment of the present application. DETAILED DESCRIPTION

[0037] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0038] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0039] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0040] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0041] In the technical solution of the present application, a method for generating auxiliary prompt information based on knowledge reasoning is proposed. Figure 1 This is a flowchart of a method for generating auxiliary prompt information based on knowledge reasoning according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the auxiliary prompt information generation method based on knowledge reasoning according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the auxiliary prompt information generation method based on knowledge reasoning according to an embodiment of the present application includes the following steps: S1, collecting current power grid operation status data through sensors, and extracting operator behavior data from interaction logs; S2, vectorizing the current power grid operation status data and the operator behavior data respectively to obtain a power grid operation status embedded coding vector and an operator behavior data semantic coding vector; S3, performing feature interaction coding on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector to obtain an operator behavior-power grid operation status significant interaction fusion representation vector; S4, obtaining an auxiliary prompt template question; S5, semantically coding the auxiliary prompt template question to obtain an auxiliary modal prompt question semantic coding feature vector; S6, generating auxiliary prompt information based on the joint features of the operator behavior-power grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic coding feature vector.

[0042] In the above-mentioned knowledge-based reasoning method for generating auxiliary prompt information, step S1 involves collecting current grid operating status data via sensors and extracting operator behavior data from interaction logs. It should be understood that by monitoring and collecting various grid operating status data, such as voltage, current, frequency, and load, through sensors, auxiliary prompt information can be generated based on the latest grid status, thereby improving the accuracy and timeliness of the prompt information. The interaction log records the operator's specific operations, decisions, and response times during grid operation. By extracting operator behavior data from it, it is possible to effectively understand the operator's real-time behavior patterns and decision-making logic during grid operation, helping to better understand the operator's potential needs or questions, thereby further improving the relevance and practicality of the auxiliary prompt information. It is worth mentioning that in a specific example of this application, the sensor can be a current sensor, which is used to measure the current in a circuit and can be divided into two types: contact and non-contact. Contact current sensors, such as current transformers and current transformers, convert current into voltage or current output based on the principle of electromagnetic induction. Non-contact current sensors, such as magnetostrictive sensors and Hall sensors, detect current by sensing changes in a magnetic field.

[0043] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, in step S2, the current power grid operation status data and the operator behavior data are respectively vectorized to obtain a power grid operation status embedded coding vector and an operator behavior data semantic coding vector. It should be understood that since the current power grid operation status data and the operator behavior data usually have different data types and characteristics, in order to convert the two into a unified data form to facilitate information interaction and fusion between different data sources, the present application further vectorizes the current power grid operation status data and the operator behavior data. It is worth mentioning that vectorization processing is a process of converting non-vector format data into vector format, which is very useful in many fields, especially in scenarios such as graphic design, remote sensing image processing and power grid operation status data processing. For power grid operation status data, it is usually necessary to convert the sensor data involved into a vector format that can be analyzed and utilized through specific vectorization processing.

[0044] In a specific example of this application, Figure 3 As shown, the step S2 includes: S21, using the power grid operation state embedding coding matrix to embed the current power grid operation state data to obtain the power grid operation state embedding coding vector; S22, semantically encoding the operator behavior data to obtain the operator behavior data semantic coding vector.

[0045] Specifically, in step S21, the current power grid operation state data is embedded and encoded using the power grid operation state embedding coding matrix to obtain the power grid operation state embedded coding vector. That is, considering that the current power grid operation state data is usually discrete numerical data, the present application uses the power grid operation state embedding coding matrix to embed the current power grid operation state data to map the discrete data to a continuous vector space, and learn the intrinsic connection between each data item in the current power grid operation state data, generate a power grid operation state embedded coding vector, and thus realize the vectorization processing of the data. In one embodiment of the present application, the power grid operation state embedding coding matrix is ​​generated by 1D CNN network training. It is worth mentioning that the 1D CNN network (one-dimensional convolutional neural network) is a deep learning model, which is mainly used to process time series data or any other form of one-dimensional sequence data, and is particularly suitable for processing numerical data such as power grid operation state data.

