Production command case-oriented interactive deduction method and system

Abstract

The invention relates to the field of intelligent interaction, and provides a production command case-oriented interaction deduction method and system, and the method employs a text analysis and processing technology based on a deep learning model to carry out the sentence segmentation and semantic coding of the text description of a production command case, carries out the semantic coding of the text description of a current trigger event, and carries out the semantic coding of the text description of the current trigger event. In this way, the auxiliary prompt content is automatically generated according to the query matching representation between the text semantic information of the current trigger event and the sentence granularity description semantics of each production command case. Therefore, the production command case fragment most related to the current event can be identified more accurately, and more accurate and personalized auxiliary prompts are dynamically generated according to the current condition, so that power grid production command training and emergency drilling are better supported.

Description

Technical Field

[0001] This application relates to the field of intelligent interaction, and more specifically, to an interactive deduction method and system for production command cases. Background Art

[0002] Power grid production command involves multiple aspects such as the operation, maintenance, and fault handling of the power system, and is the key to ensuring the stable operation of the power system. With the continuous expansion of the power grid scale and the continuous progress of technology, the production command work is facing more and more challenges, such as complex and changeable operating environments, rapid response requirements for emergencies, etc. Therefore, through training and drills, various possible emergency situations can be simulated in advance to help commanders familiarize themselves with the emergency procedures and improve their response speed and handling efficiency in actual operations.

[0003] However, traditional power grid production command training and emergency drills mostly rely on the experience imparting of experts and on-site simulation drills. This traditional method is difficult to comprehensively cover all possible production scenarios. Especially when facing emergencies, its adaptability and flexibility are significantly insufficient. Moreover, due to limited expert resources and scarce training opportunities, it is difficult to meet the large-scale and high-frequency training needs. In addition, on-site drills not only cost a lot but also come with certain safety risks, restricting the frequency and complexity of the drills, resulting in insufficient ability to handle emergencies in actual operations and limited overall training effects.

[0004] Therefore, an interactive deduction solution for production command cases is desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an interactive deduction method and system for production command cases, which use text analysis and processing technologies based on deep learning models to perform clause processing and semantic encoding on the text descriptions of production command cases, and perform semantic encoding on the text descriptions of current trigger events, so as to automatically generate the auxiliary prompt content according to the query matching representation between the semantic information of the current trigger event text and the semantic information of each production command case sentence-level description. In this way, it is possible to more accurately identify the production command case segments most relevant to the current event, and dynamically generate more accurate and personalized auxiliary prompts according to the current situation to better support power grid production command training and emergency drills.

[0006] According to one aspect of this application, an interactive deduction method for production command cases is provided, which includes:

[0007] Respond to the simulation instruction of the production command case confirmed by the user, and start the simulation;

[0008] In response to receiving the intelligent assistance prompt instruction confirmed by the user, extract the text description of the production command case and extract the text description of the currently triggered event;

[0009] Based on the text description of the production command case and the text description of the currently triggered event, generate assistance prompt content;

[0010] Display the assistance prompt content;

[0011] Record all operations of the user and their results.

[0012] According to another aspect of the present application, there is provided an interactive deduction system for production command cases, which includes:

[0013] A simulation start response module, configured to start a simulation in response to receiving a simulation instruction of a production command case confirmed by a user;

[0014] A production command case currently triggered event text description extraction module, configured to extract the text description of the production command case and extract the text description of the currently triggered event in response to receiving the intelligent assistance prompt instruction confirmed by the user;

[0015] An assistance prompt content generation module, configured to generate assistance prompt content based on the text description of the production command case and the text description of the currently triggered event;

[0016] An assistance prompt content display module, configured to display the assistance prompt content;

[0017] A user operation result recording module, configured to record all operations of the user and their results.

[0018] Compared with the prior art, the interactive deduction method and system for production command cases provided by the present application adopt text analysis and processing technologies based on deep learning models to perform clause processing and semantic encoding on the text description of production command cases, and perform semantic encoding on the text description of currently triggered events, so as to automatically generate the assistance prompt content according to the query matching representation between the text semantics of the currently triggered event and the semantic descriptions of each sentence granularity of the production command cases. In this way, it can more accurately identify the production command case segments most relevant to the current event, and dynamically generate more accurate and personalized assistance prompts according to the current situation to better support power grid production command training and emergency drills. Description of the Drawings

[0019] The above and other objects, features, and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The drawings are used 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 to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 It is a flowchart of an interactive deduction method for production command cases according to an embodiment of the present application.

[0021] Figure 2 It is a flowchart of generating auxiliary prompt content based on the text description of the production command case and the text description of the current trigger event in the interactive deduction method for production command cases according to an embodiment of the present application.

[0022] Figure 3 It is a schematic diagram of data flow for generating auxiliary prompt content based on the text description of the production command case and the text description of the current trigger event in the interactive deduction method for production command cases according to an embodiment of the present application.

