Electric power operation and maintenance emergency service remote decision interaction system and method

Abstract

The invention relates to the technical field of electric power engineering technology research and development, and discloses an electric power operation and maintenance emergency service remote decision interaction system and method, which introduces a causal inference model to carry out deep analysis on multi-source data, realizes advanced deduction of a fault risk conduction chain and virtual emergency scene construction, and is embedded with an electric power system physical rule to realize remote decision interaction of the electric power operation and maintenance emergency service. The risk deduction and disposal scheme generation process is ensured to strictly conform to the power grid operation rule, the output result is physically credible and has clear causal interpretation, and the scientificity and reliability of decision making are greatly enhanced. Furthermore, for a long-term existing data collaboration barrier, a privacy computing protocol is innovatively adopted to interact with an external system, so that the generation and computing of a joint disposal plan can be completed on the premise that the privacy of original data does not need to be shared, safe and efficient multi-department emergency collaboration is truly realized, and the safety and reliability of the system are improved. And the overall handling capability of the complex power emergency event is fundamentally improved.

Description

A remote decision-making and interaction system and method for power operation and maintenance emergency services Technical Field

[0001] This invention relates to the field of power engineering technology, and in particular to a remote decision-making and interactive system and method for power operation and maintenance emergency services. Background Technology

[0002] Traditional power emergency response models rely heavily on the personal experience of dispatchers and on-site personnel. Typically, after the monitoring system issues an alarm, personnel need to piece together the on-site situation through telephone communication and by consulting multiple independent systems (such as SCADA, GIS, and meteorological systems) before formulating a response plan based on experience. This model suffers from problems such as delayed response, incomplete information, and strong subjectivity in decision-making, and is prone to delaying the golden window for handling complex faults.

[0003] While some digital emergency management systems have emerged in existing technologies, they are essentially "passive response" and "static matching," unable to provide effective risk warnings before faults occur, nor can they generate optimized solutions for complex scenarios involving non-standard and multi-factor coupling. Secondly, the contingency plan databases upon which these systems rely are slowly updated, making it difficult to adapt to dynamic changes in power grid topology and operating modes. More importantly, these systems are often information silos, lacking the ability to collaborate with real-time, reliable data from key external departments such as meteorology, forestry, and transportation, which means that the developed response plans may be unenforceable due to external constraints. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a remote decision-making and interaction system and method for power operation and maintenance emergency services, so as to solve the technical problems of poor effectiveness in risk warning, emergency decision-making and collaborative execution during power operation and maintenance emergency services.

[0005] The first aspect of this invention discloses a remote decision-making and interaction system for power operation and maintenance emergency services. The system includes: a data fusion module for accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; a risk simulation module for outputting potential fault risk transmission chains based on the first integrated data using a causal inference model embedded with the physical rules of the power system, and generating corresponding virtual emergency scenarios; a collaborative handling module for performing collaborative computing operations with at least one external system based on the virtual emergency scenarios through a privacy computing protocol to generate joint handling plans; wherein the external systems are determined based on external data required by the virtual emergency scenarios; and a decision-making interaction module for providing a decision-making interaction interface, configuring the interface to visually display the fault risk transmission chains, virtual emergency scenarios, and joint handling plans, and providing a human-computer interaction command entry point.

[0006] Furthermore, the causal inference model is a physical information neural network; the loss function of the physical information neural network includes a data fitting term and a physical rule constraint term; wherein, the physical rule constraint term is constructed based on the differential equations of power flow calculation in the power system; during the training process, the physical rule constraint term is used to ensure that the output result of the neural network satisfies the power balance relationship of the power flow calculation at any node in the power grid topology.

[0007] Furthermore, the physical information neural network uses a graph structure to represent the power grid topology, where the nodes of the graph correspond to power equipment and the edges represent electrical connections.

[0008] Furthermore, the loss function also includes a graph structure learning loss term, used to directly optimize the adjacency matrix of the graph structure; the graph structure learning loss term includes: a reconstruction loss for reconstructing the observation data based on the graph structure and node states; the observation data includes real-time and historical measurement data accessed from the data fusion module; a norm regularization term for the adjacency matrix of the graph structure; and a physical consistency constraint term to ensure that the evolution of node states operating on the graph structure conforms to the power flow calculation differential equation.