[0046] Specifically, step S22 semantically encodes the operator behavior data to obtain a semantic encoding vector for the operator behavior data. It should be understood that the operator behavior data is typically textual data. Therefore, this application utilizes a pre-trained natural language model to semantically encode the operator behavior data. Word embedding encoding and a self-attention mechanism are used to capture semantic associations within the operator behavior data, extracting the contextual semantic meaning of the operator behavior data and thereby generating a semantic encoding vector for the operator behavior data. In an embodiment of this application, a Transformer-based pre-trained model is employed to semantically encode the operator behavior data, ensuring that the semantic information of the operator behavior data is fully preserved and expressed. It is worth mentioning that, in another embodiment of this application, semantic encoding of the operator behavior data can be implemented based on a probabilistic latent semantic analysis model. The probabilistic latent semantic analysis model is a topic model that can be used to discover implicit semantic structures in data. In the context of behavioral data, a probabilistic latent semantic analysis model can be used to identify behavioral patterns and encode them into semantic vectors. This approach can capture the semantic information of behavior by learning the topic distributions within the behavioral data and representing these distributions as vectors.

[0047] In the above-mentioned auxiliary prompt information generation method based on knowledge reasoning, the step S3 performs feature interaction encoding on the operator behavior data semantic encoding vector and the power grid operation status embedded encoding vector to obtain the operator behavior-power grid operation status significant interaction fusion representation vector. That is, in order to comprehensively utilize the multi-source complementary information of the power grid operation status data and the operator behavior data, the operator behavior data semantic encoding vector and the power grid operation status embedded encoding vector are further subjected to feature interaction encoding to achieve information complementarity and fusion between the two, thereby providing a more comprehensive background basis for the generation of auxiliary prompt information. It is worth mentioning that in order to optimize the feature interaction fusion effect, the present application proposes a feature fusion mechanism based on shared implicit features, which generates more discriminative and expressive feature representations by learning and utilizing the implicit association between the power grid operation status data and the operator behavior data.

[0048] In a specific example of this application, Figure 4 As shown, the step S3 includes: S31, extracting the joint implicit features between the operator behavior data semantic coding vector and the grid operation state embedded coding vector as a conditional feature vector; S32, based on the semantic contribution of the operator behavior data semantic coding vector and the grid operation state embedded coding vector relative to the conditional feature vector, performing feature modulation on the operator behavior data semantic coding vector and the grid operation state embedded coding vector to obtain a modulated operator behavior data semantic coding vector and a modulated grid operation state embedded coding vector; S33, performing cross-domain interactive coding based on a converter structure on the modulated operator behavior data semantic coding vector, the modulated grid operation state embedded coding vector and the conditional feature vector to obtain the operator behavior-grid operation state significant interaction fusion representation vector.

[0049] Specifically, if Figure 5 As shown, the step S31 includes: S311, inputting the operator behavior data semantic coding vector and the grid operation status embedded coding vector into a joint implicit feature capture network to obtain an operator behavior-grid operation status joint implicit feature vector; S312, performing feature activation based on a Sigmoid function on the operator behavior-grid operation status joint implicit feature vector to obtain the conditional feature vector.

[0050] Specifically, step S32 includes: calculating the first semantic contribution of the operator behavior data semantic coding vector relative to the conditional feature vector; calculating the second semantic contribution of the power grid operation state embedded coding vector to the conditional feature vector; normalizing the first semantic contribution and the second semantic contribution, and using the normalized first semantic contribution and the second semantic contribution to weighted modulate the operator behavior data semantic coding vector and the power grid operation state embedded coding vector to obtain the modulated operator behavior data semantic coding vector and the modulated power grid operation state embedded coding vector. Calculating the first semantic contribution of the operator behavior data semantic coding vector relative to the conditional feature vector includes: calculating the dotted division vector between the operator behavior data semantic coding vector and the conditional feature vector, and calculating the base-two logarithm of the absolute value of each eigenvalue in the dotted division vector to obtain a semantic relevance representation vector; calculating the dotted product vector between the semantic relevance representation vector and the operator behavior data semantic coding vector, and calculating the exponential function value with base e and the sum of the eigenvalues ​​of the dotted product vector as the exponent to obtain the first semantic contribution.