[0023] Figure 4 It is a flowchart of performing local neighborhood gating query matching on the sequence of the semantic information representation vectors of the current trigger event text and the semantic information representation vectors of the sentence-level descriptions of the production command cases to obtain the current event-production command case query matching implicit representation vector in the interactive deduction method for production command cases according to an embodiment of the present application.

[0024] Figure 5 It is a system block diagram of an interactive deduction system for production command cases according to an embodiment of the present application. Detailed implementation manners

[0025] Next, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein.

[0026] The stable operation of the power system is crucial for power grid production command, which covers multiple aspects such as system operation, maintenance, and fault handling. With the growth of the power grid scale and technological innovation, power grid production command faces many challenges, such as the need to cope with a complex and changeable operating environment and the requirement to quickly respond to emergencies. In order to improve the familiarity of commanders with the emergency process and the response speed and efficiency in actual operations, training and simulation exercises are particularly crucial.

[0027] However, traditional training and emergency drills mainly rely on the teachings of experts and on-site simulations. These methods have limitations in covering comprehensive production scenarios. Especially when dealing with emergencies, their flexibility and adaptability are insufficient. In addition, the limited availability of expert resources leads to a scarcity of training opportunities, making it difficult to meet large-scale and frequent training needs. The high cost of on-site drills and the associated safety risks limit the frequency and complexity of the drills, affecting the training effect and reducing the ability of commanders to handle emergencies in actual work.

[0028] Based on this, the present application proposes an interactive deduction method for production command cases. Figure 1 The flowchart of the interactive deduction method for production command cases according to an embodiment of the present application is as follows. Figure 1 As shown, the interactive deduction method for production command cases according to an embodiment of the present application includes: S110, starting a simulation in response to receiving a simulation instruction for a production command case confirmed by a user; S120, extracting the text description of the production command case and the text description of the currently triggered event in response to receiving the intelligent auxiliary prompt instruction confirmed by the user; S130, generating auxiliary prompt content based on the text description of the production command case and the text description of the currently triggered event; S140, displaying the auxiliary prompt content; S150, recording all operations of the user and their results.

[0029] In step S110, a simulation is started in response to receiving a simulation instruction for a production command case confirmed by a user. It should be understood that when a simulation instruction for a production command case confirmed by a user is received, the corresponding simulation environment will be started according to the content of the case. This environment simulates different actual operating states of the power grid, including fault states, emergency situations, etc., ensuring that the simulated scenario is as close as possible to the real operation scenario. The simulation allows power grid staff to perform various operations and decision-making exercises without interfering with the actual operation of the power grid. Through the simulation, staff can become familiar with the operation process of the power grid, master response measures in different states, and improve their ability to handle emergencies.

[0030] In step S120, the text description of the production command case and the text description of the currently triggered event are extracted in response to receiving the intelligent auxiliary prompt instruction confirmed by the user. In power grid production, each production command case is constructed based on real events in actual power grid operation. After the user confirms to start the intelligent auxiliary prompt function, the text description of the currently selected production command case and the text description of the currently triggered event will be extracted. These text descriptions contain key elements such as the background information, key events, and decision points of the case, which are the basis for generating the auxiliary prompt content.

[0031] In step S130, based on the text description of the production command case and the text description of the current triggering event, auxiliary prompt content is generated. In this application, based on the extracted text description, natural language processing and deep learning technologies can be used to generate auxiliary prompt content for the current event. These prompts may include emergency handling suggestions, operation steps, precautions, etc.

[0032] It should be understood that the auxiliary prompt content can provide specific guiding information to ensure that the operator performs tasks according to the correct steps. Especially in complex or emergency situations, accurate operation guidelines can prevent misoperations and reduce the likelihood of accidents. However, traditionally, a predefined rule set is used to generate prompts. Since the rules are usually static, they are difficult to change once set. This means that the system cannot adapt to new situations or unforeseen events, resulting in insufficient flexibility and adaptability.

[0033] Accordingly, in generating the auxiliary prompt content based on the text description of the production command case and the text description of the current triggering event, the technical concept of this application is to use text analysis and processing technologies based on a deep learning model to perform clause processing and semantic encoding on the text description of the production command case, and perform semantic encoding on the text description of the current triggering event, so as to automatically generate the auxiliary prompt content according to the query matching representation between the text semantic information of the current triggering event and the semantic descriptions of each production command case sentence granularity. In this way, it is possible to more accurately identify the production command case segments most relevant to the current event, and dynamically generate more accurate and personalized auxiliary prompts according to the current situation to better support power grid production command training and emergency drills.

