Intelligent management method and system for power grid resources based on 6G and deep learning

Abstract

The present application relates to the technical field of power grid resource management, and discloses a method and system for intelligent management of power grid resources based on 6G and deep learning. The method includes: collecting node data through 6G technology to obtain a topology mapping table, pre-processing data to obtain feature tensors, using a hybrid architecture of spatiotemporal graph convolution and long short-term memory network to train a model cluster, analyzing the network status based on the model, generating a scheduling plan, and issuing control instructions through 6G network slicing to achieve intelligent management of the power network. The present application improves the real-time data acquisition accuracy and intelligent analysis capabilities of the power network by combining 6G communication positioning integrated technology with a hybrid architecture of spatiotemporal graph convolution network and long short-term memory network. Based on the collaborative optimization mechanism of multi-source data fusion and deep learning model cluster, the comprehensiveness of power grid status assessment and the accuracy of resource scheduling are enhanced.

Description

Technical Field

[0001] The present application relates to the technical field of power grid resource management, and in particular to a method and system for intelligent management of power grid resources based on 6G and deep learning. Background Art

[0002] Currently, power network resource management primarily relies on traditional SCADA systems and power dispatch automation systems for monitoring and control. These systems collect power equipment operating data via fiber optic cables or low-speed wireless networks and utilize classic power flow calculations and economic dispatch algorithms for resource allocation. For communications, 4G / 5G mobile communication technologies, power line carrier communications, and narrowband Internet of Things (Narrowband Internet of Things) are primarily used, providing limited data transmission capabilities. Positioning technology primarily relies on GPS or base station positioning technology, with accuracy generally within the meter range. At the decision-making level, traditional systems primarily utilize rule-based expert systems or simple statistical prediction models, relying on manual experience to construct decision rules and lacking the support of advanced intelligent algorithms such as deep learning. These systems can meet basic requirements in relatively stable power grid environments, ensuring the safe and stable operation of the power system.

[0003] However, with the increasing proportion of renewable energy, the widespread integration of distributed energy resources, and increased volatility in electricity load, traditional power network resource management methods face significant challenges. First, traditional communication technologies suffer from limited bandwidth, high latency, and insufficient connection capacity, making it difficult to support real-time data transmission from massive amounts of power equipment. Second, traditional positioning technologies lack precision, failing to meet the demands of refined resource management. Third, traditional data processing methods lack the ability to effectively utilize massive amounts of heterogeneous data, and their value cannot be fully mined. Furthermore, traditional decision-making models have limited ability to predict grid states in complex scenarios, making it difficult to cope with the uncertainties associated with high proportions of renewable energy integration. Finally, traditional resource scheduling algorithms struggle to achieve multi-objective optimization and strike a balance between economic efficiency, reliability, and environmental performance. These shortcomings severely restrict improvements in grid resource utilization efficiency and increase operational risks. Summary of the Invention

[0004] This application provides a method and system for intelligent management of power grid resources based on 6G and deep learning. By combining 6G communication and positioning technology with a hybrid architecture of spatiotemporal graph convolutional networks and long-short-term memory networks, this system improves the accuracy of real-time data acquisition and intelligent analysis capabilities of the power grid. A collaborative optimization mechanism based on multi-source data fusion and deep learning model clustering enhances the comprehensiveness of grid status assessment and the accuracy of resource scheduling.

[0005] In the first aspect, the present application provides a method for intelligent management of power grid resources based on 6G and deep learning, which includes: collecting power network node data through 6G communication base stations and multi-source positioning sensors to obtain a power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information; performing data preprocessing on the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information to obtain a power data feature tensor; according to the power data feature tensor, training a deep learning model cluster through a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load Prediction model, status assessment model and resource optimization model; collect real-time resource data of the power grid, and perform grid status analysis on the real-time resource data of the power grid according to the load prediction model, status assessment model and resource optimization model to obtain a grid status assessment report, a fault risk distribution map and early warning instructions; calculate the resource scheduling plan according to the grid status assessment report, fault risk distribution map and early warning instructions to obtain the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan; according to the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan, control instructions are issued through 6G network slicing to obtain power network operation status data and performance evaluation report.

[0006] In a second aspect, the present application provides a power grid resource intelligent management system based on 6G and deep learning, the power grid resource intelligent management system based on 6G and deep learning comprising:

[0007] The acquisition module is used to collect power network node data through 6G communication base stations and multi-source positioning sensors to obtain the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information;

[0008] a processing module, configured to perform data preprocessing on the power network topology physical coordinate mapping table, the real-time power parameter data stream, and the communication network status information to obtain a power data feature tensor;

[0009] A training module is used to train a deep learning model cluster based on the power data feature tensor through a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load forecasting model, a state assessment model, and a resource optimization model;

[0010] An analysis module is used to collect real-time resource data of the power grid and perform grid status analysis on the real-time resource data of the power grid according to the load forecasting model, the status assessment model and the resource optimization model to obtain a grid status assessment report, a fault risk distribution map and an early warning instruction;

[0011] A calculation module is used to calculate a resource scheduling plan based on the grid status assessment report, the fault risk distribution map, and the early warning instructions, and obtain a power generation equipment output plan, an energy storage system charging and discharging strategy, and a load switching plan;

[0012] The control module is used to issue control instructions through 6G network slicing based on the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan to obtain power network operation status data and performance evaluation reports.

[0013] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned 6G and deep learning-based intelligent management method for power grid resources.

[0014] The fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the above-mentioned method for intelligent management of power grid resources based on 6G and deep learning.

[0015] In the technical solution provided by this application, the power network node data is collected through 6G communication base stations and multi-source positioning sensors to obtain the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information, which solves the problems of insufficient data collection accuracy and poor real-time performance in traditional technologies, and provides a high-quality data foundation for subsequent intelligent analysis. The technical feature of pre-processing the collected data to form a power data feature tensor effectively removes noise and redundancy in the original data, improves data quality, and enables deep learning models to more accurately extract key features of the power network. The use of a hybrid architecture of spatiotemporal graph convolutional neural networks and long short-term memory networks fully considers the spatial topological structure and time series characteristics of the power network. Compared with traditional single-structure neural networks, this hybrid architecture can simultaneously capture the spatial dependencies between power nodes and the temporal pattern of load changes. The trained load forecasting models, state assessment models and resource optimization models are more accurate and have stronger generalization capabilities. The technical feature of collecting real-time power grid resource data and using the aforementioned model to analyze the grid status enables early identification and risk assessment of abnormal grid conditions. Compared to traditional threshold-based monitoring methods, this deep learning-based state analysis can discover complex nonlinear relationships and potential risk patterns, greatly improving the accuracy of grid security warnings. The technical feature of calculating resource scheduling plans based on grid status assessment results enables the coordinated optimization of power generation, energy storage, and load. Compared with traditional single-resource optimization, this multi-resource coordinated scheduling significantly improves system flexibility and resource utilization efficiency. The technical feature of issuing control instructions through 6G network sharding ensures ultra-reliable and low-latency transmission of critical control instructions, solving the congestion and delay problems that may occur in traditional communication networks in emergency situations, and enabling control instructions to be executed accurately and promptly. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a schematic diagram of an embodiment of a method for intelligent management of power grid resources based on 6G and deep learning in an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of an embodiment of an intelligent management system for power grid resources based on 6G and deep learning in an embodiment of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The embodiments of the present application provide a method and system for intelligent management of power grid resources based on 6G and deep learning. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of the method for intelligent management of power grid resources based on 6G and deep learning includes:

[0022] Step S101: Collect power network node data through 6G communication base stations and multi-source positioning sensors to obtain a power network topology physical coordinate mapping table, real-time power parameter data stream, and communication network status information;

[0023] Step S102: pre-processing the power network topology physical coordinate mapping table, the real-time power parameter data stream, and the communication network status information to obtain a power data feature tensor;

[0024] Step S103: Based on the power data feature tensor, a deep learning model cluster is trained using a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load forecasting model, a state assessment model, and a resource optimization model;

[0025] Step S104: collecting real-time resource data of the power grid, and performing grid status analysis on the real-time resource data of the power grid according to the load forecasting model, the status assessment model, and the resource optimization model, to obtain a grid status assessment report, a fault risk distribution map, and an early warning instruction;

[0026] Step S105: Calculate a resource scheduling plan based on the grid status assessment report, fault risk distribution map, and early warning instructions to obtain a power generation equipment output plan, energy storage system charging and discharging strategy, and load switching plan;

[0027] Step S106: Based on the power generation equipment output plan, the energy storage system charging and discharging strategy, and the load switching plan, control instructions are issued through the 6G network slicing to obtain the power network operation status data and performance evaluation report.