[0051] Specifically, step S33 performs a converter-structured cross-domain interaction encoding on the modulated operator behavior data semantic encoding vector, the modulated grid operation state embedded encoding vector, and the conditional feature vector to obtain the operator behavior-grid operation state significant interaction fusion representation vector. This includes: using the modulated operator behavior data semantic encoding vector as a query vector, the modulated grid operation state embedded encoding vector as a key vector, and the conditional feature vector as a value vector; inputting the modulated operator behavior data semantic encoding vector, the modulated grid operation state embedded encoding vector, and the conditional feature vector into a converter-structured inter-feature significant guided interaction module to obtain the operator behavior-grid operation state significant interaction fusion representation vector. More specifically, the modulated operator behavior data semantic encoding vector is multiplied by the transposed vector of the modulated grid operation state embedded encoding vector, and then divided by the square root of the length of the grid operation state embedded encoding vector to obtain an operator behavior-grid operation state query attention score matrix; and the operator behavior-grid operation state query attention score matrix is ​​passed through a softmax function and then multiplied by the conditional feature vector to obtain the operator behavior-grid operation state significant interaction fusion representation vector.

[0052] Accordingly, the operator behavior data semantic encoding vector and the grid operation state embedded encoding vector are processed using the following feature interaction encoding formula to obtain the operator behavior-grid operation state significant interaction fusion representation vector, wherein the feature interaction encoding formula is:

[0053]

[0054] X c =sigmoid(X J )

[0055]

[0056] X'1=X1w1

[0057] X'2=X2w2

[0058]

[0059] Wherein, X1 represents the semantic coding vector of the operator behavior data, X2 represents the embedded coding vector of the power grid operation status, Indicates adding by position, X J represents the joint implicit feature vector of operator behavior and grid operation status, W VT and b VT Represent the weight matrix and bias parameters of the neural network layer, tanh represents the hyperbolic tangent function, sigmoid represents the sigmoid activation function, X c represents the conditional eigenvector, and represents the i-th eigenvalue in the semantic coding vector of the operator behavior data, the i-th eigenvalue in the embedded coding vector of the power grid operation status, and the i-th eigenvalue in the conditional eigenvector, respectively; |·| is the absolute value symbol; log represents the logarithmic function with base 2; exp(·) represents the exponential function with base e; n represents the length of the semantic coding vector of the operator behavior data, and the lengths of the semantic coding vector of the operator behavior data, the embedded coding vector of the power grid operation status, and the conditional eigenvector are the same; CI1 and CI2 represent the first semantic contribution and the second semantic contribution, respectively; w1 and w2 represent the normalized first semantic contribution and the second semantic contribution, respectively; X1' and X2' represent the modulated semantic coding vector of the operator behavior data and the modulated embedded coding vector of the power grid operation status, respectively; (·) T represents the transpose of a vector, represents matrix multiplication operation, softmax is the normalized exponential function, X r A fusion representation vector representing the significant interaction between operator behavior and grid operation status is provided.

[0060] Specifically, a joint implicit feature capture network is first used to extract the shared implicit structure from the operator behavior data semantic encoding vector and the grid operation status embedded encoding vector. The extracted joint implicit features are then nonlinearly transformed using the Sigmoid function to emphasize key information and suppress irrelevant noise, creating the conditional feature vector. Next, the operator behavior data semantic encoding vector and the grid operation status embedded encoding vector are evaluated for correlation with the conditional feature vector to determine their importance to the final representation. This is then used to weight the original operator behavior data semantic encoding vector and grid operation status embedded encoding vector, effectively emphasizing key information while suppressing irrelevant information. Finally, the modulated operator behavior data semantic encoding vector is used as the query vector, the modulated grid operation status embedded encoding vector as the key vector, and the conditional feature vector as the value vector. The Transformer architecture is then used to guide significant interactions between features. Through the transformer architecture, the interactions between different features are simultaneously examined from multiple perspectives, promoting effective feature fusion, and ultimately generating a fusion representation vector of significant operator behavior-grid operation status interactions.

