Power grid data connection method based on data analysis

Abstract

The invention relates to the technical field of power grid data processing, and discloses a power grid data connection method based on data analysis, and the method comprises the steps: collecting the operation data of power grid equipment through a distributed sensor network, carrying out the denoising, unifying the timestamp through a DTW algorithm, and obtaining the power grid data; establishing a distributed database based on a block chain, and classifying and storing the power grid data according to different types and levels; constructing a dynamic topological graph of the power grid, encoding the dynamic topological graph by utilizing a GNN network, and capturing a spatial dependency relationship among equipment; fusing the output of the Transform architecture and the output of the CNN network to obtain a comprehensive feature vector; inputting the comprehensive feature vector into an LSTM-GAN model to predict the load and fault probability of the power grid, and generating an equipment fault risk index in combination with the dynamic topological graph and a prediction result; an early warning signal is triggered, a control strategy is generated, and cross-regional power grid data collaboration is carried out through federated learning; according to the invention, efficient communication and deep utilization of power grid data are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid data processing, and in particular to a power grid data penetration method based on data analysis. BACKGROUND

[0002] With the continuous expansion of the power grid scale and the improvement of the intelligent degree, a large amount of data is generated in the power grid, which is scattered in different systems and departments, and the data formats, standards and interfaces are different, forming many data islands, which is difficult to realize cross-domain collaboration. The existing method relies on fixed rules or simple statistical models, and cannot quickly respond to the dynamic changes of the power grid. SUMMARY

[0003] The purpose of the present application is to solve the above problems, and a power grid data penetration method based on data analysis is designed.

[0004] The first aspect of the present application provides a power grid data penetration method based on data analysis, which comprises the following steps:

[0005] The operation data of the power grid equipment is collected by using a distributed sensor network, the operation data is denoised by using Kalman filtering, and the time stamp is unified by using a DTW algorithm to obtain power grid data;

[0006] A distributed database based on a block chain is established, and the power grid data is classified and stored according to different types and levels;

[0007] A dynamic topology graph of the power grid is constructed, and the dynamic topology graph is coded by using a GNN network to capture the spatial dependence relationship between the devices;

[0008] The time series data in the power grid data is extracted by using a Transformer architecture, the unstructured data in the power grid data is encoded by using a CNN network, and the outputs of the Transformer architecture and the CNN network are fused to obtain a comprehensive feature vector;

[0009] The comprehensive feature vector is input into an LSTM-GAN model to predict the power grid load and the fault probability, and the device fault risk index is generated in combination with the dynamic topology graph and the prediction result;

[0010] According to the device fault risk index and the real-time operation state, a warning signal is triggered and a control strategy is generated, and data collaboration of the cross-regional power grid is realized through federated learning.

[0011] Optionally, in the first implementation manner of the first aspect of the present application, the operation data of the power grid equipment is collected by using a distributed sensor network, and the operation data is denoised by using Kalman filtering, comprising:

[0012] A state space model of Kalman filtering is established for the collected operation data, and initial parameters are set;

[0013] For the data collected by the sensor at each time point, a state estimation value at the current time is predicted according to a state estimation at the previous time and a state transition matrix, and an error covariance at the current time is predicted by using an error covariance matrix and a process noise covariance matrix;

[0014] The predicted state estimation value at the current time is compared with the actual observation value, a Kalman gain is calculated, the predicted state estimation value is corrected according to the Kalman gain, an optimal state estimation value at the current time is obtained, and the error covariance matrix is updated to obtain the denoised operation data.

[0015] Optionally, in the second implementation manner of the first aspect of the present application, the step of obtaining the power grid data by unifying the time stamps by using the DTW algorithm comprises the following steps.