[0028] It is understandable that the execution subject of this application can be a power grid resource intelligent management system based on 6G and deep learning, or a terminal or server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0029] Specifically, 6G communication base stations and multi-source positioning sensors are deployed at each node in the power network to collect node data. Specifically, 6G communication modules installed at substations, key points on transmission lines, and distribution equipment measure signal strength and transmission delay between nodes to generate a node communication connection matrix. Intelligent power sensors collect voltage, current, active power, reactive power, and frequency values every second, forming a real-time power parameter data stream. GPS receivers and base station-assisted positioning systems collect device locations and, after Kalman filtering, obtain centimeter-level spatial coordinates. The node communication matrix is matched point-to-point with the device spatial coordinates to construct a mapping table of the physical coordinates of the power network topology. The signal-to-noise ratio, bit error rate, and available bandwidth of the 6G communication link are measured to generate communication network status information. The collected data is then preprocessed. The real-time power parameter data stream is decomposed into high-frequency noise components and low-frequency trend components through wavelet transform to form a power signal component matrix; the autoencoder detects and repairs outliers and missing values in the matrix to generate a cleaned power parameter matrix; a node adjacency graph is constructed based on the topological physical coordinate mapping table, and the connection weight values between each node are calculated; the signal-to-noise ratio, bit error rate and bandwidth data in the communication network status information are normalized to form a communication status feature vector; the cleaned power parameter matrix is associated and fused with the node adjacency graph and the communication status feature vector to construct a multi-dimensional association matrix; the matrix is tensor-decomposed using the Tucker decomposition algorithm, setting the time dimension core tensor rank to 24, the space dimension core tensor rank to 30% of the number of network nodes, and the physical characteristic dimension core tensor rank to 8 to obtain the power data feature tensor.

[0030] A deep learning model cluster is trained based on the power data feature tensor. The feature tensor is divided into training and validation sets according to the time dimension. A power network node relationship graph is established to construct a model training dataset. A spatiotemporal graph convolutional network processes this dataset, extracting spatial dependency features between power nodes and generating a spatial correlation representation of power nodes. A long-short-term memory network receives this representation, capturing the temporal variation of power load and obtaining temporal correlation features. A load forecasting model is constructed based on the temporal correlation features and historical load data. A Bayesian neural network is trained based on the spatial correlation representation of power nodes and device state parameters to quantify grid state uncertainty and form a state assessment model. The prediction results of the load forecasting model and the evaluation results of the state assessment model are used to train a deep reinforcement learning network to optimize resource allocation strategies and obtain a resource optimization model. Real-time power grid resource data is collected and grid state analysis is performed. The distributed sensor network collects the real-time operating parameters of each node, including load level, equipment temperature, transmission line impedance and phase angle difference; the load forecasting model calculates the load distribution forecast value for the next 24 hours and generates a load forecast result map; the state assessment model performs Bayesian probabilistic reasoning on real-time resource data and load forecast results, calculates the stability margin index and fault probability value of each node, and constructs a power grid stability assessment matrix; threshold analysis is performed on this matrix to identify high-risk nodes and generate a risk level classification table; the risk level classification table is mapped to the power network topology map to form a fault risk distribution map; based on the fault risk distribution map and risk propagation assessment results, a power grid state assessment report and early warning instructions are generated.

[0031] Calculate resource scheduling plans based on grid status assessment results. Extract key system stability indicators and load forecast data from the grid status assessment report to construct a set of resource scheduling constraints. Based on the high-risk areas of the fault risk distribution map, calculate the risk factor and emergency power demand for each area, and form a power supply priority table for risk areas. Based on the fault type information and time estimate in the early warning instructions, formulate a grid scheduling schedule and generate a scheduling cycle matrix. Calculate the optimal output distribution ratio for each type of power generation equipment and generate a power generation equipment output plan. Calculate power shortages and surpluses, allocate charging and discharging time periods and power values to the energy storage system, and form a charging and discharging strategy for the energy storage system. Determine the load nodes that can be cut and the switching sequence, and generate a load switching plan.

[0032] Control instructions are issued and feedback data is obtained through 6G network slicing. The scheduling plan is converted into a standardized control instruction format to generate instruction data packets; the instruction data packets are prioritized and divided into three levels: critical control instructions, routine scheduling instructions, and auxiliary management instructions; communication resources are dynamically allocated through 6G network slicing technology to create different types of network slices; a distributed consistency algorithm is used to synchronize control instructions at nodes to ensure the time synchronization and logical consistency of multi-node control; the synchronized control instruction set is issued to the control execution unit through each network slice, and execution feedback data is collected to form power network operation status data; a multi-dimensional comparison and analysis of the real-time operating parameters of the power network is performed, and parameter deviation values and fluctuation curves are recorded, and the execution status of control instructions is correlated to form a performance evaluation report.

[0033] For example, when abnormal voltage fluctuations were detected at a power network node, the state assessment model calculated a 78% probability of failure at that node through spatiotemporal correlation analysis of the power data feature tensor, marking it as a high-risk area. Based on this, the resource scheduling solution configured the adjacent energy storage system to discharge mode, setting the power to 2MW and preparing 5MW of shelving load as an emergency measure. These control commands were issued via 6G ultra-reliable, low-latency network slices, ensuring command transmission latency of less than 5 milliseconds, effectively preventing the spread of grid faults.

[0034] In an embodiment of the present application, 6G communication base stations and multi-source positioning sensors are used to collect power network node data, obtaining a power network topology physical coordinate mapping table, real-time power parameter data streams, and communication network status information. This solves the problems of insufficient data collection accuracy and poor real-time performance in traditional technologies and provides a high-quality data foundation for subsequent intelligent analysis. The technical feature of preprocessing the collected data to form a power data feature tensor effectively removes noise and redundancy from the raw data, improving data quality and enabling deep learning models to more accurately extract key features of the power network. The hybrid architecture of spatiotemporal graph convolutional neural networks and long-short-term memory networks fully considers the spatial topology and time series characteristics of the power network. Compared with traditional single-structure neural networks, this hybrid architecture can simultaneously capture the spatial dependencies between power nodes and the temporal patterns of load changes. The trained load forecasting models, state assessment models, and resource optimization models have higher accuracy and stronger generalization capabilities. The technical feature of collecting real-time power grid resource data and using the above-mentioned models for grid state analysis enables early identification and risk assessment of abnormal grid conditions. Compared with traditional threshold-based monitoring methods, this deep learning-based state analysis can discover complex nonlinear relationships and potential risk patterns, greatly improving the accuracy of grid security warnings. The technical feature of calculating resource scheduling plans based on grid state assessment results enables the coordinated optimization of power generation, energy storage, and load. Compared with traditional single-resource optimization, this multi-resource coordinated scheduling significantly improves system flexibility and resource utilization efficiency. The technical feature of issuing control instructions through 6G network sharding ensures ultra-reliable and low-latency transmission of critical control instructions, solving the congestion and delay issues that may occur in traditional communication networks in emergency situations, and enabling the precise and timely execution of control instructions.

[0035] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0036] (1) Deploy 6G communication modules at each substation, key points of transmission lines, and distribution equipment in the power network to measure the signal strength and transmission delay between nodes and generate a node communication connection matrix;

[0037] (2) The intelligent power sensors collect the voltage, current, active power, reactive power and frequency values of each node in the power network every second to form a real-time power parameter data stream;

[0038] (3) Use GPS receivers and base station assisted positioning systems to collect the precise location of the device, use the Kalman filter algorithm to process the raw positioning data, and obtain centimeter-level device spatial coordinates;

[0039] (4) Perform point-to-point matching of the link status data in the node communication connection matrix with the device spatial coordinates to construct a power network topology physical coordinate mapping table that includes physical distance and link quality;

[0040] (5) By measuring the signal-to-noise ratio, bit error rate, and available bandwidth of the 6G communication link, recording the real-time status parameters of each communication channel, and generating communication network status information;

[0041] (6) Timestamp and align the data points in the real-time power parameter data stream to eliminate the sampling time differences between different devices and complete the data synchronization processing of the power network nodes.

[0042] Specifically, 6G communication modules are deployed at key locations in the power grid. These modules are installed at substations, key points on transmission lines, and distribution equipment, forming a grid-like coverage structure. Each communication module is equipped with millimeter-wave and terahertz wave transceivers, enabling ultra-high-speed data transmission. The communication modules measure each other's signal strength by sending a probe signal and recording the received power. The round-trip signal time is calculated to obtain the transmission delay. The received power is converted into a link quality indicator using a signal strength attenuation model, generating a connection status score for each pair of nodes. These scores are organized into a matrix, where the matrix elements represent the connection quality between nodes, thus forming a node communication connection matrix.