[0034] Figure 2 It is a flowchart for generating auxiliary prompt content based on the text description of the production command case and the text description of the current triggering event in the interactive deduction method for production command cases according to an embodiment of the present application. Figure 3 It is a schematic diagram of data flow for generating auxiliary prompt content based on the text description of the production command case and the text description of the current triggering event in the interactive deduction method for production command cases according to an embodiment of the present application. As Figure 2 and Figure 3As shown, based on the text description of the production command case and the text description of the current triggering event, auxiliary prompt content is generated, including: S131, performing clause segmentation on the text description of the production command case to obtain a sequence of production command case sentence-granularity descriptions; S132, performing semantic encoding on each production command case sentence-granularity description in the sequence of production command case sentence-granularity descriptions to obtain a sequence of production command case sentence-granularity description semantic information representation vectors; S133, performing semantic encoding on the text description of the current triggering event to obtain a current triggering event text semantic information representation vector; S134, performing local neighborhood gated query matching on the current triggering event text semantic information representation vector and the sequence of production command case sentence-granularity description semantic information representation vectors to obtain a current event-production command case query matching implicit representation vector; S135, based on the current event-production command case query matching implicit representation vector, obtaining the auxiliary prompt content.

[0035] In step S131, clause segmentation is performed on the text description of the production command case to obtain a sequence of production command case sentence-granularity descriptions. Correspondingly, considering that the text description of the production command case contains a large amount of information at the production command sentence-granularity level, and each sentence may contain different power grid operation, target, or condition information, etc., based on this, in the technical solution of this application, clause segmentation is performed on the text description of the production command case to divide it into more concentrated and fine-grained information, obtaining a sequence of production command case sentence-granularity descriptions.

[0036] In step S132, semantic encoding is performed on each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions to obtain a sequence of production command case sentence granularity description semantic information representation vectors. Specifically, in the embodiments of the present application, performing semantic encoding on each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions to obtain a sequence of production command case sentence granularity description semantic information representation vectors includes: inputting each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions into a sentence semantic encoder to obtain the sequence of production command case sentence granularity description semantic information representation vectors. It should be understood that each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions contains implicit semantic information between the language contexts of specific power grid operations or conditions. Therefore, in order to better capture and extract the semantic meanings contained in each sentence granularity description, in the technical solution of the present application, semantic encoding is performed on each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions to obtain a sequence of production command case sentence granularity description semantic information representation vectors. In particular, in a specific embodiment of the present application, semantic encoding and extraction can be performed on each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions by using a sentence semantic encoder to obtain the sequence of production command case sentence granularity description semantic information representation vectors.

[0037] In step S133, semantic encoding is performed on the text description of the current trigger event to obtain a current trigger event text semantic information representation vector. Specifically, in the embodiments of the present application, performing semantic encoding on the text description of the current trigger event to obtain a current trigger event text semantic information representation vector includes: inputting the text description of the current trigger event into a current trigger event semantic encoder based on the Bert model to obtain the current trigger event text semantic information representation vector. It should be understood that the text description of the current trigger event contains the semantic information of the key trigger event. Therefore, in order to accurately understand the current needs of the user, in the technical solution of the present application, semantic encoding is performed on the text description of the current trigger event to refine and capture the important semantic features in the event, and a current trigger event text semantic information representation vector is obtained. In particular, in a specific embodiment of the present application, semantic extraction can be performed on the text description of the current trigger event by using a current trigger event semantic encoder based on the Bert model to obtain the current trigger event text semantic information representation vector.

[0038] In step S134, local neighborhood gated query matching is performed on the sequence of the current trigger event text semantic information representation vector and the production command case sentence granularity description semantic information representation vectors to obtain a current event-production command case query matching implicit representation vector. Accordingly, considering that the current trigger event text semantic information representation vector represents the semantic information of a specific event or state change occurring in the current production environment. The sequence of the production command case sentence granularity description semantic information representation vectors represents the power grid operation steps, objectives, expected results, and other important semantic information of each sentence in the case. Therefore, in order to effectively identify the association between the current trigger event and historical production command cases and thus generate useful auxiliary prompts, in the technical solution of this application, local neighborhood gated query matching is performed on the sequence of the current trigger event text semantic information representation vector and the production command case sentence granularity description semantic information representation vectors to obtain a current event-production command case query matching implicit representation vector.

[0039] Specifically, first, a one-dimensional convolutional encoding is performed on the sequence of semantic information representation vectors of the production command case sentence granularity descriptions to capture and extract the local semantic domain information of each sentence granularity description information, obtaining a sequence of local neighborhood semantic association feature vectors of the production command case sentence granularity descriptions. Next, the value embedding encoding matrix is used to process the semantic information representation vector (query feature) of the current trigger event text, generating a semantic value feature vector of the current trigger event text, providing data input for the subsequent processing of the attention mechanism. Then, the sequence of semantic information representation vectors of the production command case sentence granularity descriptions and the sequence of local neighborhood semantic association feature vectors of the production command case sentence granularity descriptions are respectively used as the sequence of key vectors, and the sequences of query vectors, value vectors, and key vectors are input into the first multi-head attention query module based on the transformer structure and the second multi-head attention query module based on the transformer structure to obtain a sequence of semantic query matching implicit vectors of the production command case sentence granularity descriptions and a sequence of semantic query matching implicit vectors of the production command case neighborhood granularity descriptions. That is, the first multi-head attention query module focuses on the features of the key vectors (semantic information representation vectors of the production command case sentence granularity descriptions) themselves, while the second multi-head attention query module considers the features of the neighborhood between the key vectors and their contexts, which can more comprehensively capture the internal connection between the production command case and the query data, thereby enhancing the accuracy of the matching. In this way, by combining the local details and overall features of the production command case, a comprehensive analysis of the information is achieved, and the reliability of the matching result is improved. Subsequently, the mean vectors of the sequence of semantic query matching implicit vectors of the production command case sentence granularity descriptions and the sequence of semantic query matching implicit vectors of the production command case neighborhood granularity descriptions are calculated to obtain a comprehensive semantic query matching feature vector of the production command case sentence granularity descriptions and a comprehensive semantic query matching feature vector of the production command case neighborhood granularity descriptions. Finally, the features of the two are fused to obtain a current event-production command case query matching implicit representation vector that can reflect the complex relationship information between the query and the key vectors.