[0016] The operation data after the Kalman filtering denoising processing is used to construct a time series distance matrix, wherein the distance matrix stores the distance between each data point in two time series;

[0017] An accumulated distance matrix is initialized, the accumulated distance value of each element in the accumulated distance matrix is calculated, the accumulated distance matrix is gradually filled from the starting point, and all elements are calculated until all elements are calculated, wherein the accumulated distance matrix has the same size as the time series distance matrix;

[0018] The optimal alignment path is found by backtracking from the right lower corner element of the accumulated distance matrix, the time stamps of the two time series are unified according to the optimal alignment path, and finally the power grid data with unified time stamps is obtained.

[0019] Optionally, in the third implementation manner of the first aspect of the present application, the step of constructing the power grid dynamic topology graph and using the GNN network to encode the dynamic topology graph to capture the spatial dependency between devices comprises the following steps.

[0020] The power grid dynamic topology graph is constructed by taking the power grid device as the node of the topology graph and taking the physical connection between the devices as the edge of the topology graph;

[0021] For each node in the dynamic topology graph, the corresponding operation data and attribute information of the node are extracted as the node features, and the node features are encoded and converted to form the feature vector of each node;

[0022] The feature vector of each node is input into the GNN network, the feature information of each node is propagated to its neighbor nodes along the edges, and each node aggregates the information from the neighbor nodes with the original feature to obtain a new intermediate feature representation;

[0023] The intermediate features are processed by an activation function to update the feature vector of the node, and the spatial dependence between devices is captured through multi-layer iteration of the GNN network.

[0024] Optionally, in a fourth implementation form of the first aspect of the application, the feature extraction of the time series data in the power grid data by using the Transformer architecture comprises:

[0025] The time series data is divided into multiple sequence segments by using a time window, and each segment is taken as an input of the Transformer architecture;

[0026] The Transformer architecture captures the dependence between different positions of the time series in different representation subspaces through a multi-head attention mechanism, and extracts long-term dependence features in the time series data.

[0027] Optionally, in a fifth implementation form of the first aspect of the application, the feature encoding of the unstructured data in the power grid data by using the CNN network and the fusion of the outputs of the Transformer architecture and the CNN network to obtain a comprehensive feature vector comprise:

[0028] The convolutional layers and the pooling layers of the CNN network are used to gradually extract local features and global features of the unstructured data to obtain high-level semantic feature representations;

[0029] The outputs of the Transformer architecture and the CNN network are spliced or fused by using a weighted summation method, a gating mechanism is introduced, and the weights of the features are adaptively adjusted to obtain a comprehensive feature vector.

[0030] Optionally, in a sixth implementation form of the first aspect of the application, the input of the comprehensive feature vector into the LSTM-GAN model to predict the power grid load and the fault probability, and the generation of a device fault risk index in combination with a dynamic topology graph and a prediction result comprise:

[0031] The comprehensive feature vector is input into the LSTM-GAN model, the LSTM network learns the dynamic change rule of the time series through a memory unit and a gating mechanism;

[0032] The GAN network is composed of a generator and a discriminator, and the fitting of the data distribution is processed through the adversarial training of the generator and the discriminator;

[0033] The LSTM-GAN model generates a predicted value of future load, and predicts a probability of a device fault in a future period of time according to device operation data and the comprehensive feature vector, and generates a device fault risk index in combination with a dynamic topology graph and a prediction result.

[0034] The second aspect of the application provides a power grid data through system based on data analysis, which comprises:

[0035] A collection module is configured to collect operation data of power grid equipment by using a distributed sensor network, to denoise the operation data by using Kalman filtering, and to unify timestamps by using a DTW algorithm to obtain power grid data.

[0036] A storage module is configured to establish a distributed database based on a block chain, and to store the power grid data according to different types and levels.

[0037] An encoding module is configured to construct a dynamic topology graph of the power grid, to encode the dynamic topology graph by using a GNN network, and to capture spatial dependency between devices.

[0038] A fusion module is configured to extract features of time series data in the power grid data by using a Transformer architecture, to encode features of unstructured data in the power grid data by using a CNN network, to fuse outputs of the Transformer architecture and the CNN network, and to obtain a comprehensive feature vector.

[0039] A prediction module is configured to input the comprehensive feature vector into an LSTM-GAN model to predict power grid load and failure probability, to combine a dynamic topology graph and a prediction result, and to generate a device failure risk index.