[0043] At the same time, smart power sensors collect data at high frequencies at various power nodes. These sensors, including smart meters, current transformers, and voltage transformers, are equipped with high-precision analog-to-digital converters and acquire power parameters at a sampling rate of one second. The collected data includes line voltage (in volts), current (in amperes), active power (in watts), reactive power (in vars), and frequency (in hertz). This data is timestamped and packaged into a data stream format. It is transmitted in real time to the data processing center via a 6G communication module, forming a real-time power parameter data stream.

[0044] Multi-source positioning technology is used to accurately locate equipment. Each power device is equipped with a GPS receiver to receive satellite positioning signals to obtain initial location data. Simultaneously, positioning signals transmitted by nearby 6G base stations are used to calculate the distance between the device and the base station using a time difference of arrival algorithm. Because the raw positioning data contains noise and errors, a Kalman filter algorithm is applied to process it. This algorithm consists of two phases: prediction and correction. The prediction phase predicts the current state based on the previous state and the system model; the correction phase corrects the prediction based on observed data. The Kalman filter recursively estimates the device's true position, filtering out random fluctuations and ultimately achieving centimeter-level positioning accuracy.

[0045] After obtaining the node communication connection matrix and device spatial coordinates, point-to-point matching is required. First, a unique identifier is assigned to each communication node, and a corresponding index is created for each device location coordinate. The two datasets are linked by identifiers, integrating communication quality data with physical distance information. The Euclidean distance between any two points is calculated to represent the physical separation between the nodes. Link communication quality metrics, including bandwidth capacity and stability scores, are also retained. This data is organized into a physical coordinate mapping table for the power network topology, which contains node identifiers, geographic coordinates, a list of connected nodes, and corresponding link quality metrics.

[0046] To obtain information on the communication network status, 6G communication links must be monitored in real time. A dedicated network analyzer continuously measures key parameters for each link: the signal-to-noise ratio (in decibels) reflects interference immunity; the bit error rate (BER) indicates the percentage of errors per bit; and the available bandwidth (in megabits per second) indicates data transmission capacity. These measurements are statistically processed to calculate the average, peak, and standard deviation, generating a description of the status of each communication channel. These descriptions are organized into structured data, forming a complete picture of the communication network status. Time synchronization is performed on the collected real-time power parameter data streams. Due to slight differences in sampling clocks between devices, sampling times may not be completely consistent. First, a precise timestamp is added to each data point to mark the sampling moment. A unified time base is then established, typically Coordinated Universal Time (UTC). Linear interpolation is used to align data points at different times to a common timeline, eliminating sampling time discrepancies. The aligned data is then reorganized into a unified data stream format, completing data synchronization across power network nodes.

[0047] For example, a 6G communication module deployed at a substation measured a signal strength of -70 dBm with a transmission delay of 2 milliseconds. Consequently, the node communication connection matrix recorded a connection quality of 0.85 (out of a maximum score of 1). Simultaneously, smart power sensors collected data from the substation's main transformer, indicating a voltage of 220 kV, a current of 500 A, and an active power of 105 MW. Raw location data captured by the GPS receiver showed a latitude deviation of up to 1.2 meters. After Kalman filtering, positioning accuracy was improved to 8 centimeters. This information was integrated into a physical coordinate mapping table within the power network topology, recording the physical distance between the substation and adjacent nodes and the communication quality. Communication link measurements showed a signal-to-noise ratio of 21 dB, a bit error rate of 10^-6, and an available bandwidth of 1.2 Gbps. When a 15-millisecond discrepancy in sampling time was detected between different devices, timestamp alignment was performed to align the data to a reference time point, ensuring data synchronization for subsequent analysis.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] (1) Perform wavelet transform decomposition on the real-time power parameter data stream to separate the high-frequency noise components and low-frequency trend components in the power signal and form a power signal component matrix;

[0050] (2) Use the autoencoder to detect and repair abnormal values and missing values in the power signal component matrix to generate a cleaned power parameter matrix;

[0051] (3) Construct a node adjacency graph based on the physical coordinate mapping table of the power network topology and calculate the connection weight value between each node;

[0052] (4) Normalize the signal-to-noise ratio, bit error rate, and bandwidth data in the communication network status information to form a communication status feature vector;

[0053] (5) The cleaned power parameter matrix is associated and fused with the node adjacency graph and the communication status feature vector to construct a multi-dimensional association matrix;

[0054] (6) The multi-dimensional correlation matrix is decomposed into tensors using the Tucker decomposition algorithm. The time dimension core tensor rank is set to 24, the space dimension core tensor rank is set to 30% of the number of network nodes, and the physical characteristic dimension core tensor rank is set to 8. Multi-linear feature extraction is performed to obtain the power data feature tensor.

[0055] Specifically, the real-time power parameter data stream is decomposed by wavelet transform using discrete wavelet transform (DWT) technology. Wavelet transform has multi-resolution analysis capabilities and decomposes the signal through low-pass filters and high-pass filters. The specific operation is to perform multi-level decomposition of power time series signals such as voltage, current, and power using the Daubechies wavelet basis (db4), usually a 5-level decomposition. Each level of decomposition divides the signal into approximate coefficients (low-frequency part) and detail coefficients (high-frequency part). The low-frequency part reflects the overall trend of the signal and contains the basic change pattern of the power load; the high-frequency part contains short-term fluctuations, measurement noise, and interference information. The decomposed coefficients at each level are organized into a matrix form, with each row representing a power parameter and each column representing a different frequency component, forming a power signal component matrix.

[0056] An autoencoder is used to detect outliers and repair missing values in the power signal component matrix. The autoencoder is a neural network structure consisting of an encoder and a decoder. The encoder compresses the input data into a low-dimensional latent space, while the decoder attempts to reconstruct the original data from the latent space. Normal data points can be reconstructed well, while outliers usually have large reconstruction errors. The specific operation is to input the power signal component matrix into the autoencoder in segments according to time windows and calculate the reconstruction error. If the reconstruction error of a data point exceeds the set threshold (usually set to 3 times the standard deviation of the average reconstruction error), it is marked as an outlier. For missing values, a method combining forward filling and autoencoder generated values is used to repair them. For outliers, they are replaced with autoencoder reconstructed values. The processed data is reorganized into a matrix form to obtain a cleaned power parameter matrix. Constructing a node adjacency relationship graph based on the physical coordinate mapping table of the power network topology is a key step. The power network can be regarded as a graph structure, in which substations, key points of transmission lines and distribution equipment are nodes and physical connection relationships are edges. The calculation of connection weight values involves two factors: physical distance and communication quality, and can be expressed by the following formula:

[0057]

[0058] in, Representation node and nodes The connection weight value between them ranges from [0,1]; Represents the physical Euclidean distance between nodes; is the distance attenuation parameter, which controls the influence of distance on weight; Indicates the quality of the communication link, which is derived from a comprehensive evaluation of signal strength and delay; To balance parameters, the relative contributions of physical distance and communication quality to the weight are controlled, with values ranging from [0 to 1]. Larger weights indicate stronger connections between nodes. The calculated weights are stored in the adjacency matrix, completing the construction of the node adjacency graph.

[0059] Normalizing communication network status information is essential to ensure comparability of metrics across different dimensions. Raw communication status information includes parameters such as signal-to-noise ratio (dB), bit error rate (dimensionless), and bandwidth (bps), which can vary widely in value range. The Z-score normalization method is used to normalize each parameter:

[0060]

[0061] in, is the standardized parameter value, is the original parameter value, is the mean value of the parameter, is the standard deviation. After standardization, each parameter has a mean of 0 and a standard deviation of 1, and is on the same order of magnitude. The standardized signal-to-noise ratio, bit error rate (taking a negative value so that smaller values indicate better performance), and bandwidth values are concatenated into a vector to form a communication status feature vector with a unified dimension.

[0062] The cleaned power parameter matrix is associated and fused with the node adjacency graph and the communication state feature vector to construct a multi-dimensional association matrix. First, the topological structure information of the node adjacency graph is combined with the power parameter matrix to construct a graph-enhanced power parameter representation. The specific operation is to use graph convolution to weightedly mix the power parameters of each node with the parameters of its neighboring nodes:

[0063]

[0064] in, For the enhanced power parameter representation, is the adjacency matrix, is the degree matrix (diagonal elements are the number of edges of each node), is the power parameter matrix after cleaning, is the learnable parameter matrix, is a nonlinear activation function. Then, the communication state feature vector is used as an additional feature and connected with the graph-enhanced power parameter representation to form a multi-dimensional correlation matrix containing time, space, and physical characteristics.