[0040] Specifically, Figure 4 In the interactive deduction method for production command cases according to an embodiment of the present application, it is a flowchart for performing local neighborhood gated query matching on the semantic information representation vector of the current trigger event text and the sequence of semantic information representation vectors of the production command case sentence granularity descriptions to obtain a current event-production command case query matching implicit representation vector. As Figure 4As shown, a local neighborhood gated query matching is performed on the sequence of the semantic information representation vectors of the current trigger event text and the semantic information representation vectors of the production command case sentence granularity descriptions to obtain a current event-production command case query matching implicit representation vector, including: S1341, performing a one-dimensional convolutional encoding on the sequence of the semantic information representation vectors of the production command case sentence granularity descriptions to obtain a sequence of local neighborhood semantic association feature vectors of the production command case sentence granularity descriptions, where the scale of the one-dimensional convolutional encoding is an integer multiple of the length of the semantic information representation vector of the production command case sentence granularity description; S1342, processing the semantic information representation vector of the current trigger event text using a value embedding encoding matrix to obtain a semantic value feature vector of the current trigger event text; S1343, using the semantic information representation vector of the current trigger event text as a query vector, the semantic value feature vector of the current trigger event text as a value vector, and the sequence of the semantic information representation vectors of the production command case sentence granularity descriptions as a sequence of key vectors, inputting the sequence of the query vector, the value vector, and the key vectors into a first multi-head attention query module based on a transformer structure to obtain a sequence of semantic query matching implicit vectors of the production command case sentence granularity descriptions; S1344, using the semantic information representation vector of the current trigger event text as a query vector, the semantic value feature vector of the current trigger event text as a value vector, and the sequence of the local neighborhood semantic association feature vectors of the production command case sentence granularity descriptions as a sequence of key vectors, inputting the sequence of the query vector, the value vector, and the key vectors into a second multi-head attention query module based on a transformer structure to obtain a sequence of semantic query matching implicit vectors of the production command case sentence neighborhood granularity descriptions; S1345, calculating a position-wise mean vector of the sequence of the semantic query matching implicit vectors of the production command case sentence granularity descriptions and the sequence of the semantic query matching implicit vectors of the production command case sentence neighborhood granularity descriptions to obtain a semantic query matching feature vector of the production command case sentence granularity descriptions and a semantic query matching feature vector of the production command case sentence neighborhood granularity descriptions; S1346, fusing the semantic query matching feature vector of the production command case sentence granularity descriptions and the semantic query matching feature vector of the production command case sentence neighborhood granularity descriptions to obtain the current event-production command case query matching implicit representation vector.

[0041] More specifically, in the embodiment of the present application, processing the semantic information representation vector of the current trigger event text using a value embedding encoding matrix to obtain a semantic value feature vector of the current trigger event text includes: multiplying the semantic information representation vector of the current trigger event text by the value embedding encoding matrix, and then performing a position-wise addition of the obtained semantic embedding vector of the current trigger event text and a value bias vector to obtain the semantic value feature vector of the current trigger event text.

[0042] More specifically, in the embodiments of the present application, using the current trigger event text semantic information representation vector as the query vector, the current trigger event text semantic value feature vector as the value vector, and the sequence of the production command case sentence granularity description semantic information representation vectors as the sequence of key vectors, inputting the query vector, the value vector, and the sequence of key vectors into the first multi-head attention query module based on the transformer structure to obtain the sequence of production command case sentence granularity description semantic query matching hidden vectors, including: calculating the product between the current trigger event text semantic information representation vector and the transposed vector of the production command case sentence granularity description semantic information representation vector to obtain the current event-production command case association matrix, and then performing element-wise division of the current event-production command case association matrix by the square root of the length of the production command case sentence granularity description semantic information representation vector to obtain the current event-production command case association modulation matrix; inputting the current event-production command case association modulation matrix into the softmax function to obtain the current event-production command case association activation matrix; multiplying the current event-production command case association activation matrix by the current trigger event text semantic value feature vector to obtain the production command case sentence granularity description semantic query matching hidden vector.