[0061] In the aforementioned knowledge-based reasoning method for generating auxiliary prompt information, step S4 involves obtaining auxiliary prompt template questions. Specifically, to ensure more standardized and accurate auxiliary prompt information is generated, auxiliary prompt template questions are further obtained to clarify the specific problems and needs that operators may face, thereby providing clear guidance for generating auxiliary prompt information. In the technical solution of this application, the auxiliary prompt template questions have been optimized and verified to cover common problem types and scenarios.

[0062] In the above-mentioned method for generating auxiliary prompt information based on knowledge reasoning, in step S5, the auxiliary prompt template question is semantically encoded to obtain a semantic encoding feature vector of the auxiliary modal prompt question. In a specific example of the present application, in order to convert the auxiliary prompt template question into a unified data form, the present application also uses a pre-trained natural language model to perform semantic encoding processing on the auxiliary prompt template question, so as to capture the semantic information in the auxiliary prompt template question, deeply understand the query intent of the question, and thus generate a semantic encoding feature vector of the auxiliary modal prompt question.

[0063] In the aforementioned knowledge-based reasoning method for generating auxiliary prompt information, step S6, generating auxiliary prompt information based on the combined features of the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector, includes: concatenating the features of the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector, and then inputting the concatenated features into an auxiliary prompt information decoder generator based on an RNN model to generate the auxiliary prompt information. Specifically, the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector are concatenated to integrate the consultation question description and real-time background information about grid operation, forming a complete feature representation of the current scenario. Then, an RNN model is used to construct an auxiliary prompt information decoder generator to decode and generate the concatenated fusion features. It is worth noting that RNN is a recurrent neural network, a deep learning model for processing sequential data. Its core concept is to pass each element in the sequence through a loop structure, so that the output of each element depends not only on the current input but also on the output of the previous element. This structure enables RNN to capture long-term dependencies in input data, effectively process and memorize historical output information, and gradually generate text sequences of auxiliary prompt information, realize dynamic generation of auxiliary prompt information, and ensure that the output information is coherent and logical.

[0064] In the technical solution of the present application, the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt problem semantic coding feature vector respectively represent the operator behavior and grid operation status semantic collaborative interaction coding features and the semantic coding features of the auxiliary modal prompt problem. After feature splicing, they are input into the auxiliary prompt information decoding generator based on the RNN model to generate auxiliary prompt information. Considering the differences in feature order and feature mode between the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt problem semantic coding feature vector, this causes the auxiliary prompt information decoding generator based on the RNN model to distort and mismatch the decoding of local features during its decoding and generation process, thereby affecting the generation accuracy of the auxiliary prompt information.

[0065] In response to the above technical problem, in one example of the present application, a process of concatenating the operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector and inputting the concatenated vector into an auxiliary prompt information decoding generator based on an RNN model to generate auxiliary prompt information includes:

[0066] Feature splicing is performed on the operator behavior-grid operation state significant interaction fusion representation vector and the auxiliary modal prompt question semantic coding feature vector to obtain an operation behavior-grid state-prompt question collaborative splicing feature coding vector;

[0067] The center distance and dispersion between each feature dimension pair in the operation behavior-grid status-prompt problem collaborative splicing feature coding vector are calculated to obtain the first-order correlation measurement matrix between collaborative splicing feature dimensions and the influence amplitude quantization matrix between collaborative splicing feature dimensions, which are expressed as:

[0068]

[0069] v i ,vj∈V

[0070] Among them, v i and v j represents the i-th position and j-th position of the collaborative splicing feature coding vector of the operation behavior-grid status-prompt problem, V represents the collaborative splicing feature coding vector of the operation behavior-grid status-prompt problem, M μ (i, j) represents the eigenvalue of the (i, j) position of the first-order correlation measurement matrix between the collaborative splicing feature dimensions, M σ (i, j) represents the eigenvalue at the (i, j) position of the influence amplitude quantization matrix between the collaborative splicing feature dimensions.