[0009] Furthermore, the loss function also includes an intervention prediction loss term; the training data of the causal inference model includes power grid observation data used to train the model's basic representation and reconstruction capabilities, as well as adversarial samples generated by simulating various fault intervention measures based on the observation data and power grid topology, specifically for optimizing the model's intervention prediction capabilities; the intervention prediction loss term is constructed by calculating the difference between the causal inference model's predicted results of the system state after intervention and the actual state.

[0010] Furthermore, the loss function also includes a causal interpretability constraint term; the causal interpretability constraint term is used to maximize the mutual information between the feature vector of the hidden layer of the physical information neural network and a set of predefined causal mediator variables of the power system; the causal mediator variables of the power system represent intermediate state variables that can characterize the fault risk transmission process and have clear physical meaning in the fault risk transmission chain.

[0011] Furthermore, the privacy computing protocol is a zero-knowledge proof protocol; the zero-knowledge proof protocol is implemented based on a trusted execution environment; before executing the collaborative computing operation, the collaborative processing module is also used to receive a verifiable statement generated by an external system based on zero-knowledge proof that has completed a specific risk inspection, and to trigger the collaborative computing operation after the verification is passed.

[0012] Furthermore, the collaborative computing operation process includes: interacting with an external system under the privacy computing protocol framework to obtain first information calculated based on the internal data of the external system; the first information includes at least one of the following: local risk assessment results for the virtual emergency scenario, status and scheduling feasibility information of external controllable resources, and key constraints of the external environment; based on the type of the virtual emergency scenario, matching one or more basic contingency plan frameworks from the contingency plan library, and using the internal power grid status and the first information as parameters, calling an optimization computing model to fill, adjust, and evaluate the basic contingency plan frameworks to generate the final joint response plan.

[0013] Furthermore, the collaborative processing module is also used to send a transaction containing participant identifiers, data element information, and contribution evaluation parameters to a smart contract deployed on the blockchain network for quantifying data contributions after the collaborative computing operation is completed. The smart contract has predefined contribution evaluation rules, and automatically calculates and issues tokens representing data contributions based on the received transaction, and records the issuance in the blockchain ledger.

[0014] The second aspect of this invention discloses a remote decision-making interaction method for power operation and maintenance emergency services. This method is applied to the system disclosed in the first aspect. The method includes: accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; outputting potential fault risk transmission chains based on the first integrated data through a causal inference model embedded with the physical rules of the power system, and generating corresponding virtual emergency scenarios; based on the virtual emergency scenarios, performing collaborative computing operations with at least one external system through a privacy computing protocol to generate a joint response plan; wherein the external system is determined based on the external data required by the virtual emergency scenarios; and visually displaying the fault risk transmission chains, the virtual emergency scenarios, and the joint response plan.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: By introducing a causal inference model to conduct in-depth analysis of multi-source data, this invention enables advanced simulation of the fault risk transmission chain and the construction of virtual emergency scenarios, significantly improving the initiative in response. Furthermore, by embedding physical rules of the power system, the model ensures that the risk simulation and response plan generation process strictly conforms to the operating laws of the power grid. Its output results are not only physically reliable but also possess clear causal explanatory power, greatly enhancing the scientific rigor and reliability of decision-making. Moreover, addressing long-standing data collaboration barriers, this invention innovatively employs a privacy computing protocol to interact with external systems. This allows for the generation and calculation of joint response plans without sharing original data, ensuring confidentiality and truly achieving safe and efficient multi-department emergency collaboration, fundamentally improving the overall handling capability of complex power emergency events. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the embodiments of the present invention and form part of this application, do not constitute a limitation thereof. In the drawings: Figure 1 is a schematic diagram of the structure of a remote decision-making and interactive system for power operation and maintenance emergency services disclosed in Embodiment 1 of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example 1: The first aspect of this invention discloses a remote decision-making interaction system for power operation and maintenance emergency services, which relates to power engineering technology research and development. Please refer to Figure 1, which is a structural schematic diagram of a remote decision-making interaction system for power operation and maintenance emergency services disclosed in this embodiment. The system includes: a data fusion module, used to access and integrate multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; a risk inference module, used to output potential fault risk transmission chains based on the first integrated data using a causal inference model embedded with the physical rules of the power system, and generate corresponding virtual emergency scenarios; a collaborative handling module, used to perform collaborative computing operations with at least one external system based on the virtual emergency scenarios through a privacy computing protocol to generate joint handling plans; wherein the external systems are determined based on the external data required by the virtual emergency scenarios; and a decision-making interaction module, used to provide a decision-making interaction interface, configured to visually display the fault risk transmission chain, the virtual emergency scenarios, and the joint handling plans, and provide a human-computer interaction command entry point.