[0043] More specifically, in the embodiments of the present application, fusing the production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector to obtain the current event-production command case query matching hidden representation vector, including: respectively calculating the element-wise subtraction, element-wise dot product, and element-wise addition between the production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector to obtain the production command case sentence granularity-sentence neighborhood granularity difference vector, the production command case sentence granularity-sentence neighborhood granularity dot product vector, and the production command case sentence granularity-sentence neighborhood granularity sum vector; performing a concatenation process on the production command case sentence granularity-sentence neighborhood granularity difference vector, the production command case sentence granularity-sentence neighborhood granularity dot product vector, and the production command case sentence granularity-sentence neighborhood granularity sum vector to obtain the production command case sentence granularity-sentence neighborhood granularity concatenated vector; performing one-dimensional convolutional encoding on the production command case sentence granularity-sentence neighborhood granularity concatenated vector, and then performing local window-based max pooling on the obtained production command case sentence granularity-sentence neighborhood granularity convolutional encoded vector to obtain the current event-production command case query matching hidden representation vector.

[0044] In the embodiments of the present application, specifically, a local neighborhood gated query matching is performed on the sequence of the semantic information representation vectors of the current trigger event text and the semantic information representation vectors of the production command case sentence granularity descriptions to obtain the current event-production command case query matching implicit representation vector, which can be expressed by the formula:

[0045] K = {k1, k2,..., k i ,..., k n}

[0046] K′ = Conv 1×l ({k1, k2,..., k i ,..., k n})

[0047] K′ = {k1′, k2′,..., k i ′,..., k m ′}

[0048] v v = v q W q + b q

[0049]

[0050] Wherein, K represents the sequence of the semantic information representation vectors of the production command case sentence granularity descriptions, k1, k2,..., k i ,..., k n are the 1st, 2nd,..., ith,..., nth production command case sentence granularity description semantic information representation vectors in the sequence of the production command case sentence granularity description semantic information representation vectors, Conv 1×l (·) is a one-dimensional convolutional encoding, l is an integer multiple of the length of the production command case sentence granularity description semantic information representation vector, K′ is the sequence of the production command case sentence granularity description local neighborhood semantic association feature vectors, k1′, k2′,..., k i ′,..., k m ′ are the 1st, 2nd,..., ith,..., mth production command case sentence granularity description local neighborhood semantic association feature vectors in the sequence of the production command case sentence granularity description local neighborhood semantic association feature vectors, v q is the semantic information representation vector of the current trigger event text, W q is the value embedding encoding matrix, b q is the value bias vector, v v is the semantic value feature vector of the current trigger event text, k i T and ki ′ T are k i and k i ′s transposed vectors respectively. d1 and d2 are the lengths of the semantic information representation vector of the production command case sentence granularity description and the local neighborhood semantic association feature vector of the production command case sentence granularity description. softmax(·) is the softmax function. v ki is the i-th production command case sentence granularity description semantic query matching hidden vector in the sequence of the production command case sentence granularity description semantic query matching hidden vectors. n is the number of feature vectors in the sequence of the production command case sentence granularity description semantic query matching hidden vectors. v ni is the i-th production command case neighborhood granularity description semantic query matching hidden vector in the sequence of the production command case neighborhood granularity description semantic query matching hidden vectors. m is the number of feature vectors in the sequence of the production command case neighborhood granularity description semantic query matching hidden vectors. v k is the production command case sentence granularity description semantic query matching feature vector. v n is the production command case neighborhood granularity description semantic query matching feature vector. ⊙ and are subtraction by position, multiplication by position, and addition by position respectively. [·, ·, ·] is the concatenation process. conv1D(·) is a one-dimensional convolutional encoding operation. MaxPool(·) is a max pooling operation. v p is the current event - production command case query matching hidden representation vector.

[0051] In step S135, the auxiliary prompt content is obtained by querying and matching the implicit representation vector based on the current event - production command case. Specifically, in the embodiment of the present application, obtaining the auxiliary prompt content by querying and matching the implicit representation vector based on the current event - production command case includes: inputting the implicit representation vector obtained by querying and matching the current event - production command case into an auxiliary prompt content generation module based on a decoder to obtain the auxiliary prompt content. That is, the implicit representation vector obtained by querying and matching the sequence of the current trigger event text semantic information representation vector and the production command case sentence - granularity description semantic information representation vector is decoded, so as to automatically generate the auxiliary prompt content. In this way, it is possible to more accurately identify the production command case segment most relevant to the current event, and dynamically generate more accurate and personalized auxiliary prompts according to the current situation, so as to better support power grid production command training and emergency drills. In particular, in a specific embodiment of the present application, a fully - connected layer is used to construct the decoder, and the decoder performs de - convolution decoding processing on the implicit representation vector obtained by querying and matching the current event - production command case to obtain the auxiliary prompt content.