[0071] Based on the first-order correlation measurement matrix between the collaborative splicing feature dimensions and the influence amplitude quantification matrix between the collaborative splicing feature dimensions, a driven aggregation analysis based on a single path dependency is performed on the operation behavior-grid status-prompt problem collaborative splicing feature encoding vector to obtain the collaborative splicing feature basic dominant vector and the collaborative splicing feature disturbance influence vector, which are expressed as:

[0072]

[0073] Among them, ⊙ represents the Hadamard product, Represents matrix multiplication, M μ represents the first-order correlation measure matrix between the collaborative splicing feature dimensions, M σ represents the quantization matrix of the influence amplitude between the collaborative splicing feature dimensions, Softmax represents the normalized exponential function, V μ Denotes the basis dominant vector of collaborative splicing features, V σ represents the collaborative splicing feature perturbation influence vector.

[0074] An autocorrelation analysis is performed on the collaborative splicing feature base dominant vector and the collaborative splicing feature perturbation influence vector to obtain a collaborative splicing feature dominant-perturbation covariance spectrum matrix, which is expressed as:

[0075]

[0076] Among them, T represents the transpose of the vector, M ∑ represents the co-splicing feature dominance-perturbation covariation spectrum matrix.

[0077] Based on the collaborative splicing feature dominant-disturbance covariance spectrum matrix, the collaborative splicing feature basic dominant vector and the collaborative splicing feature disturbance influence vector are described by correlation constraints to obtain an optimized operation behavior-grid state-prompt problem collaborative splicing feature encoding vector, which is expressed as:

[0078]

[0079] Among them, α and β represent weight hyperparameters, and V′ represents the optimized operation behavior-grid status-prompt problem collaborative splicing feature encoding vector.

[0080] The optimized operation behavior-grid status-prompt problem collaborative splicing feature coding vector is input into an auxiliary prompt information decoding generator based on the RNN model to generate auxiliary prompt information.

[0081] That is, in the technical solution of the present application, on the basis of analyzing the core aggregation driving force of each eigenvalue of the operation behavior-grid state-prompt problem collaborative splicing feature coding vector through a single path dependency evaluation based on the mean and variance, a complex decomposition model is used to correlate and analyze the independent contribution of local structural features and local characteristic variables of the eigenvector. Thus, with the help of autocorrelation analysis, the limitations of the universal applicability of the basic assumptions are revealed to establish a descriptive model of how the correlation of structural elements affects the steady-state parameters of the basic homogeneous environment, thereby improving the simulation accuracy of local phenomena under specific situational constraints, and realizing the flattened correlation tracing of the impact of local structural features on the overall system behavior, so as to promote the consistency and robustness of the decoding of local features by the auxiliary prompt information decoding generator based on the RNN model. In this way, the generation accuracy of the auxiliary prompt information is improved.

[0082] In summary, a method for generating auxiliary prompt information based on knowledge reasoning based on the embodiment of the present application is explained, which uses artificial intelligence technology based on deep learning to perform semantic feature extraction and feature interaction fusion analysis on the current power grid operation status data and operator behavior data, so as to dig out the implicit association between operator behavior and power grid operation status, provide a rich contextual background for the generation of auxiliary prompt information, and then combine the auxiliary prompt template problem to intelligently generate targeted auxiliary prompt information. In this way, by deeply analyzing the operation status of the power grid and the operation behavior of the personnel, auxiliary prompt information that conforms to the current situation can be dynamically generated, thereby improving the intelligence level of power grid operation monitoring and management, and enhancing the decision-making support ability of operators when facing complex situations.

[0083] Furthermore, a knowledge-based reasoning-based auxiliary prompt information generation system is also provided.