[0019] Specifically, in this embodiment of the invention, the power grid monitoring data mainly includes real-time measurement information from the SCADA (Supervisory Control and Data Acquisition) system, such as the temperature and voltage of each node device, the active and reactive power of branches, the status and position of circuit breakers and disconnectors, and the action signals and alarm information of relay protection and automatic devices. Meteorological and environmental data are obtained from professional service agencies, covering meteorological elements such as temperature, humidity, wind speed, wind direction, and precipitation in the target area, as well as environmental monitoring information such as lightning location, satellite cloud images, and landslide warnings. Data from external collaborative departments is dynamically accessed according to the emergency scenario. For example, in the scenario of preventing wildfires, data on forest fire risk levels, hotspot monitoring coordinates, and vegetation types are accessed from the forestry department; in the case of responding to external damage, construction permits and large machinery location information may be accessed from the planning or housing and construction departments.

[0020] The data integration process described above involves data preprocessing and correlation fusion. First, the data streams are cleaned and standardized, with timestamps unified to the same time scale, and unstructured data (such as text alarms) converted into structured formats. Second, using the power grid physical topology and geographic information system as a common spatial reference, data such as equipment measurements, meteorological grids, and external events are mapped to a unified coordinate system, establishing multi-dimensional correlations. For example, the real-time load data of a transmission line is correlated and bound to the real-time wind speed of its corridor area and the coordinates of nearby construction points. Finally, a spatiotemporally synchronized and correlated dataset is generated, which constitutes a unified, high-quality input for subsequent risk simulation analysis.

[0021] It should be noted that the process of generating the corresponding virtual emergency scenario is mainly based on and uses the potential fault risk transmission chain output by the aforementioned causal inference model as its core input. Specifically, the risk simulation module first analyzes the risk chain output by the model. This chain is essentially a structured causal graph, which clearly identifies the source of the risk, such as the leaning of trees and bamboo in a specific section under strong winds; key transmission links, such as changes in conductor sag leading to ground discharge; and potential final fault consequences, such as triggering grid protection actions and causing local power outages. Generating a virtual emergency scenario is to transform this abstract causal logic chain into a concrete, contextualized, and decision-making-rich dynamic digital sandbox. The system automatically associates and injects specific parameters from the first integrated data based on the key nodes and paths in the chain. For example, strong winds are concretized as specific wind speed ranges and wind directions in meteorological data, and specific sections are located to their precise coordinates and feeders in the power grid geographic information system. At the same time, the scenario estimates the possible time window for fault evolution and, based on the real-time operating status of the power grid, simulates and simulates secondary impacts such as power flow shifts, voltage limits exceeding limits, and load losses after a fault occurs.

[0022] The resulting virtual emergency scenario is a comprehensive descriptive framework that integrates a clear source of risk, a clear evolution logic, specific spatiotemporal boundaries, and quantitative impact assessment, providing a key hub for closed-loop decision-making from risk perception to response actions.

[0023] Furthermore, the causal inference model is a physical information neural network; the loss function of the physical information neural network includes a data fitting term and a physical rule constraint term; wherein, the physical rule constraint term is constructed based on the differential equations of power flow calculation in the power system; during the training process, the physical rule constraint term is used to ensure that the output result of the neural network satisfies the power balance relationship of the power flow calculation at any node in the power grid topology.

[0024] Furthermore, the physical information neural network uses a graph structure to represent the power grid topology, where the nodes of the graph correspond to power equipment and the edges represent electrical connections.

[0025] Furthermore, the loss function also includes a graph structure learning loss term, used to directly optimize the adjacency matrix of the graph structure; the graph structure learning loss term includes: a reconstruction loss for reconstructing the observation data based on the graph structure and node states; the observation data includes real-time and historical measurement data accessed from the data fusion module; a norm regularization term for the adjacency matrix of the graph structure; and a physical consistency constraint term to ensure that the evolution of node states operating on the graph structure conforms to the power flow calculation differential equation.

[0026] Furthermore, the loss function also includes an intervention prediction loss term; the training data of the causal inference model includes power grid observation data used to train the model's basic representation and reconstruction capabilities, as well as adversarial samples generated by simulating various fault intervention measures based on the observation data and power grid topology, specifically for optimizing the model's intervention prediction capabilities; the intervention prediction loss term is constructed by calculating the difference between the causal inference model's predicted results of the system state after intervention and the actual state.