[0065] Finally, the multi-dimensional correlation matrix is decomposed into tensors using the Tucker decomposition algorithm. Tucker decomposition is a high-order singular value decomposition that decomposes a high-dimensional tensor into the product of a core tensor and multiple factor matrices:

[0066]

[0067] in, represents the original multidimensional incidence matrix, represents the kernel tensor, 、 、 Represent the factor matrices in the dimensions of time, space and physical characteristics, Represents the tensor-matrix product along the vth dimension. Based on the characteristics of the power grid, the time-dimensional core tensor rank is set to 24 (corresponding to a 24-hour periodicity), the spatial-dimensional core tensor rank is set to 30% of the number of network nodes (preserving key spatial patterns), and the physical-property-dimensional core tensor rank is set to 8 (covering key power parameter characteristics). By selecting appropriate core tensor ranks, data dimensionality reduction and feature extraction are achieved, resulting in a compact and information-rich power data feature tensor.

[0068] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0069] (1) Divide the power data feature tensor into a training set and a validation set according to the time dimension, establish a power network node relationship diagram, and construct a model training dataset;

[0070] (2) Applying a spatiotemporal graph convolutional network to the model training dataset to extract the spatial dependency features between power nodes and generate a representation of the spatial correlation of power nodes;

[0071] (3) Inputting the spatial correlation representation of power nodes into the long short-term memory network to capture the temporal variation of power load and obtain the temporal correlation characteristics;

[0072] (4) Based on the time correlation characteristics and historical load data, a load forecasting model is constructed, the loss function is optimized by the gradient descent algorithm, and the model parameters are adjusted until the prediction error converges;

[0073] (5) Based on the spatial correlation representation of power nodes and equipment state parameters, a Bayesian neural network is trained to quantify the uncertainty of the power grid state and form a state assessment model;

[0074] (6) Using the prediction results of the load forecasting model and the evaluation results of the state assessment model, the deep reinforcement learning network is trained, and the resource allocation strategy is optimized through the reward and punishment mechanism to obtain a resource optimization model.

[0075] Specifically, the power data feature tensor is partitioned. This feature tensor is a three-dimensional data structure obtained after preprocessing the power network data. It contains a time dimension (recording data at different time points), a spatial dimension (recording data at different nodes), and a feature dimension (recording different power parameters). Data partitioning uses a sliding window method, taking data from a continuous period as input and predicting data for subsequent time periods. The input window is typically 7-30 days, and the prediction window is 1-7 days. Specifically, the entire data is arranged in chronological order, and samples from the first 80% of the time period are used as the training set, and the last 20% as the validation set. For example, for 90 days of historical data, the first 72 days of data enter the training set, and the last 18 days enter the validation set. This partitioning method maintains the continuity of the time series and avoids the potential leakage of temporal information that may result from random sampling.

[0076] At the same time, based on the physical coordinate mapping table of the power network topology obtained in the preprocessing phase, the power network is represented as a graph structure, consisting of a set of nodes (such as substations, key points on transmission lines, and distribution equipment) and a set of edges (representing the physical connections between nodes). Each node contains a multidimensional attribute vector that records parameters such as voltage, current, and power at that node; each edge contains a weight value that represents the strength of the connection between nodes. The edge weight value is derived from the connection weight value calculated in the aforementioned data preprocessing phase, taking into account physical distance and communication quality. The training dataset is constructed by combining the node features in the power data feature tensor with the topological structure of the node relationship graph to form a complete dataset with spatiotemporal correlations. Each training sample contains both time series information and graph structure information, providing a foundation for subsequent spatiotemporal joint modeling.

[0077] The model training dataset was applied to a spatiotemporal graph convolutional network (STN), a deep learning architecture specialized for processing graph-structured data with spatiotemporal characteristics. The network architecture comprises multiple spatiotemporal convolutional blocks, each consisting of a spatial graph convolution layer and a temporal convolution layer. The spatial graph convolution layer aggregates the feature information of each node and its neighbors through convolution operations defined on the graph. Specifically, for each node, its first-order neighbors are determined using an adjacency matrix. The features of the node and its neighbors are then weighted and aggregated, with the weights determined by the strength of the connections. This allows physically or electrically close nodes to influence each other, forming a feature representation that accounts for spatial structure. Through multi-layer graph convolution operations, information is transferred over a wider range, capturing multi-order dependencies between nodes. The temporal convolution layer uses a one-dimensional convolutional sliding window to process the feature sequence of each node in the temporal dimension, capturing short-term temporal patterns. Spatial graph convolution and temporal convolution are performed alternately, extracting and fusing spatially and temporally dependent features layer by layer, ultimately generating a spatial correlation representation of power nodes that incorporates the mutual influence between nodes. The spatial correlation representation of power nodes is fed into a long-short-term memory (LSTM) network to further process long-term dependencies in time series. The LSTM network is a special type of recurrent neural network designed to address the vanishing gradient problem of conventional recurrent networks when processing long sequences. The core of the network is the memory unit, which consists of three control units: the input gate, the forget gate, and the output gate. The input gate controls the extent to which current input information enters the memory unit; the forget gate determines how much historical memory is retained; and the output gate controls the degree to which the memory unit state influences the current output. The gating mechanism is implemented using a sigmoid function, with output values between 0 and 1 representing the proportion of information passing through. Power load data often exhibits complex temporal characteristics, such as short-term fluctuations, intraday cycles, intraweek patterns, and seasonal variations. The LSTM network, through its memory units and gating mechanism, can learn and memorize patterns across different time scales. Specifically, the spatial correlation representation of power nodes is fed into the network in chronological order. The network gradually processes the sequence information, retaining important historical information through the memory mechanism and adjusting the importance of information at different time points through the gating mechanism. Ultimately, it outputs a feature representation rich in temporal information, namely the temporal correlation feature.

[0078] A load forecasting model is constructed based on time-correlation features and historical load data. The model employs a multi-layer feedforward neural network architecture. The input layer receives time-correlation features and processed historical load data (e.g., statistical features across different time windows). The network consists of multiple hidden layers, each composed of a fully connected layer and a nonlinear activation function. Common activation functions include rectified linear units (ReLUs), which effectively mitigate the vanishing gradient problem. The output layer typically does not use an activation function or uses the identity function, directly outputting the predicted value. Model training utilizes a gradient descent algorithm to optimize the mean squared error loss function. The training process consists of four main steps: 1) Forward propagation calculates the predicted value, passing the input data through each layer of the network to obtain the output predicted value; 2) Loss calculation, comparing the predicted value with the actual value to calculate the mean squared error; 3) Backward propagation calculates the gradient, calculating the parameter gradient layer by layer from the output layer to the input layer; and 4) Parameter update, adjusting the network parameters based on the gradient direction. To improve training efficiency, mini-batch gradient descent is typically used, with a small batch of randomly selected samples selected for training. The Adam algorithm is used as the optimizer, combined with momentum and adaptive learning rate methods to accelerate convergence. To prevent overfitting, regularization techniques such as weight decay (adding a parameter norm penalty to the loss function) and early stopping (terminating training when validation set performance no longer improves) are introduced. The training process continues for multiple rounds, typically hundreds of iterations, until the prediction error on the validation set stabilizes or reaches a preset threshold, indicating that the model parameters have converged.

[0079] Based on the spatial correlation representation of power nodes and device state parameters, a Bayesian neural network is trained to quantify grid state uncertainty. Bayesian neural networks are probabilistic neural networks. Unlike traditional deterministic neural networks, their parameters are probability distributions rather than single numerical values. This property enables them to not only provide predictions but also quantify the uncertainty of the predictions. Training data includes the spatial correlation representation of power nodes (reflecting the mutual influence between nodes) and device state parameters (such as device temperature, vibration frequency, operating hours, and maintenance records). Due to the complexity of Bayesian neural networks, directly calculating the exact posterior distribution is generally not feasible. Therefore, variational inference is used to approximate it. Variational inference approximates the parameter posterior distribution to a simple, tractable distribution (such as a Gaussian distribution) and optimizes the approximation by minimizing the KL divergence between the two distributions. Specifically, Monte Carlo dropout can be used to randomly disable neurons during training and prediction. This is equivalent to averaging multiple different networks, resulting in a predicted distribution rather than a single point estimate. After training, the state assessment model not only provides device state assessment results (such as failure probability), but also provides confidence intervals for these estimates, quantifying the uncertainty of the predictions. This allows decision makers to distinguish between assessments with high certainty (which should lead to immediate action) and assessments with high uncertainty (which may require further monitoring or review).