[0040] A data collaboration module is configured to trigger an early warning signal and generate a control strategy according to the device failure risk index and a real-time operation state, and to perform data collaboration of cross-regional power grids by using federated learning.

[0041] The third aspect of the application provides a power grid data through device based on data analysis, which comprises a memory and at least one processor, and the memory stores instructions; the at least one processor invokes the instructions in the memory, so that the power grid data through device based on data analysis performs each step of the power grid data through method based on data analysis.

[0042] The fourth aspect of the application provides a computer readable storage medium, which stores instructions, and the instructions are executed by a processor to realize each step of the power grid data through method based on data analysis.

[0043] The technical scheme provided by the application adopts a distributed sensor network to collect operation data of power grid equipment, adopts Kalman filtering to denoise the operation data, and obtains power grid data through a DTW algorithm to unify timestamps; a distributed database based on a block chain is established, and the power grid data is stored in a classified manner according to different types and levels; a power grid dynamic topology graph is constructed, a GNN network is used to code the dynamic topology graph, and spatial dependency between equipment is captured; a Transformer architecture is used to extract features of time series data in the power grid data, a CNN network is used to code features of unstructured data in the power grid data, the outputs of the Transformer architecture and the CNN network are fused to obtain a comprehensive feature vector; the comprehensive feature vector is input into an LSTM-GAN model to predict power grid load and fault probability, a dynamic topology graph and a prediction result are combined to generate an equipment fault risk index; according to the equipment fault risk index and a real-time operation state, an early warning signal is triggered and a control strategy is generated, data of cross-regional power grids is cooperated through federated learning; the application realizes real-time modeling of the influence of topology changes on data association, unified processing of structured and unstructured data, improvement of feature extraction capability, accurate data collection and preliminary processing, effective removal of noise, improvement of data accuracy and reliability, guarantee of data security, integrity and traceability based on the distributed database based on the block chain, prevention of malicious tampering and loss of data, deep analysis and mining of massive power grid data, more scientific decision support for power grid operation management, breaking of data silos, efficient penetration and deep utilization of power grid data, and improvement of the intelligent level and work efficiency of power grid operation management. BRIEF DESCRIPTION OF DRAWINGS

[0044] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not considered as limiting the application.

[0045] Figure 1 A first embodiment schematic diagram of the power grid data penetration method based on data analysis provided by the embodiment of the application;

[0046] Figure 2 A second embodiment schematic diagram of the power grid data penetration method based on data analysis provided by the embodiment of the application;

[0047] Figure 3 A structure schematic diagram of the power grid data penetration equipment based on data analysis provided by the embodiment of the application. DETAILED DESCRIPTION

[0048] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application, and above-described drawings, if any, are used to distinguish between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so construed can be interchanged, under appropriate circumstances, and that the embodiments described herein can be carried out in other sequences than the one illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, article, or apparatus that comprises a list of steps or units can not necessarily be limited to those steps or units which are clearly recited, but can include other not expressly recited steps or units, or steps or units which are sensible additions carrying forward the purpose of those processes, methods, articles, or apparatus.

[0049] For the sake of understanding, the specific flow of the embodiments of the present application is described below, please refer to Figure 1 The first embodiment of the power grid data through method based on data analysis provided by the embodiments of the present application is shown in the figure, and the method specifically comprises the following steps:

[0050] In step 101, the operation data of the power grid equipment is collected by using the distributed sensor network, the Kalman filter is used to denoise the operation data, and the DTW algorithm is used to unify the time stamp, so as to obtain the power grid data.