[0027] Furthermore, the loss function also includes a causal interpretability constraint term; the causal interpretability constraint term is used to maximize the mutual information between the feature vector of the hidden layer of the physical information neural network and a set of predefined causal mediator variables of the power system; the causal mediator variables of the power system represent intermediate state variables that can characterize the fault risk transmission process and have clear physical meaning in the fault risk transmission chain.

[0028] In this embodiment of the invention, the training of the causal inference model is driven by a comprehensive loss function. The multi-component structure of this loss function is designed to systematically address the fundamental shortcomings of purely data-driven models in high-reliability power applications.

[0029] Specifically, the core of the model is constructed as a physical information neural network. The fundamental reason for this is that the behavior of a power system must strictly adhere to physical laws. Relying solely on data fitting may lead to results that violate fundamental operational principles, resulting in decision failure. Therefore, in the loss function, in addition to the data fitting term that minimizes the error between predicted and actual measurements, this invention also introduces a physical rule constraint term. This term is based on the core differential equations of power flow calculation, preferably the nodal power balance equations. During training, this term acts as a strong constraint, ensuring that the output of the neural network at any grid topology node (such as voltage amplitude and phase angle) strictly satisfies this power balance relationship. Through these settings, each deduction of the model operates within a physically feasible solution space, and the generated fault chains and scenarios possess solid credibility at the level of fundamental principles.

[0030] Secondly, given the structural characteristics of the power grid itself, this invention further configures the neural network to use a graph structure to represent the power grid topology, where nodes represent equipment such as generators, loads, or buses, and edges represent connections such as transmission lines or transformers. This structure is adopted because the propagation of faults or risks essentially occurs along topological paths, and the graph structure provides the most natural framework for explicitly modeling this propagation.

[0031] To further optimize this structure, this invention incorporates a graph structure learning loss term into the loss function. This loss term consists of three parts: first, a reconstruction loss, which forces the model to use the learned graph structure and node states to inversely analyze the massive amounts of observed SCADA measurement data, ensuring that the graph can interpret reality; second, an L1 norm regularization term, which acts on the adjacency matrix, penalizing excessive connections to promote the model's discovery of a sparse causal graph that conforms to the engineering prior of localized propagation of power faults; and third, a physical consistency constraint term on the graph, ensuring that the node state evolution simulated on the dynamic graph structure (such as power flow transition after a fault) still satisfies the power flow equations. Through these operations, the model not only passively fits the data but also actively discovers the strength of potential causal influences between devices and implicit in the data, and automatically constructs an easily interpretable and physically consistent risk propagation network.

[0032] To further empower emergency decision-making, the loss function of this invention also integrates an intervention prediction loss term. It is understood that the core of emergency response is making the optimal choice among different measures; therefore, the model must possess the ability to predict how the system state will change if a certain measure is taken. Thus, the training data not only includes real historical data but also generates adversarial samples simulating various scheduling, isolation, and load shedding operations by modifying the graph adjacency matrix (e.g., masking edges representing a certain line) or node attributes. This loss term specifically trains the model's counterfactual reasoning ability by calculating the difference between the model's predictions of these post-intervention states and high-fidelity simulations or post-event real data. Based on the above settings, the model can quantitatively evaluate the effectiveness of different contingency plans in advance, providing core algorithmic support for generating scientific and comparable joint response plans.

[0033] Furthermore, to address concerns about the "black box" nature of artificial intelligence models, this invention introduces a causal interpretability constraint. This constraint, by maximizing the mutual information between the hidden layer features of the model and a set of predefined causal mediator variables with clear engineering significance (such as line load rate, voltage stability margin, and protection operation probability) during training, forces the neural network to align its internal abstract representation with concepts understandable to experts or staff. When the model outputs a risk chain such as "tree obstruction leads to line overload, ultimately triggering protection tripping," the change curve of the predicted line load rate corresponding to the "overload" link in the chain can be traced and displayed, improving the transparency, verifiability, and trustworthiness of the abstract model's reasoning process.

[0034] Through the above operations, the loss function in this embodiment of the invention deeply integrates the four core requirements of physical conservation, graph structure learning, intervention inference, and interpretability, systematically shaping a powerful and reliable causal intelligent engine specifically for power emergency decision-making. This not only improves the accuracy of predictions but also fundamentally guarantees the physical feasibility, logical sparsity, decision predictability, and process interpretability of the output, thus providing high-quality assurance for the core risk warning and solution generation functions.