[0052] In a preferred embodiment of the present application, in the process of inputting the implicit representation vector obtained by querying and matching the current event - production command case into an auxiliary prompt content generation module based on a decoder to obtain the auxiliary prompt content, semantic encoding enhancement is performed on the implicit representation vector obtained by querying and matching the current event - production command case. This process includes:

[0053] Subtracting the zero - measure basis element of the implicit representation vector obtained by querying and matching the current event - production command case from its Riemannian metric to generate an event - production command case query - matching isolated differential form, denoted as:

[0054] n = L - ‖V‖0

[0055] where L represents the Riemannian metric of the implicit representation vector obtained by querying and matching the current event - production command case, ‖V‖0 represents the zero - measure basis element of the implicit representation vector obtained by querying and matching the current event - production command case, and n represents the event - production command case query - matching isolated differential form;

[0056] Calculating the exterior product self - mapping of the event - production command case query - matching isolated differential form and its dual - space entropy to obtain an event - production command case query - matching connection coefficient, denoted as:

[0057] γ = log2(n 2 +n)

[0058] where γ represents the event - production command case query - matching connection coefficient;

[0059] Construct a cotangent bundle lift with the eigenvalues at each position of the implicitly represented vector matching the current event - production command case as the base and the isolated differential form matching the event - production command case query minus one as the exponent, and perform a contraction operation with the connection coefficients matching the event - production command case query to generate a principal bundle section matching the event - production command case query;

[0060] V1 = V ⊙(n-1) ⊙γ

[0061] where V ⊙(n-1) represents constructing a cotangent bundle lift with the eigenvalues at each position of the implicitly represented vector matching the current event - production command case as the base and the isolated differential form matching the event - production command case query minus one as the exponent, ⊙ represents the contraction operation, and V1 represents the principal bundle section matching the event - production command case query;

[0062] Perform a contraction operation on the result of the implicitly represented vector matching the current event - production command case query and the isolated differential form matching the event - production command case query minus one, and then process it through the Hodge dual operator to obtain the principal bundle section matching the event - production command case query, expressed as:

[0063] V2 = V⊙(n - 1)⊙n -1

[0064] where V2 represents the principal bundle section matching the event - production command case query;

[0065] Apply an exponential layer expansion transformation to the principal bundle section matching the event - production command case query to output the holonomy compensation quantity matching the event - production command case query, expressed as:

[0066] V3 = exp(V2)

[0067] where V3 represents the holonomy compensation quantity matching the event - production command case query;

[0068] Obtain the optimized implicitly represented vector matching the current event - production command case query by performing a position - by - position addition of the principal bundle section matching the event - production command case query and the holonomy compensation quantity matching the event - production command case query, expressed as:

[0069]

[0070] where represents position - by - position addition, and V′ represents the optimized implicitly represented vector matching the current event - production command case query.

[0071] For the current event - production command case query, match the multi-scale feature space topology of the implicit representation vector, and construct a dynamic gradient field model based on the convergence dimension of the attribute parameters of the feature set. Through the extraction of hierarchical correlation identifiers of the non-continuous cumulants of the current event - production command case query matching implicit representation vector in the spatial domain, parameter neighborhood anchoring is realized, and a gradient compensation excitation mechanism for the inversion field stability criterion is introduced to drive the regression distribution of the current event - production command case query matching implicit representation vector to meet the projection path locking requirement. Thereby, the feature set is prompted to perceive spatial correlation conduction, and by enhancing the semantic coupling resolvability, the text quality of the auxiliary prompt content obtained by optimizing the input of the current event - production command case query matching implicit representation vector into the decoder-based auxiliary prompt content generation module is optimized.

[0072] In step S140, the auxiliary prompt content is displayed. It should be understood that the generated auxiliary prompt content will be displayed on the user interface in real time to ensure that the command personnel can view and refer to it at any time during the operation. Through the intuitive interface, the user experience and operation convenience are enhanced, and the emergency response efficiency is improved.

[0073] In step S150, all operations of the user and their results are recorded. It should be understood that by recording the operation process and results of the user, the organization can more objectively evaluate the work performance and ability level of employees, and thus formulate a more scientific and reasonable training plan. At the same time, these recorded data can also provide strong support for the safety management and risk control of power grid production.

[0074] The interactive deduction method provided by this application can provide valuable practical experience for power grid staff in a relatively safe virtual environment to help them better prepare for challenges in reality. At the same time, it also provides an effective tool for the organization to evaluate the ability level of employees and carry out targeted training.

[0075] In summary, the interactive deduction method for production command cases based on the embodiments of this application is clarified. It uses text analysis and processing techniques based on deep learning models to perform clause processing and semantic encoding on the text descriptions of production command cases, and perform semantic encoding on the text description of the current triggering event. Based on this, the auxiliary prompt content is automatically generated according to the query matching representation between the semantic information of the current triggering event text and the semantic descriptions of each production command case sentence granularity. In this way, it is possible to more accurately identify the production command case segments most relevant to the current event, and dynamically generate more accurate and personalized auxiliary prompts according to the current situation to better support power grid production command training and emergency drills.