[0051] In this embodiment, according to the distribution of the power grid equipment and the monitoring demand, the distributed sensor network architecture is designed, and multiple types of sensors are deployed at the key equipment nodes of the transformer, circuit breaker and transmission line, including voltage sensor, current sensor, temperature sensor, vibration sensor and the like. According to different equipment characteristics and monitoring parameters, the type, quantity and installation position of the sensor are determined to ensure that the operation data of the power grid equipment can be comprehensively and real-timely collected. The distributed sensor network is started, and the operation parameters of the power grid equipment are collected in real time according to the preset sampling frequency. The voltage sensor collects the voltage value of the equipment, the current sensor collects the current value, the temperature sensor monitors the temperature change of the key part of the equipment, and the vibration sensor detects the vibration frequency of the equipment. Each sensor works independently, records the collected original data in time sequence, and forms multi-dimensional time sequence data.

[0052] In this embodiment, the running data after Kalman filter denoising processing is used to construct a time series distance matrix, wherein the distance matrix stores the distance between each data point in the two time series; an accumulated distance matrix is initialized, the accumulated distance value of each element in the accumulated distance matrix is calculated, the accumulated distance matrix is gradually filled from the starting point, and all elements are calculated until all elements are calculated, wherein the accumulated distance matrix has the same size as the time series distance matrix; starting from the right lower corner element of the accumulated distance matrix, the optimal alignment path is found backtracking, and the timestamps of the two time series are unified according to the optimal alignment path, and finally the power grid data with unified timestamps is obtained.

[0053] In step 102, a distributed database based on a blockchain is established, and the power grid data is classified and stored according to different types and levels.

[0054] In this embodiment, the number and deployment position of the nodes of the blockchain network are determined, the blockchain nodes are set at the key nodes such as the core data centers and substations in each region of the power grid, the practical Byzantine fault tolerance algorithm is selected as the consensus mechanism, the network communication configuration of each node is initialized, the secure communication channel between the nodes is established, it is ensured that the nodes can discover and connect with each other, and the distributed database underlying network architecture based on the blockchain is constructed; the experts in the power field, the data management personnel and the security personnel are organized, the detailed data classification standard is formulated according to the source, the use and the sensitivity of the power grid data, the power grid data is divided into main types such as user electricity consumption data, power grid equipment asset data and equipment operation data; further, according to the sensitive level of the data, the data is divided into three levels of public data, internal data and sensitive data, for example, the basic electricity consumption of the user belongs to the internal data, and the personal privacy information such as the user's ID card number belongs to the sensitive data.

[0055] For different levels of power grid data, a suitable encryption algorithm is selected, for sensitive data, a hybrid encryption method of asymmetric encryption algorithm combined with symmetric encryption algorithm is used, the data is first quickly encrypted by using the AES algorithm, and then the key of the AES is encrypted by using the RSA algorithm; for internal data, the symmetric encryption algorithm is used for encryption; for public data, no encryption or only simple hash algorithm is used to generate data digest for data integrity verification;

[0056] According to the characteristics and storage requirements of the power grid data, the data sharding rule is formulated, the equipment operation data is divided into transformer data shards, circuit breaker data shards and the like according to the equipment type; the user electricity consumption data is divided into shards of different regions according to the region; each type of data is further divided into shards of different time periods according to the data generation time, the storage node allocation strategy of each shard is determined, for example, the device data shards of the same substation are stored on adjacent several blockchain nodes, so as to improve the data reading efficiency;

[0057] The collected original power grid data is classified and labeled according to the established data classification standard, the data that needs to be encrypted is encrypted according to the corresponding encryption strategy, the data that meets the fragmentation rule is fragmented, and a unique identifier is generated for each data fragment, and metadata information is added to each data fragment, including data type, level, creation time, device or user identifier, etc., so as to facilitate subsequent data management and retrieval;

[0058] After each node receives the preprocessed power grid data fragment, the relevant information of the data fragment such as the identifier and metadata is packaged into a transaction record, and the node broadcasts the transaction record to the blockchain network according to the process of the PBFT consensus mechanism. The nodes in the network verify the transaction record, including data integrity verification and encryption compliance verification. When more than a statutory number of nodes, such as 2 / 3 of the normal nodes, reach a consensus, the transaction record is packaged into a new block and added to the blockchain, completing the storage of the power grid data in the distributed database.