[0035] In its implementation, the causal inference model is specifically implemented as a physical information graph neural network designed for power networks. The model's input is the first integrated data generated by the data fusion module. This data, after standardization, is input into the model in a structured graph form. Specific inputs include, but are not limited to, an adjacency matrix defining the connection relationships of power grid equipment, a node feature matrix composed of real-time measurements of each node and associated environmental indicators, and a global feature vector describing the overall system state. The model outputs multi-level structured prediction results, including both the failure risk probability of each device node and directed weighted edges representing the strength of causal influence between devices, which can be automatically combined into a risk transmission subgraph. Ultimately, these results are integrated into a comprehensive virtual emergency scenario description containing the estimated failure sequence, impact range, and key evolution nodes.

[0036] The model's network architecture employs an encoder, processor, and decoder framework to deeply integrate physical rules. The encoder, composed of a multi-layer graph neural network, is responsible for mapping the input graph data to a high-dimensional latent space to capture complex spatial relationships. The processor, as the core, includes a differentiable physical constraint layer. This layer uses the previous output as input to directly calculate the power imbalance residuals conforming to the power flow differential equations, feeding these residuals back into the training process as part of the physical loss, while simultaneously providing physical corrections to subsequent layers of the network. Finally, the decoder, composed of fully connected layers, is responsible for decoding the physically corrected latent representation into specific prediction targets such as risk probabilities and causal strengths.

[0037] Before training, the model requires the construction of a high-quality dataset containing a large amount of historical time-series data of normal and fault cases, as well as various rare faults and adversarial intervention samples generated through simulation. During training, a stochastic gradient descent algorithm is used, with the aforementioned composite loss function as the optimization objective. Backpropagation is performed on a graphics processing unit cluster to simultaneously optimize the weight parameters of the neural network and the parameters of the graph adjacency matrix representing potential causal relationships. After training is completed, the model parameters are fixed and deployed, enabling forward inference on real-time integrated data to complete the entire task from risk deduction to scenario generation.

[0038] Furthermore, the privacy computing protocol is a zero-knowledge proof protocol; the zero-knowledge proof protocol is implemented based on a trusted execution environment; before executing the collaborative computing operation, the collaborative processing module is also used to receive a verifiable statement generated by an external system based on zero-knowledge proof that has completed a specific risk inspection, and to trigger the collaborative computing operation after the verification is passed.

[0039] Specifically, the aforementioned privacy-preserving computation protocol aims to resolve the core contradiction in cross-departmental data collaboration: the conflict between data use and privacy protection. Privacy-preserving computation refers to a set of technologies that, during data processing and analysis, can complete computational tasks and obtain expected results while maintaining data opacity and confidentiality. To achieve the aforementioned secure collaboration, this embodiment of the invention preferably employs a zero-knowledge proof protocol. This protocol is a classic cryptographic scheme that allows a prover (e.g., an external collaborating department) to prove to a verifier (this system) that a statement is true without revealing any additional information beyond the truthfulness of the statement. This aligns with the need in emergency scenarios to verify whether external units have fulfilled their responsibilities without requiring them to disclose sensitive inspection details.

[0040] For example, suppose the collaborating party is a forestry department with an internal forest fire risk patrol system that records detailed information such as patrol personnel routes, patrol photos, and hazard markings. When the risk simulation module of this system generates a virtual emergency scenario of "a forest fire risk caused by tree obstructions along a mountain corridor," the collaborative response module will initiate a collaboration request to the forestry system. The forestry system, acting as the certifier, uses its locally running zero-knowledge proof algorithm to process its private patrol database, generating a cryptographically verifiable statement. This statement is essentially a short, digital credential tied to a specific patrol task, summarizing as "This department has completed a risk patrol of the target area within the specified time, and the results meet the requirements," while the original data such as patrol time, personnel, and photos are completely hidden. Upon receiving this statement, the collaborative response module of this system uses a pre-agreed public verification algorithm for rapid verification. Once verification is successful, it is certain that the other party has fulfilled its responsibilities, triggering subsequent, more in-depth data collaborative computation.