[0080] Using the predictions from the load forecasting model and the evaluation results from the state assessment model, a deep reinforcement learning network is trained to optimize resource allocation strategies. Deep reinforcement learning combines deep learning and reinforcement learning, learning optimal policies through the interaction between an intelligent agent and its environment. First, a state space is defined, encompassing information such as load forecast results (future load distribution), state assessment results (equipment status and its uncertainty), and current resource states (such as power generation output, energy storage system charge and discharge status, and adjustable load conditions). Next, an action space is defined, encompassing possible actions such as power generation scheduling instructions (such as increasing or decreasing output), energy storage system control instructions (such as charging, discharging, and maintenance), and load switching operations (such as peak shaving and valley filling). Designing a reward function is a key step, comprehensively considering multiple objectives such as economic costs (fuel consumption and power losses), system reliability (voltage stability and frequency stability), and environmental performance (carbon emissions). A dual network architecture is employed, consisting of a policy network and a value network. The policy network inputs the current state and outputs a probability distribution over different actions, guiding the agent's action selection. The value network evaluates the long-term value of a state or state-action pair, assisting in policy optimization. The training process utilizes deep Q-learning or policy gradient methods, accumulating experience through interaction with the simulated environment. An experience replay buffer stores historical interaction data (state, action, reward, and next state), breaking temporal correlation between samples and improving training stability. To balance exploration and exploitation, an epsilon-greedy strategy is employed, selecting random actions with a certain probability instead of the currently optimal action to explore possible better strategies. After sufficient training, the resource optimization model can generate an optimal resource allocation strategy based on the current system state, balancing multiple optimization objectives.

[0081] For example, a regional power network consists of 50 major nodes (including 5 large substations, 15 medium-sized substations, and 30 distribution substations) and the transmission lines connecting them. Historical data consists of three months of continuous monitoring records. The data is first segmented chronologically, using the first 72 days of data as the training set and the last 18 days of data as the validation set. A sliding window method is used, inputting 7 consecutive days of data at a time, to predict the load distribution for the next day. The connection weights between nodes in the node relationship graph are calculated based on physical distance and communication quality. For example, the connection weight between two substations 2 kilometers apart with good communication quality is 0.82, while the connection weight between nodes 10 kilometers apart with average communication quality is only 0.45.

[0082] A spatiotemporal graph convolutional network (GCN) was applied to the training data to extract spatially dependent features. The network consists of three spatiotemporal convolutional blocks. The graph convolution layer in each block aggregates information from neighboring nodes, while the temporal convolution layer captures short-term fluctuations. Through network processing, it was found that load changes in a given area show a clear correlation with those in neighboring areas. For example, an increase in load in an industrial area is often accompanied by a slight decrease in load in surrounding residential areas. These extracted spatial correlation representations were fed into a long short-term memory (LSTM) network, which consists of two layers, each with 100 units. A gating mechanism was used to capture complex temporal patterns: load peaks occur between 7:00 and 9:00 a.m. and 6:00 and 8:00 p.m. on weekdays, while the load distribution is more flat on weekends. In summer, air conditioning load exhibits a lagged correlation with temperature, with a significant load increase occurring 1-2 hours after the temperature rises. The load forecasting model utilizes a four-layer feedforward neural network with a hidden layer size of [256, 128, 64] and a ReLU activation function. During training, the initial learning rate was set to 0.001 and decayed by 10% every 50 iterations. The model converged after 300 iterations on the training set, reducing the mean squared error from an initial 5.6MW² to 0.8MW². The error on the validation set was 1.2MW², demonstrating good generalization. The condition assessment model, using a Bayesian neural network, analyzed state parameters such as transformer temperature and vibration frequency and detected an anomaly in transformer No. 3. The resulting failure probability was 0.72, with a confidence interval of [0.57, 0.87], indicating a high risk and medium certainty. The resource optimization model, based on deep reinforcement learning, has a state space dimension of 128 and an action space containing control commands for five generators, three energy storage systems, and ten adjustable loads. The reward function comprehensively considers power loss, equipment lifespan, and carbon emissions, with weights of 0.5, 0.3, and 0.2, respectively. The model was trained using a dual deep Q-network and converged after 100,000 environmental interactions. The resulting resource scheduling strategy included adjusting the output of Generator Units 1 and 2 by 20MW and 15MW, respectively, based on predicted peak loads (8:00 AM and 7:00 PM). The energy storage system was scheduled to charge during low-load periods (2:00 AM to 4:00 AM) and discharge during peak periods, with a power setting of 10MW. Furthermore, the load on Transformer 3 was shifted, reducing its load factor from 85% to 65%, mitigating overload risks. This comprehensive scheduling solution met load demands, reduced the risk of equipment failure, and took both economic and environmental considerations into account.

[0083] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0084] (1) Collecting real-time operating parameters of each node in the power grid through a distributed sensor network, including load level, equipment temperature, transmission line impedance and phase angle difference, to form real-time resource data of the power grid;

[0085] (2) Input the real-time resource data of the power grid into the load forecasting model, calculate the load distribution forecast value for the next 24 hours, and generate a load forecast result map;

[0086] (3) Based on the real-time resource data of the power grid and the load forecast result map, Bayesian probabilistic reasoning is performed through the state assessment model to calculate the stability margin index and fault probability value of each node and construct the power grid stability assessment matrix;

[0087] (4) Perform threshold analysis on the grid stability assessment matrix, identify high-risk nodes and potential failure points, and generate a risk level classification table;

[0088] (5) Map the risk level classification table onto the power network topology map, mark the risk level and impact range, and form a fault risk distribution map;

[0089] (6) Based on the fault risk distribution map and risk propagation assessment results, a grid status assessment report and early warning instructions containing risk description, impact assessment and treatment suggestions are generated.

[0090] Specifically, a distributed sensor network collects real-time operating parameters from each node in the power grid. These distributed sensors, including intelligent power sensors, temperature sensors, vibration sensors, and phasor measurement units, are deployed at substations, key points along transmission lines, and distribution equipment. The data collected by these sensors includes load levels (indicating the magnitude of power load in megawatts), equipment temperatures (indicating the operating temperature of equipment such as transformers and circuit breakers in degrees Celsius), transmission line impedance (indicating the resistance and reactance of the line in ohms), and phase angle differences (indicating the voltage phase difference between different network nodes in degrees). This data is transmitted in real time to a data processing center via a 6G communication network, typically with a sampling frequency of seconds or minutes, generating real-time power grid resource data with high temporal and spatial resolution.

[0091] The collected real-time power grid resource data is fed into a trained load forecasting model to calculate a 24-hour load distribution forecast. This load forecasting model, derived from the aforementioned deep learning model training phase, comprises a hybrid architecture consisting of a spatiotemporal graph convolutional network and a long short-term memory network. The input data undergoes preprocessing, including missing value filling, outlier handling, and feature normalization. This preprocessed data is organized in the format required by the model, including historical load series, meteorological data (such as temperature and humidity), and temporal features (such as hour, day of the week, and holiday markers). The model generates hourly load forecasts for the next 24 hours, including the power load at each node and the total load for the entire network. The forecast results are organized in both temporal and spatial dimensions to form a load forecast map, visually displaying the future load distribution, including peak and off-peak periods, and spatial distribution characteristics.

[0092] Based on real-time power grid resource data and load forecast results, a state assessment model performs Bayesian probabilistic reasoning to calculate the stability margin index and failure probability of each node. Bayesian probabilistic reasoning is a statistical inference method based on Bayes' theorem that can handle uncertainty and integrate multi-source information. In this solution, the state assessment model is implemented using a Bayesian neural network to evaluate the state of each node. The formula for calculating the stability margin index (SII) is:

[0093]

[0094] in, Representation node The stability margin index ranges from [0,1], and the larger the value, the higher the stability margin; Representation node Maximum load capacity; Representation node The current load level; Node representing the load forecast model prediction Future load value; Representation node The current load value; Representation node The health status coefficient is calculated based on parameters such as equipment temperature and vibration frequency.

[0095] The failure probability value of the node is calculated through the Bayesian network:

[0096]

[0097] in, Represents a given observation data Under these conditions, the node the posterior probability of a failure; Representation node Data observed when a failure occurs The likelihood probability of ; Representation node The prior probability of failure is set based on historical statistical data; Represents observation data The calculated stability margin index and fault probability values for each node are organized into a matrix to form a power grid stability assessment matrix. The rows of the matrix represent different nodes, and the columns represent different assessment indicators (such as stability margin index, fault probability, temperature anomaly, etc.). This matrix representation facilitates subsequent threshold analysis and risk level classification.