[0084] Figure 6 FIG is a block diagram of an auxiliary prompt information generation system based on knowledge reasoning according to an embodiment of the present application. Figure 6 As shown, according to an embodiment of the present application, the auxiliary prompt information generation system 100 based on knowledge reasoning includes: a data acquisition module 110, which is used to collect current power grid operation status data through sensors and extract operator behavior data from interaction logs; a data vectorization module 120, which is used to vectorize the current power grid operation status data and the operator behavior data respectively to obtain a power grid operation status embedded coding vector and an operator behavior data semantic coding vector; a feature interaction coding module 130, which is used to perform feature interaction coding on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector to obtain an operator behavior-power grid operation status significant interaction fusion representation vector; a prompt template acquisition module 140, which is used to obtain an auxiliary prompt template question; a semantic coding module 150, which is used to semantically encode the auxiliary prompt template question to obtain an auxiliary modal prompt question semantic coding feature vector; an information generation module 160, which is used to generate auxiliary prompt information based on the joint features of the operator behavior-power grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic coding feature vector.

[0085] Here, those skilled in the art will appreciate that the specific operations of each module in the auxiliary prompt information generation system based on knowledge reasoning have been described in detail above. Figures 1 to 5 It has been introduced in detail in the description of the auxiliary prompt information generation method based on knowledge reasoning, and therefore, its repeated description will be omitted.

[0086] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for generating auxiliary prompt information based on knowledge reasoning, characterized in that: include: Collect current grid operation status data through sensors and extract operator behavior data from interaction logs; Performing vectorization processing on the current power grid operation state data and the operator behavior data respectively to obtain a power grid operation state embedded coding vector and an operator behavior data semantic coding vector; Performing feature interaction coding on the operator behavior data semantic coding vector and the grid operation state embedded coding vector to obtain an operator behavior-grid operation state significant interaction fusion representation vector; Get auxiliary prompt template questions; Performing semantic encoding on the auxiliary prompt template question to obtain an auxiliary modal prompt question semantic encoding feature vector; generating auxiliary prompt information based on the combined features of the operator behavior-grid operation state significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector; Among them, feature interaction coding is performed on the operator behavior data semantic coding vector and the power grid operation state embedded coding vector, including: extracting the joint implicit features between the operator behavior data semantic coding vector and the power grid operation state embedded coding vector as a conditional feature vector; based on the semantic contribution of the operator behavior data semantic coding vector and the power grid operation state embedded coding vector relative to the conditional feature vector, feature modulation is performed on the operator behavior data semantic coding vector and the power grid operation state embedded coding vector to obtain a modulated operator behavior data semantic coding vector and a modulated power grid operation state embedded coding vector; cross-domain interaction coding is performed on the modulated operator behavior data semantic coding vector, the modulated power grid operation state embedded coding vector and the conditional feature vector based on a converter structure to obtain the operator behavior-power grid operation state significant interaction fusion representation vector.

2. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 1, characterized in that: The current power grid operation state data and the operator behavior data are respectively vectorized to obtain a power grid operation state embedded coding vector and an operator behavior data semantic coding vector, including: Embedding the current power grid operation state data using a power grid operation state embedding coding matrix to obtain the power grid operation state embedding coding vector; The operator behavior data is semantically encoded to obtain a semantic encoding vector of the operator behavior data.

3. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 2, characterized in that: Extracting a joint implicit feature between the operator behavior data semantic encoding vector and the grid operation state embedded encoding vector as a conditional feature vector includes: Inputting the operator behavior data semantic encoding vector and the grid operation state embedding encoding vector into a joint implicit feature capture network to obtain an operator behavior-grid operation state joint implicit feature vector; A sigmoid function-based feature activation is performed on the operator behavior-grid operation state joint implicit feature vector to obtain the conditional feature vector.

4. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 3, characterized in that: Inputting the operator behavior data semantic encoding vector and the grid operation state embedding encoding vector into a joint implicit feature capture network to obtain an operator behavior-grid operation state joint implicit feature vector, including: The operator behavior data semantic coding vector and the power grid operation status embedding coding vector are fused according to position points and passed through a neural network layer based on a tanh function to obtain the operator behavior-power grid operation status joint implicit feature vector.

5. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 4, characterized in that: Based on the semantic contributions of the operator behavior data semantic coding vector and the power grid operation state embedded coding vector relative to the condition feature vector, feature modulation is performed on the operator behavior data semantic coding vector and the power grid operation state embedded coding vector to obtain modulated operator behavior data semantic coding vector and modulated power grid operation state embedded coding vector, including: Calculating a first semantic contribution of the operator behavior data semantic encoding vector relative to the conditional feature vector; Calculating a second semantic contribution of the power grid operation state embedded coding vector to the conditional feature vector; The first semantic contribution and the second semantic contribution are normalized, and the normalized first semantic contribution and the second semantic contribution are used to perform weighted modulation on the operator behavior data semantic coding vector and the power grid operation status embedded coding vector to obtain the modulated operator behavior data semantic coding vector and the modulated power grid operation status embedded coding vector.

6. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 5, characterized in that: Calculating a first semantic contribution of the operator behavior data semantic encoding vector relative to the conditional feature vector includes: Calculating a dotted-point vector between the operator behavior data semantic encoding vector and the conditional feature vector, and calculating a base-two logarithm of the absolute value of each eigenvalue in the dotted-point vector to obtain a semantic correlation representation vector; Calculate the dot product vector between the semantic relevance representation vector and the operator behavior data semantic encoding vector, and calculate an exponential function value with e as the base and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the first semantic contribution.

7. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 6, characterized in that: The modulated operator behavior data semantic coding vector, the modulated grid operation state embedded coding vector, and the conditional feature vector are subjected to cross-domain interactive coding based on a converter structure to obtain the operator behavior-grid operation state significant interaction fusion representation vector, including: Using the semantic coding vector of the modulated operator behavior data as a query vector, the embedded coding vector of the modulated power grid operation status as a key vector and the conditional feature vector as a value vector, the semantic coding vector of the modulated operator behavior data, the embedded coding vector of the modulated power grid operation status and the conditional feature vector are input into a converter structure-based feature-to-feature significant guided interaction module to obtain the operator behavior-power grid operation status significant interaction fusion representation vector.

8. The method for generating auxiliary prompt information based on knowledge reasoning according to claim 7, characterized in that: Based on the joint features of the operator behavior-grid operation state significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector, auxiliary prompt information is generated, including: The operator behavior-grid operation status significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector are feature spliced ​​and input into the auxiliary prompt information decoding generator based on the RNN model to generate the auxiliary prompt information.

9. A knowledge-based reasoning-based auxiliary prompt information generation system, characterized in that: include: The data acquisition module is used to collect current grid operation status data through sensors and extract operator behavior data from interaction logs; a data vectorization module, configured to perform vectorization processing on the current power grid operation state data and the operator behavior data respectively to obtain a power grid operation state embedded coding vector and an operator behavior data semantic coding vector; a feature interaction coding module, configured to perform feature interaction coding on the operator behavior data semantic coding vector and the grid operation state embedded coding vector to obtain an operator behavior-grid operation state significant interaction fusion representation vector; Prompt template acquisition module, used to obtain auxiliary prompt template questions; A semantic encoding module, configured to perform semantic encoding on the auxiliary prompt template question to obtain a semantic encoding feature vector of the auxiliary modal prompt question; An information generation module is used to generate auxiliary prompt information based on the joint features of the operator behavior-grid operation state significant interaction fusion representation vector and the auxiliary modal prompt question semantic encoding feature vector.

10. The auxiliary prompt information generation system based on knowledge reasoning according to claim 9 is characterized in that: The feature interaction encoding module includes: an implicit feature extraction unit, configured to extract a joint implicit feature between the operator behavior data semantic encoding vector and the grid operation state embedded encoding vector as a conditional feature vector; a feature modulation unit, configured to perform feature modulation on the operator behavior data semantic coding vector and the grid operation state embedded coding vector based on semantic contributions of the operator behavior data semantic coding vector and the grid operation state embedded coding vector relative to the condition feature vector to obtain a modulated operator behavior data semantic coding vector and a modulated grid operation state embedded coding vector; A cross-domain interaction coding unit is used to perform cross-domain interaction coding based on a converter structure on the modulated operator behavior data semantic coding vector, the modulated grid operation state embedded coding vector and the conditional feature vector to obtain the operator behavior-grid operation state significant interaction fusion representation vector.

Images