[0041] Furthermore, to enable complex data fusion even after efficient verification, this embodiment of the invention combines zero-knowledge proof protocols with Trusted Execution Environment (TEE) technology. A TEE is a hardware-level secure area embedded in the main processor, ensuring that code and data loaded within it are strictly protected in terms of confidentiality and integrity, preventing them from being spied on or tampered with even if the operating system is compromised. Its role is to provide a highly trusted execution path for the generation and verification of zero-knowledge proofs, as well as subsequent joint computation. In specific implementations, key proof generation and verification logic, as well as encrypted risk data provided by external systems (such as encrypted fire risk index grids), are decrypted and computed within the secure enclave of the TEE, extending its ability to handle complex collaborative computation through a hardware root of trust.

[0042] Furthermore, the collaborative computing operation process includes: interacting with an external system under the privacy computing protocol framework to obtain first information calculated based on the internal data of the external system; the first information includes at least one of the following: local risk assessment results for the virtual emergency scenario, status and scheduling feasibility information of external controllable resources, and key constraints of the external environment; based on the type of the virtual emergency scenario, matching one or more basic contingency plan frameworks from the contingency plan library, and using the internal power grid status and the first information as parameters, calling an optimization computing model to fill, adjust, and evaluate the basic contingency plan frameworks to generate the final joint response plan.

[0043] In this embodiment of the invention, collaborative computing is a refined process that integrates external intelligence into internal decision-making while protecting privacy. This process begins with secure interaction with external systems within a privacy-preserving computing protocol framework. The first information acquired is not raw data, but rather structured decision support information that has been anonymized or encrypted and provided by the external system within its data sovereignty boundaries, based on a virtual emergency scenario request issued by this system. For example, the local risk assessment result could be that when a virtual scenario involves strong winds causing objects to suspend from power lines, the traffic management department might not provide the original surveillance video, but rather a quantitative indicator output by its video intelligent analysis system regarding the stability risk level and expected probability of detachment of a specific billboard or plastic greenhouse near the target power line.

[0044] Information on the status and scheduling feasibility of externally controllable resources can provide municipal drainage departments with real-time location, discharge capacity, and commitment to arrival time of their available pumping and drainage equipment when dealing with flooding scenarios such as power distribution room flooding.

[0045] Key constraints of the external environment include the ability of the traffic police command center to provide control areas, traffic priority plans, and estimated passage time windows when formulating emergency repair routes at fire scenes.

[0046] The above information is transmitted to the collaborative processing module of this system through a secure channel.

[0047] Subsequently, the system initiates the contingency plan generation logic. Based on the type of the virtual emergency scenario (e.g., wildfire tripping, construction work severing cables), it matches one or more basic contingency plan frameworks from the contingency plan library. This framework is a template containing standard procedures, role assignments, and typical measures, but with gaps left for key parameters. Next, the system calls an optimization calculation model to instantiate the framework. In this embodiment, the model is preferably a mixed-integer programming model. In its specific implementation, the model is constructed based on the classic mathematical framework of power emergency resource scheduling and operation sequence optimization. Its core lies in defining a linear or nonlinear objective function that minimizes weighted load loss and recovery time. Decision variables are designed as a series of binary integer variables to represent discrete decisions such as "whether a switch is operated," "whether a repair team is dispatched to a specific fault point," and "whether an external resource is called upon," while continuous variables are used to represent continuous quantities such as power flow and time. The model's constraint system systematically integrates multiple types of rules, including but not limited to physical constraints characterizing the safe operation of the power grid (such as power balance and line transmission capacity), time constraints characterizing operational logic and sequence (such as the sequence of switching operations and the phased nature of fault isolation and power restoration), and most importantly, quantifying the initial information obtained through external collaboration into specific resource availability constraints, time window constraints, or geographical accessibility constraints. For example, road travel time information obtained from the transportation department will be transformed into a lower bound constraint on the movement time of the emergency repair team. The model itself is a deterministic mathematical model, solved using commercial or open-source mathematical programming solvers. Its required parameters, such as equipment operation time, team movement speed, and resource efficiency, are derived from data mining and statistical summarization of historical operation tickets and simulation case libraries. Historical operation tickets provide the time cost and logical sequence of real-world handling steps, while the simulation case library generated by the digital twin system greatly enriches parameter estimation under various extreme and rare scenarios, jointly ensuring the completeness of the model parameter set.

[0048] The system inputs blank parameters of the basic framework (such as which team to send, when to arrive, and which switch to disconnect) as variables to be optimized, along with the power grid status and initial information, into the model. The model then performs calculations and outputs one or more sets of fully-equipped draft contingency plans and their key performance indicators (such as estimated recovery time, required cost, and probability of success). Based on these evaluation results, the system automatically adjusts the logical branches of the basic framework (e.g., if the main road is blocked, it automatically replaces the backup route plan), forming the final joint response plan.