[0098] Threshold analysis is performed on the grid stability assessment matrix to identify high-risk nodes and potential failure points. Threshold analysis consists of two parts: single-indicator threshold determination and multi-indicator comprehensive evaluation. Single-indicator threshold determination sets the risk threshold for each indicator. For example, a stability margin index below 0.3 is considered high risk, and a failure probability value above 0.6 is considered high risk. Multi-indicator comprehensive evaluation considers the combined impact of multiple indicators and calculates a comprehensive risk score using weighted summation or fuzzy logic methods. Based on the threshold analysis results, nodes are classified into three levels: high risk, medium risk, and low risk. The sources of risk (such as overload risk, equipment aging risk, and short-term load surge risk) are recorded in detail. This information is organized into a structured risk classification table, which contains fields such as node identification, risk level, risk source, and risk indicator value.

[0099] The risk classification table is mapped onto the power network topology map to visually display the risk distribution. The power network topology map is a visual representation constructed based on the physical coordinate mapping table of the power network topology. Nodes represent entities such as substations and distribution equipment, and edges represent transmission lines. During the mapping process, nodes of different risk levels are marked with different colors: high-risk nodes are red, medium-risk nodes are yellow, and low-risk nodes are green. Simultaneously, the risk impact range is calculated by analyzing the downstream nodes and regions that high-risk nodes may affect, determined through topological connectivity and power flow analysis. The impact range is marked on the topology map with dashed lines or shaded areas, supplemented by a quantitative description of the impact level. The resulting fault risk distribution map visually illustrates the spatial distribution and potential propagation paths of power grid risks. Based on the fault risk distribution map and risk propagation assessment results, a power grid status assessment report and early warning instructions are generated. The power grid status assessment report is a structured document consisting of four main sections: an overview of the overall system status, a detailed analysis of high-risk nodes, an analysis of risk propagation paths, and response recommendations. The overall system status overview summarizes the risk distribution across the entire network, such as the number and proportion of high-, medium-, and low-risk nodes. A detailed analysis of high-risk nodes details the risk indicators, risk causes, and risk levels for each high-risk node. The risk propagation path analysis analyzes the potential propagation paths and impact range of faults based on the grid topology and power flow characteristics. Response recommendations provide specific measures for different risks, such as equipment maintenance, load transfer, and emergency power generation. Early warning instructions are a structured set of instructions, graded by urgency, containing information such as risk node identification, risk type, processing time limit, and response measures, facilitating the rapid implementation of risk control measures.

[0100] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0101] (1) Extract key system stability indicators and load forecast data from the grid status assessment report to construct a resource scheduling constraint set;

[0102] (2) Based on the high-risk areas identified in the fault risk distribution map, calculate the risk coefficient and emergency power supply demand of each area to form a power supply priority table for the risk areas;

[0103] (3) Based on the fault type information and time estimate in the early warning instructions, formulate the grid dispatch schedule of different time scales and generate the dispatch cycle matrix;

[0104] (4) Based on the risk area power supply priority table and the dispatch cycle matrix, the optimal output allocation ratio is calculated for each type of power generation equipment, and a power generation equipment output plan including power generation start and stop time, power curve and ramp rate is generated;

[0105] (5) Based on the power generation equipment output plan and the grid load curve, calculate the power gap and surplus, allocate charging and discharging time periods and power values for the energy storage system, and form a charging and discharging strategy for the energy storage system that includes charging and discharging time, power size, and duration;

[0106] (6) Based on the power supply priority table of the risk area and the stability assessment results of each node, determine the load-shedding nodes and the load shedding order, and generate a load switching plan including the load switching time point, switching capacity and recovery order.

[0107] Specifically, key system stability indicators and load forecast data are extracted from the grid status assessment report to construct a resource scheduling constraint set. The grid status assessment report is a structured document that contains information such as an overview of the system's overall status, an analysis of high-risk nodes, risk propagation paths, and response recommendations. Natural language processing technology is used in the extraction process to identify key indicators in the report, including node stability margin index, failure probability, voltage deviation rate, and frequency stability. Load forecast data is also extracted, including the load forecast value and fluctuation range for each node for the next 24 hours. After standardization, this extracted data is converted into resource scheduling constraints, including generation capacity constraints, line transmission constraints, node voltage constraints, and system frequency constraints. These constraints collectively constitute the resource scheduling constraint set, providing boundary conditions for subsequent optimized scheduling.

[0108] Based on the high-risk areas identified in the fault risk distribution map, calculate the risk factor and emergency power supply demand for each area. The fault risk distribution map intuitively shows the spatial distribution of risks in the power grid, marking nodes and areas with different risk levels with different colors. For each high-risk area, calculate its risk factor:

[0109]

[0110] in, Indicates area The risk factor of Indicates the node index within the region; represents the failure probability of node k; represents the failure probability threshold, usually 0.6; Representation node The stability margin index of Indicates the stability margin threshold, usually 0.3; and is the weight coefficient, which adjusts the relative importance of failure probability and stability margin in risk assessment; Indicates the criticality of node k, taking into account the importance and influence of the node.

[0111] At the same time, calculate the emergency power supply demand of each area:

[0112]

[0113] in, Indicates area Emergency power supply demand; represents the predicted load value of node k; represents the safety load threshold of node k; is an indicator function, which takes the value of 1 when the stability margin index of node k is lower than the threshold, and takes the value of 0 otherwise.

[0114] Based on the fault type information and time estimate in the early warning instructions, a grid dispatch schedule with different time scales is developed. The early warning instructions contain information such as the degree of urgency, fault type, estimated occurrence time, and duration. Based on this information, the dispatch time is divided into multiple time scales: emergency response time scale (minutes), short-term dispatch time scale (hours), and medium-term dispatch time scale (days). For each time scale, the start and end times, dispatch cycle, and dispatch step size of the dispatch are determined. This information is organized into a dispatch cycle matrix, where the rows of the matrix represent different time scales and the columns represent the dispatch parameters (start and end times, cycle, step size). The dispatch cycle matrix provides a time framework for subsequent resource scheduling, ensuring that the corresponding dispatch operations are executed at the appropriate time.

[0115] Based on the risk area power supply priority table and the dispatch cycle matrix, the optimal output allocation ratio of each type of power generation equipment is calculated. For each power generation equipment, its technical characteristics are first considered, including maximum output, minimum output, ramp rate limit, start and stop time constraints, etc. The output allocation of power generation equipment uses the following formula:

[0116]

[0117] in, represents the output value of the power generation equipment g at time point t; and Respectively represent the minimum and maximum output of the power generation equipment g; Indicates area The power demand at time t; Indicates area Power supply priority; Indicates the power generation equipment g for the area The power supply accessibility is represented by h, which is an index variable representing all power generation equipment in the system.

[0118] At the same time, consider the ramp rate constraints of power generation equipment:

[0119]

[0120] in, Indicates the ramp rate limit of the power generation equipment g; Indicates the interval between adjacent scheduling time points.

[0121] Based on the power generation equipment output plan and the grid load curve, the system calculates power shortages and surpluses, allocating charging and discharging periods and power levels to the energy storage system. Based on the power supply priority table for risky areas and the stability assessment results of each node, it determines the nodes that can be shelved and the order in which they should be shelved, generating a load switching plan.

[0122] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0123] (1) Convert the power generation equipment output plan, energy storage system charging and discharging strategy, and load switching plan into a standardized control instruction format to generate an instruction data packet;

[0124] (2) Prioritize the command data packets and classify them into three levels according to their urgency: critical control commands, routine scheduling commands, and auxiliary management commands;

[0125] (3) Dynamically allocate communication resources through 6G network slicing technology, creating ultra-reliable low-latency network slices for critical control instructions, mobile broadband slices for routine scheduling instructions, and machine-type communication slices for auxiliary management instructions;

[0126] (4) Perform node synchronization on the control instructions to ensure the time synchronization and logical consistency of multi-node control and generate a synchronous control instruction set;

[0127] (5) Send the synchronous control instruction set to each control execution unit through each network slice, and collect execution feedback data to form power network operation status data including parameter measured values, equipment status and instruction execution results;

[0128] (6) Based on the power network operation status data, a multi-dimensional comparative analysis of the real-time operation parameters of the power network is carried out, the deviation values and fluctuation curves of various parameters are recorded, and the corresponding control instruction execution status is associated to form a performance evaluation report including the power system operation quality evaluation, control instruction execution effect and resource scheduling rationality.

[0129] Specifically, the power generation equipment output plan, energy storage system charge and discharge strategy, and load switching plan are converted into a standardized control instruction format to generate an instruction data packet. This involves three types of data conversion: the power generation equipment output plan is converted into a power setting instruction, which contains parameters such as the target power value, execution time point, and ramp rate limit; the energy storage system charge and discharge strategy is converted into an energy storage control instruction, which contains information such as the charge and discharge mode, power level, and duration; and the load switching plan is converted into a load control instruction, which contains data such as the disconnection / restore flag, target load amount, and execution time point. The conversion uses a unified data structure, which usually includes three parts: an instruction header (sending time, receiving object, and instruction type), an instruction body (specific parameter settings), and a checksum (to ensure data integrity). This standardized format helps different types of equipment understand and execute control instructions, while also facilitating the transmission and management of instructions.