[0076] Figure 5 It is a system block diagram of an interactive deduction system for production command cases according to an embodiment of this application. As Figure 5As shown in the figure, the interactive deduction system 100 for production command cases according to an embodiment of the present application includes: a simulation start response module 110, configured to start a simulation in response to receiving a simulation instruction for a production command case confirmed by a user; a text description extraction module 120 for the current trigger event of the production command case, configured to extract the text description of the production command case and the text description of the current trigger event in response to receiving the intelligent assistance prompt instruction confirmed by the user; an assistance prompt content generation module 130, configured to generate assistance prompt content based on the text description of the production command case and the text description of the current trigger event; an assistance prompt content display module 140, configured to display the assistance prompt content; and a user operation result recording module 150, configured to record all operations of the user and their results.

[0077] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned interactive deduction system 100 for production command cases have been introduced in detail in the description of the Figures 1 to 4 interactive deduction method for production command cases, and therefore, the repeated description thereof will be omitted.

[0078] In summary, the interactive deduction system 100 for production command cases based on an embodiment of the present application is clarified. It uses text analysis and processing technologies based on deep learning models to perform sentence splitting processing and semantic encoding on the text description of production command cases, and perform semantic encoding on the text description of the current trigger event, so as to automatically generate the assistance prompt content according to the query matching representation between the text semantics of the current trigger event and the semantic descriptions of each production command case sentence granularity. In this way, it can more accurately identify the production command case segments most relevant to the current event, and dynamically generate more accurate and personalized assistance prompts according to the current situation to better support power grid production command training and emergency drills.

[0079] As described above, the interactive deduction system 100 for production command cases according to an embodiment of the present application can be implemented in various wireless terminals, such as a server for interactive deduction of production command cases. In one example, the interactive deduction system 100 for production command cases according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the interactive deduction system 100 for production command cases can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the interactive deduction system 100 for production command cases can also be one of many hardware modules of the wireless terminal.

[0080] Alternatively, in another example, the interactive deduction system 100 for production command cases and the wireless terminal may also be separate devices, and the interactive deduction system 100 for production command cases can be connected to the wireless terminal through a wired and / or wireless network and transmit interactive information in accordance with a predefined data format.

[0081] It should be noted that in the systems and methods of the present application, each module or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present application.

[0082] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present application. Thus, the present application is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0083] The above description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to the form disclosed herein. Although several example aspects and embodiments have been discussed above, those skilled in the art will recognize some of their variations, modifications, alterations, additions, and subcombinations.

Claims

1. An interactive deduction method for production command cases, characterized in that: include: In response to receiving a simulation instruction of a production command case confirmed by a user, starting a simulation; In response to receiving the intelligent auxiliary prompt instruction confirmed by the user, extracting a text description of the production command case and extracting a text description of the current triggering event; Generate auxiliary prompt content based on the text description of the production command case and the text description of the current trigger event; Display the auxiliary prompt content; All actions of the user and their results are recorded.

2. The interactive deduction method for production command cases according to claim 1 is characterized in that: Based on the text description of the production command case and the text description of the current trigger event, auxiliary prompt content is generated, including: Sentence processing is performed on the text description of the production command case to obtain a sequence of sentence granularity descriptions of the production command case; Semantically encoding each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions to obtain a sequence of production command case sentence granularity description semantic information representation vectors; Performing semantic encoding on the text description of the current trigger event to obtain a semantic information representation vector of the text of the current trigger event; Performing local neighborhood gated query matching on the sequence of the semantic information representation vector of the current triggering event text and the semantic information representation vector of the granular description of the production command case sentence to obtain a current event-production command case query matching implicit representation vector; The auxiliary prompt content is obtained based on the current event-production command case query matching implicit representation vector.

3. The interactive deduction method for production command cases according to claim 2 is characterized in that: Semantically encoding each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions to obtain a sequence of production command case sentence granularity description semantic information representation vectors, including: inputting each production command case sentence granularity description in the sequence of production command case sentence granularity descriptions into a sentence semantic encoder to obtain a sequence of production command case sentence granularity description semantic information representation vectors.

4. The interactive deduction method for production command cases according to claim 3 is characterized in that: The text description of the current trigger event is semantically encoded to obtain a current trigger event text semantic information representation vector, including: inputting the text description of the current trigger event into a current trigger event semantic encoder based on the Bert model to obtain the current trigger event text semantic information representation vector.