[0049] This joint response plan is a structured and executable set of operational instructions. Its specific content includes, but is not limited to, a step-by-step list of operational instructions, collaborative task assignment (clearly listing the specific matters requiring external cooperation and time requirements), resource allocation plans, key time nodes, and the expected recovery status. Through a decision-making interface, it is clearly presented to command personnel in a combination of a visual timeline, overlaid geographic information maps, and structured text instruction boxes, allowing them to confirm and issue the plan with a single click or make minor adjustments, thereby transforming cross-departmental intelligent analysis into a unified action plan.

[0050] Furthermore, the collaborative processing module is also used to send a transaction containing participant identifiers, data element information, and contribution evaluation parameters to a smart contract deployed on the blockchain network for quantifying data contributions after the collaborative computing operation is completed. The smart contract has predefined contribution evaluation rules, and automatically calculates and issues tokens representing data contributions based on the received transaction, and records the issuance in the blockchain ledger.

[0051] Specifically, after the collaborative computing operation is completed, the collaborative processing module automatically assembles and sends a structured transaction to the blockchain network. The core data packet of this transaction includes unique digital identifiers for the participants, data metadata describing the characteristics and quality of the provided data (such as type, spatiotemporal range, and refresh frequency), and contribution evaluation parameters automatically generated by the system based on the data's criticality in this decision-making process. The smart contract deployed on the chain has a built-in contribution measurement formula. Upon receiving the transaction, it automatically parses these parameters and calculates the corresponding contribution metric value according to predefined public rules. Based on this, a specific number of digital tokens are issued as contribution credentials, and the complete record of this issuance is permanently and immutably written into the distributed ledger, thereby achieving transparent measurement and automated incentives for cross-departmental collaborative contributions.

[0052] By leveraging the immutability of blockchain and the automatic execution of smart contracts, the abstract value of data collaboration is transformed into quantifiable and auditable on-chain digital assets. This technically establishes a trustworthy incentive link across departments, not only solving the problems of insufficient trust and motivation in data sharing, but also promoting a fundamental shift in emergency collaboration from temporary administrative coordination to a sustainable and self-driven data ecosystem, ensuring the vitality of data supply and the quality of collaboration for the long-term operation of the system.

[0053] The second aspect of this invention discloses a remote decision-making interaction method for power operation and maintenance emergency services. This method is applied to the system disclosed in the first aspect. The method includes: accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; outputting potential fault risk transmission chains based on the first integrated data through a causal inference model embedded with the physical rules of the power system, and generating corresponding virtual emergency scenarios; based on the virtual emergency scenarios, performing collaborative computing operations with at least one external system through a privacy computing protocol to generate a joint response plan; wherein the external system is determined based on the external data required by the virtual emergency scenarios; and visually displaying the fault risk transmission chains, the virtual emergency scenarios, and the joint response plan.

[0054] Example 2: The second aspect of this invention discloses a remote decision-making interaction method for power operation and maintenance emergency services. The method includes: accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; outputting potential fault risk transmission chains based on the first integrated data using a causal inference model embedded with power system physical rules, and generating corresponding virtual emergency scenarios; based on the virtual emergency scenarios, performing collaborative computing operations with at least one external system through a privacy computing protocol to generate a joint response plan; wherein the external system is determined based on the external data required by the virtual emergency scenarios; and visually displaying the fault risk transmission chains, the virtual emergency scenarios, and the joint response plan.

[0055] It should be noted that the specific implementation process of Embodiment 2 is similar to that of Embodiment 1, and will not be repeated in this embodiment.

[0056] Finally, it should be noted that the above-described embodiments include multiple parallel implementations of the present invention. Deleting or otherwise adjusting one or more implementations will not affect the implementation of the solution. Furthermore, the remote decision-making and interactive system and method for power operation and maintenance emergency services disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote decision-making and interactive system for power operation and maintenance emergency services, characterized in that, The system includes: a data fusion module for accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; a risk simulation module for outputting potential fault risk transmission chains based on the first integrated data using a causal inference model embedded with the physical rules of the power system, and generating corresponding virtual emergency scenarios; a collaborative response module for generating joint response plans based on the virtual emergency scenarios and performing collaborative computing operations with at least one external system through a privacy computing protocol; wherein the external system is determined based on the external data required by the virtual emergency scenarios; and a decision interaction module for providing a decision interaction interface, configuring the decision interaction interface to visually display the fault risk transmission chains, virtual emergency scenarios, and joint response plans, and providing a human-computer interaction command entry point.

2. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 1, characterized in that, The causal inference model is a physical information neural network; the loss function of the physical information neural network includes a data fitting term and a physical rule constraint term; wherein, the physical rule constraint term is constructed based on the differential equation of power flow calculation in the power system; during the training process, the physical rule constraint term is used to ensure that the output result of the neural network satisfies the power balance relationship of the power flow calculation at any node in the power grid topology.

3. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 2, characterized in that, The physical information neural network uses a graph structure to represent the power grid topology, where the nodes of the graph correspond to power equipment and the edges represent electrical connections.

4. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 3, characterized in that, The loss function further includes a graph structure learning loss term, used to directly optimize the adjacency matrix of the graph structure; the graph structure learning loss term includes: a reconstruction loss for reconstructing observation data based on the graph structure and node states; the observation data includes real-time and historical measurement data accessed from the data fusion module; a norm regularization term for the adjacency matrix of the graph structure; and a physical consistency constraint term to ensure that the evolution of node states operating on the graph structure conforms to the power flow calculation differential equation.

5. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 4, characterized in that, The loss function also includes an intervention prediction loss term; the training data of the causal inference model includes power grid observation data used to train the model's basic representation and reconstruction capabilities, and adversarial samples generated by simulating various fault intervention measures based on the observation data and power grid topology, which are used to specifically optimize the model's intervention prediction capabilities; the intervention prediction loss term is constructed by calculating the difference between the causal inference model's predicted results of the system state after intervention and the actual state.

6. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 5, characterized in that, The loss function also includes a causal interpretability constraint term; the causal interpretability constraint term is used to maximize the mutual information between the feature vector of the hidden layer of the physical information neural network and a set of predefined causal mediator variables of the power system; the causal mediator variables of the power system represent intermediate state variables that can characterize the fault risk transmission process and have clear physical meaning in the fault risk transmission chain.

7. The remote decision-making and interactive system for power operation and maintenance emergency services according to any one of claims 1-6, characterized in that, The privacy computing protocol is a zero-knowledge proof protocol; the zero-knowledge proof protocol is implemented based on a trusted execution environment; before executing the collaborative computing operation, the collaborative processing module is also used to receive a verifiable statement generated by an external system based on zero-knowledge proof that has completed a specific risk inspection, and to trigger the collaborative computing operation after the verification is passed.

8. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 7, characterized in that, The collaborative computing operation includes: interacting with an external system under the privacy computing protocol framework to obtain first information calculated based on the internal data of the external system; the first information includes at least one of the following: local risk assessment results for the virtual emergency scenario, status and scheduling feasibility information of external controllable resources, and key constraints of the external environment; based on the type of the virtual emergency scenario, matching one or more basic contingency plan frameworks from the contingency plan library, and using the internal power grid status and the first information as parameters, calling an optimization computing model to fill, adjust, and evaluate the basic contingency plan frameworks to generate the final joint response plan.

9. The remote decision-making and interactive system for power operation and maintenance emergency services according to claim 8, characterized in that, The collaborative processing module is also used to send a transaction containing participant identifiers, data element information, and contribution evaluation parameters to a smart contract deployed on the blockchain network for quantifying data contributions after the collaborative computing operation is completed. The smart contract has predefined contribution evaluation rules, and automatically calculates and issues tokens representing data contributions based on the received transaction, and records the issuance in the blockchain ledger.

10. A remote decision-making and interaction method for power operation and maintenance emergency services, wherein the method is applied to the system described in any one of claims 1-9, characterized in that, The method includes: accessing and integrating multi-source heterogeneous data from power grid monitoring, meteorological environment, and external collaborative departments to generate first integrated data; outputting potential fault risk transmission chains based on the first integrated data through a causal inference model embedding power system physical rules, and generating corresponding virtual emergency scenarios; based on the virtual emergency scenarios, performing collaborative computing operations with at least one external system through a privacy computing protocol to generate a joint response plan; wherein, the external system is determined based on the external data required by the virtual emergency scenarios; and visually displaying the fault risk transmission chains, virtual emergency scenarios, and joint response plans.

Images