[0130] Generated command packets are prioritized and divided into three levels based on urgency. Critical control instructions involve core operations for safe and stable system operation, such as emergency power generation adjustments and emergency load shedding to address sudden failures. These instructions have the highest priority and must be executed immediately without delay. Routine dispatch instructions, including normal operation instructions such as daily power balancing and economic optimization dispatch, have a medium priority and can tolerate short delays but must be completed as planned. Auxiliary management instructions involve non-critical operations such as status monitoring, data collection, and parameter fine-tuning. They have a relatively low priority and less stringent execution time requirements. Priority marking uses a binary flag method, adding a priority field to the instruction header. A two-bit binary code represents three levels (00 for auxiliary management, 01 for routine dispatch, and 10 for critical control). Priority marking ensures that critical instructions are transmitted and executed first when network resources are limited.

[0131] 6G network slicing technology dynamically allocates communication resources, creating dedicated network slices for commands of different priorities. Network slicing, a core technology of 6G networks, uses virtualization to create multiple logically independent end-to-end networks on the same physical network, each with unique quality of service characteristics. Ultra-reliable, low-latency network slices are created for critical control commands. These slices have a transmission latency of less than 1 millisecond and utilize dedicated spectrum resources and redundant transmission paths to ensure reliable transmission of commands even in emergency situations. Enhanced mobile broadband slices are created for routine scheduling commands, offering high bandwidth and medium latency (5-10 milliseconds) to meet the transmission requirements of large amounts of scheduling data. Massive machine-type communication slices are created for auxiliary management commands. These slices support large numbers of connected devices. Although individual connections have lower bandwidth, they can effectively handle large amounts of non-critical data exchange. Network slices are created using software-defined networking technology, with a centralized controller dynamically allocating network resources and adjusting slice parameters based on real-time communication needs.

[0132] A distributed consensus algorithm is used to synchronize control instructions among nodes, ensuring the time synchronization and logical consistency of multi-node control. In power grids, control operations at multiple nodes often require coordinated execution, such as the joint scheduling of multiple generators and coordinated load transfer across multiple distribution areas. The distributed consensus algorithm is based on a two-phase commit protocol. It first sends a pre-execution message to all execution nodes to collect their readiness status. Once all nodes are confirmed to be ready, a formal execution message is sent to trigger synchronous execution. To address time asynchrony caused by network latency differences, the Precision Time Protocol (PTP) is used to achieve microsecond-level clock synchronization, allowing all nodes to execute instructions based on the same time base. This mechanism generates a set of synchronized control instructions containing a global timestamp and execution sequence number, ensuring the coordination of complex operations.

[0133] Synchronous control instruction sets are distributed to each control execution unit via their respective network slices, and execution feedback data is collected. This distribution process utilizes a combination of point-to-point and multicast methods. Critical control instructions are typically transmitted directly point-to-point to ensure reliable delivery; routine scheduling instructions and auxiliary management instructions can be transmitted via multicast to improve transmission efficiency. After the instructions are distributed, each execution unit returns confirmation information, including instruction receipt status, execution progress, and results. During execution, the control execution unit continuously collects equipment operating data, including measured values of parameters such as actual output power, voltage and current, and device temperature, as well as equipment operating status information (e.g., normal, alarm, or fault). This data, along with the instruction execution results (success, partial success, or failure), is uploaded to the data processing center via the 6G network to generate power network operating status data. This data is stored in a time series database to facilitate subsequent time series analysis and performance evaluation.

[0134] Based on power grid operating status data, a multi-dimensional comparative analysis of real-time operating parameters is conducted. First, parameter deviation analysis is performed, calculating the difference between actual values and command setpoints, such as power generation deviation and voltage deviation, to assess control accuracy. Next, a time response analysis is performed, measuring the time from command issuance to execution to assess control timeliness. Next, a stability analysis is performed, recording parameter fluctuation curves, calculating standard deviations and maximum fluctuation amplitudes, and assessing operational stability. Finally, a correlation analysis is performed, correlating parameter changes with control command execution to identify causal relationships. These analyses generate a performance evaluation report that includes an assessment of power system operational quality, control command execution effectiveness, and resource scheduling rationality. This report not only summarizes the current control effectiveness but also provides data support for subsequent scheduling strategy optimization.

[0135] The above describes the intelligent management method of power grid resources based on 6G and deep learning in the embodiment of the present application. The following describes the intelligent management system of power grid resources based on 6G and deep learning in the embodiment of the present application. Figure 2 In the embodiments of the present application, an embodiment of an intelligent management system for power grid resources based on 6G and deep learning includes:

[0136] The acquisition module is used to collect power network node data through 6G communication base stations and multi-source positioning sensors to obtain the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information;

[0137] A processing module is used to pre-process the power network topology physical coordinate mapping table, the real-time power parameter data stream and the communication network status information to obtain the power data feature tensor;

[0138] The training module is used to train a deep learning model cluster based on the power data feature tensor through a hybrid architecture of spatiotemporal graph convolutional neural network and long short-term memory network to obtain load forecasting model, state assessment model and resource optimization model;

[0139] An analysis module is used to collect real-time resource data of the power grid and perform grid status analysis on the real-time resource data of the power grid according to a load forecasting model, a status assessment model and a resource optimization model to obtain a grid status assessment report, a fault risk distribution map and an early warning instruction;

[0140] A calculation module is used to calculate a resource scheduling plan based on the grid status assessment report, the fault risk distribution map, and the early warning instructions, and obtain a power generation equipment output plan, an energy storage system charging and discharging strategy, and a load switching plan;

[0141] The control module is used to issue control instructions through 6G network slicing based on the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan to obtain power network operation status data and performance evaluation reports.

[0142] Through the collaborative efforts of these components, 6G communication base stations and multi-source positioning sensors collect data from power network nodes, generating a mapping table of the power network topology's physical coordinates, real-time power parameter data streams, and communication network status information. This addresses the issues of insufficient data acquisition accuracy and poor real-time performance in traditional technologies, providing a high-quality data foundation for subsequent intelligent analysis. The technical feature of preprocessing collected data to form a power data feature tensor effectively removes noise and redundancy from the raw data, improving data quality and enabling deep learning models to more accurately extract key features of the power network. The hybrid architecture of a spatiotemporal graph convolutional neural network and a long-short-term memory network fully considers the spatial topology and time series characteristics of the power network. Compared to traditional single-structure neural networks, this hybrid architecture can simultaneously capture the spatial dependencies between power nodes and the temporal patterns of load changes. The resulting load forecasting models, state assessment models, and resource optimization models are more accurate and have stronger generalization capabilities. The technical feature of collecting real-time power grid resource data and using the aforementioned model to analyze the grid status enables early identification and risk assessment of abnormal grid conditions. Compared to traditional threshold-based monitoring methods, this deep learning-based state analysis can discover complex nonlinear relationships and potential risk patterns, greatly improving the accuracy of grid security warnings. The technical feature of calculating resource scheduling plans based on grid status assessment results enables the coordinated optimization of power generation, energy storage, and load. Compared with traditional single-resource optimization, this multi-resource coordinated scheduling significantly improves system flexibility and resource utilization efficiency. The technical feature of issuing control instructions through 6G network sharding ensures ultra-reliable and low-latency transmission of critical control instructions, solving the congestion and delay problems that may occur in traditional communication networks in emergency situations, and enabling control instructions to be executed accurately and promptly.

[0143] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0144] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0145] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0146] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0149] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for intelligent management of power grid resources based on 6G and deep learning, characterized in that: The 6G and deep learning-based intelligent management method for power grid resources includes: Through 6G communication base stations and multi-source positioning sensors, power network node data is collected to obtain the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information; Performing data preprocessing on the power network topology physical coordinate mapping table, the real-time power parameter data stream, and the communication network status information to obtain a power data feature tensor; Based on the power data feature tensor, a deep learning model cluster is trained through a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load forecasting model, a state assessment model, and a resource optimization model; Collecting real-time resource data of the power grid, and performing grid status analysis on the real-time resource data of the power grid according to the load forecasting model, the status assessment model and the resource optimization model, to obtain a grid status assessment report, a fault risk distribution map and an early warning instruction; Calculate resource scheduling plans based on the grid status assessment report, fault risk distribution map, and early warning instructions to obtain power generation equipment output plans, energy storage system charging and discharging strategies, and load switching plans; According to the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan, control instructions are issued through 6G network slicing to obtain power network operation status data and performance evaluation reports.

2. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: The method collects power network node data through 6G communication base stations and multi-source positioning sensors to obtain a power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information, including: Deploy 6G communication modules at each substation, key points of transmission lines, and distribution equipment in the power network to measure the signal strength and transmission delay between nodes and generate a node communication connection matrix; Intelligent power sensors collect voltage, current, active power, reactive power, and frequency values at each node of the power network every second, forming a real-time power parameter data stream; Use GPS receivers and base station-assisted positioning systems to collect the precise location of the device, and use the Kalman filter algorithm to process the raw positioning data to obtain centimeter-level device spatial coordinates; Performing point-to-point matching of the link status data in the node communication connection matrix with the device space coordinates to construct a power network topology physical coordinate mapping table including physical distance and link quality; By measuring the signal-to-noise ratio, bit error rate, and available bandwidth of the 6G communication link, the real-time status parameters of each communication channel are recorded to generate communication network status information; The data points in the real-time power parameter data stream are timestamped and aligned to eliminate sampling time differences between different devices and complete data synchronization processing of power network nodes.

3. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: The data preprocessing of the power network topology physical coordinate mapping table, the real-time power parameter data stream and the communication network status information to obtain the power data feature tensor includes: Performing wavelet transform decomposition on the real-time power parameter data stream to separate high-frequency noise components and low-frequency trend components in the power signal to form a power signal component matrix; Using an autoencoder to detect and repair abnormal values and missing values in the power signal component matrix to generate a cleaned power parameter matrix; Constructing a node adjacency relationship graph based on the power network topology physical coordinate mapping table and calculating the connection weight value between each node; Normalizing the signal-to-noise ratio, bit error rate, and bandwidth data in the communication network state information to form a communication state feature vector; Associating and fusing the cleaned power parameter matrix with the node adjacency graph and the communication state feature vector to construct a multi-dimensional association matrix; The multi-dimensional correlation matrix is tensor decomposed using the Tucker decomposition algorithm. The time dimension core tensor rank is set to 24, the space dimension core tensor rank is set to 30% of the number of network nodes, and the physical property dimension core tensor rank is set to 8. Multilinear feature extraction is performed to obtain the power data feature tensor.

4. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: According to the power data feature tensor, a deep learning model cluster is trained through a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load forecasting model, a state assessment model, and a resource optimization model, including: Dividing the power data feature tensor into a training set and a validation set according to the time dimension, establishing a power network node relationship diagram, and constructing a model training data set; Applying a spatiotemporal graph convolutional network to the model training dataset to extract spatial dependency features between power nodes and generate a representation of the spatial correlation of power nodes; Inputting the spatial correlation representation of the power nodes into the long short-term memory network to capture the temporal variation law of the power load and obtain the time correlation characteristics; Based on the time correlation characteristics and historical load data, a load forecasting model is constructed, a loss function is optimized by a gradient descent algorithm, and model parameters are adjusted until the forecast error converges; Based on the spatial correlation representation of the power nodes and the device state parameters, a Bayesian neural network is trained to quantify the uncertainty of the power grid state and form a state assessment model; The prediction results of the load forecasting model and the evaluation results of the state evaluation model are used to train a deep reinforcement learning network, and the resource allocation strategy is optimized through a reward and punishment mechanism to obtain a resource optimization model.

5. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: The collecting of real-time resource data of the power grid and performing grid status analysis on the real-time resource data of the power grid according to the load forecasting model, the status assessment model and the resource optimization model to obtain a grid status assessment report, a fault risk distribution map and an early warning instruction include: The distributed sensor network collects real-time operating parameters of each node in the power grid, including load level, equipment temperature, transmission line impedance and phase angle difference, to form real-time resource data of the power grid; Inputting the real-time resource data of the power grid into the load forecasting model, calculating the load distribution forecast value for the next 24 hours, and generating a load forecast result map; Based on the real-time resource data of the power grid and the load forecast result map, Bayesian probabilistic reasoning is performed through the state assessment model to calculate the stability margin index and fault probability value of each node, and construct a power grid stability assessment matrix; Performing threshold analysis on the grid stability assessment matrix, identifying high-risk nodes and potential failure points, and generating a risk level classification table; Mapping the risk level classification table onto the power network topology map, marking the risk level and impact range, and forming a fault risk distribution map; Based on the fault risk distribution map and the risk propagation assessment results, a power grid status assessment report and early warning instructions including risk description, impact assessment and treatment suggestions are generated.

6. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: The resource scheduling plan is calculated based on the grid status assessment report, fault risk distribution map and early warning instructions to obtain the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan, including: Extracting key system stability indicators and load forecast data from the power grid status assessment report to construct a resource scheduling constraint set; Calculate the risk coefficient and emergency power supply demand of each area according to the high-risk areas identified by the fault risk distribution map, and form a power supply priority table for the risk areas; Based on the fault type information and time estimate in the warning instruction, formulate a grid dispatch schedule of different time scales and generate a dispatch cycle matrix; Based on the risk area power supply priority table and the scheduling cycle matrix, the optimal output allocation ratio is calculated for each type of power generation equipment, and a power generation equipment output plan including power generation start and stop times, power curves, and ramp rates is generated; Based on the power generation equipment output plan and the grid load curve, the power gap and surplus are calculated, and the charging and discharging time periods and power values are allocated to the energy storage system to form a charging and discharging strategy for the energy storage system that includes charging and discharging time, power size, and duration; Based on the risk area power supply priority table and the stability assessment results of each node, the load-shedding nodes and the load shedding order are determined, and a load switching plan including load switching time points, switching capacity and restoration order is generated.

7. The method for intelligent management of power grid resources based on 6G and deep learning according to claim 1, characterized in that: According to the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan, control instructions are issued through 6G network slicing to obtain power network operation status data and performance evaluation reports, including: Convert the power generation equipment output plan, energy storage system charging and discharging strategy, and load switching plan into a standardized control instruction format to generate an instruction data packet; Priority marking is performed on the instruction data packets, and they are divided into three levels according to the urgency: critical control instructions, routine scheduling instructions, and auxiliary management instructions; Dynamically allocate communication resources through 6G network slicing technology to create ultra-reliable, low-latency network slices for the critical control instructions, mobile broadband slices for the routine scheduling instructions, and machine-type communication slices for the auxiliary management instructions; Perform node synchronization on the control instructions to ensure time synchronization and logical consistency of multi-node control and generate a synchronous control instruction set; The synchronous control instruction set is issued to each control execution unit through each network slice, and execution feedback data is collected to form power network operation status data including parameter measured values, equipment status and instruction execution results; Based on the power network operation status data, a multi-dimensional comparison and analysis is performed on the real-time operation parameters of the power network, the deviation values and fluctuation curves of various parameters are recorded, and the corresponding control instruction execution status is associated to form a performance evaluation report that includes power system operation quality evaluation, control instruction execution effect and resource scheduling rationality.

8. An intelligent management system for power grid resources based on 6G and deep learning, for implementing the intelligent management method for power grid resources based on 6G and deep learning as described in any one of claims 1 to 7, characterized in that: The 6G and deep learning-based intelligent management system for power grid resources includes: The acquisition module is used to collect power network node data through 6G communication base stations and multi-source positioning sensors to obtain the power network topology physical coordinate mapping table, real-time power parameter data stream and communication network status information; a processing module, configured to perform data preprocessing on the power network topology physical coordinate mapping table, the real-time power parameter data stream, and the communication network status information to obtain a power data feature tensor; A training module is used to train a deep learning model cluster based on the power data feature tensor through a hybrid architecture of a spatiotemporal graph convolutional neural network and a long short-term memory network to obtain a load forecasting model, a state assessment model, and a resource optimization model; An analysis module is used to collect real-time resource data of the power grid and perform grid status analysis on the real-time resource data of the power grid according to the load forecasting model, the status assessment model and the resource optimization model to obtain a grid status assessment report, a fault risk distribution map and an early warning instruction; A calculation module is used to calculate a resource scheduling plan based on the grid status assessment report, the fault risk distribution map, and the early warning instructions, and obtain a power generation equipment output plan, an energy storage system charging and discharging strategy, and a load switching plan; The control module is used to issue control instructions through 6G network slicing based on the power generation equipment output plan, energy storage system charging and discharging strategy and load switching plan to obtain power network operation status data and performance evaluation reports.

9. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements the power grid resource intelligent management method based on 6G and deep learning as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the method for intelligent management of power grid resources based on 6G and deep learning as described in any one of claims 1 to 7.

Images