5. The interactive deduction method for production command cases according to claim 4 is characterized in that: Performing local neighborhood gated query matching on the sequence of the semantic information representation vector of the current triggering event text and the semantic information representation vector of the granular description of the production command case sentence to obtain the current event-production command case query matching implicit representation vector, including: Performing one-dimensional convolution coding on the sequence of the semantic information representation vectors of the granularity description of the production command case sentence to obtain a sequence of the local neighborhood semantic association feature vectors of the granularity description of the production command case sentence, wherein the scale of the one-dimensional convolution coding is equal to an integer multiple of the length of the semantic information representation vector of the granularity description of the production command case sentence; Using a value embedding coding matrix, the current trigger event text semantic information representation vector is processed to obtain a current trigger event text semantic value feature vector; Taking the semantic information representation vector of the current trigger event text as the query vector, the semantic value feature vector of the current trigger event text as the value vector and the sequence of the semantic information representation vector of the granular description of the production command case sentence as the sequence of key vectors, the query vector, the value vector and the sequence of the key vectors are input into a first multi-head attention query module based on a converter structure to obtain a sequence of implicit vectors matching the semantic query of the granular description of the production command case sentence; Using the semantic information representation vector of the current trigger event text as a query vector, the semantic value feature vector of the current trigger event text as a value vector, and the sequence of the local neighborhood semantic association feature vectors of the production command case sentence granularity description as a sequence of key vectors, the query vector, the value vector, and the sequence of the key vectors are input into a second multi-head attention query module based on a converter structure to obtain a sequence of implicit vectors matching the production command case sentence neighborhood granularity description semantic query; Calculate the positional mean vector of the sequence of the production command case sentence granularity description semantic query matching implicit vectors and the sequence of the production command case sentence neighborhood granularity description semantic query matching implicit vectors to obtain the production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector; The production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector are integrated to obtain the current event-production command case query matching implicit representation vector.

6. The interactive deduction method for production command cases according to claim 5 is characterized in that: The current trigger event text semantic information representation vector is processed using a value embedding coding matrix to obtain a current trigger event text semantic value feature vector, including: multiplying the current trigger event text semantic information representation vector with the value embedding coding matrix, and then adding the obtained current trigger event text semantic embedding vector and the value bias vector by position to obtain the current trigger event text semantic value feature vector.

7. The interactive deduction method for production command cases according to claim 6 is characterized in that: The method uses the semantic information representation vector of the current trigger event text as a query vector, the semantic value feature vector of the current trigger event text as a value vector, and the sequence of the semantic information representation vector of the production command case sentence granularity description as a sequence of key vectors, and inputs the query vector, the value vector, and the sequence of the key vectors into a first multi-head attention query module based on a converter structure to obtain a sequence of implicit vectors matching the semantic query of the production command case sentence granularity description, including: After calculating the product between the current trigger event text semantic information representation vector and the transposed vector of the production command case sentence granularity description semantic information representation vector to obtain a current event-production command case association matrix, the current event-production command case association matrix is ​​divided by the square root of the length of the production command case sentence granularity description semantic information representation vector by position to obtain a current event-production command case association modulation matrix; Input the current event-production command case association modulation matrix into a softmax function to obtain a current event-production command case association activation matrix; The current event-production command case association activation matrix is ​​multiplied by the current trigger event text semantic value feature vector to obtain the production command case sentence granularity description semantic query matching implicit vector.

8. The interactive deduction method for production command cases according to claim 7 is characterized in that: The production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector are integrated to obtain the current event-production command case query matching implicit representation vector, including: Respectively calculating the positional subtraction, positional dot product, and positional addition between the production command case sentence granularity description semantic query matching feature vector and the production command case sentence neighborhood granularity description semantic query matching feature vector to obtain a production command case sentence granularity-sentence neighborhood granularity difference vector, a production command case sentence granularity-sentence neighborhood granularity dot product vector, and a production command case sentence granularity-sentence neighborhood granularity sum vector; Cascading the sentence granularity-sentence neighborhood granularity difference vector of the production command case, the sentence granularity-sentence neighborhood granularity dot product vector of the production command case, and the sentence granularity-sentence neighborhood granularity sum vector of the production command case to obtain a production command case sentence granularity-sentence neighborhood granularity cascade vector; After performing one-dimensional convolution encoding on the sentence granularity-sentence neighborhood granularity cascade vector of the production command case, the obtained production command case sentence granularity-sentence neighborhood granularity convolution encoding vector is subjected to local window-based maximum pooling processing to obtain the current event-production command case query matching implicit representation vector.

9. The interactive deduction method for production command cases according to claim 8 is characterized in that: Based on the current event-production command case query matching implicit representation vector, the auxiliary prompt content is obtained, including: inputting the current event-production command case query matching implicit representation vector into a decoder-based auxiliary prompt content generation module to obtain the auxiliary prompt content.

10. An interactive deduction system for production command cases, characterized in that: include: A simulation start response module, for starting a simulation in response to receiving a simulation instruction of a production command case confirmed by a user; A production command case current trigger event text description extraction module, configured to extract the text description of the production command case and the text description of the current trigger event in response to receiving the intelligent auxiliary prompt instruction confirmed by the user; An auxiliary prompt content generation module, used to generate auxiliary prompt content based on the text description of the production command case and the text description of the current trigger event; An auxiliary prompt content display module, used to display the auxiliary prompt content; The user operation result recording module is used to record all operations of the user and their results.

Images