[0059] Step 103, build a dynamic topology graph of the power grid, and use GNN network to encode the dynamic topology graph to capture the spatial dependency between devices;

[0060] In this embodiment, all kinds of devices in the power grid are comprehensively sorted out, including transformers, circuit breakers, transmission lines, busbars, etc. The unique identifier, model, rated parameter and other basic information of each device are obtained. At the same time, the physical connection relationship between devices is recorded in detail, and it is clear which devices are directly connected through transmission lines and which devices are indirectly associated through busbars. These data are collected from multiple source systems such as power grid dispatching system and device account database, and integrated and verified to ensure the accuracy and integrity of the data.

[0061] The power grid devices are taken as the nodes of the topology graph, and the physical connection between devices is taken as the edge of the topology graph. The graph data structure is used to create a dynamic topology graph of the power grid. The basic information attributes of the devices are assigned to each node, and the length, transmission capacity and other attributes of the connecting lines are assigned to each edge. According to the collected current running state of the power grid, the on-off state of each edge is determined, such as the edge corresponding to the normally running transmission line is in a connected state, and the edge corresponding to the line in maintenance or fault state is in a disconnected state. Thus, an initial dynamic topology graph reflecting the current running state of the power grid is constructed.

[0062] According to the characteristics and data size of the dynamic topology graph of the power grid, a GNN network is selected, the number of layers of the GNN network, the number of neurons of each layer, the activation function and other hyperparameters are determined, and the weight matrix and bias vector in the network are initialized. These parameters will be continuously optimized through learning in the subsequent training process to adapt to the feature extraction requirements of the power grid topology data; for each node in the dynamic topology graph of the power grid, the relevant operating data and attribute information of the node are extracted as node features, for example, for a transformer node, the voltage, current, load rate, oil temperature and other operating parameters are extracted; for a transmission line node, the transmission power, line loss and other data are extracted, the numerical data are normalized to make them distributed in a suitable interval, and some category data are encoded and converted, and finally the feature vector of each node is formed to provide input for the calculation of the GNN network;

[0063] In the selected GNN network, information propagation and node feature updating operations are performed, starting from each node of the topology graph, the node propagates its own feature information to its neighbor nodes along the edges according to the rules set by the network, each node receives information from neighbor nodes, aggregates the information with its own original features, obtains new intermediate feature representations through specific operations such as weighted summation, splicing, etc., and updates the node feature vector through activation function processing, so that the updated feature vector contains the relevant information of the node itself and its neighbor nodes, thereby capturing the preliminary spatial dependency relationship between devices;

[0064] The information propagation and node feature updating process of the fifth step is repeated, and iterative calculation is performed between multiple layers of the GNN network. After passing through each network layer, the node feature vector further integrates the information of neighbor nodes in a wider range. With the increase of the number of layers, the node feature can capture the dependency relationship between devices at a farther distance. After each layer of calculation is completed, the updated node feature is taken as the input of the next layer of network, until all the set network layers are passed through, and finally the node feature code containing rich spatial dependency relationship information is obtained, completing the coding process of the dynamic topology graph of the power grid.

[0065] In step 104, the time series data in the power grid data is extracted by using the Transformer architecture, the unstructured data in the power grid data is encoded by using the CNN network, and the comprehensive feature vector is obtained by fusing the outputs of the Transformer architecture and the CNN network.

[0066] In this embodiment, the time series data is divided into multiple sequence segments by using a time window, and each segment is taken as the input of the Transformer architecture. The Transformer architecture captures the dependency relationship between different positions in the time series in different representation subspaces through the multi-head attention mechanism, and extracts the long-term dependency features in the time series data.

[0067] In this embodiment, the local features and global features of the unstructured data are gradually extracted by using the convolutional layer and the pooling layer of the CNN network to obtain high-level semantic feature representation; the outputs of the Transformer architecture and the CNN network are spliced or fused through weighted summation, a gating mechanism is introduced, and the weights of the features are adaptively adjusted to obtain a comprehensive feature vector.

[0068] Step 105, inputting the comprehensive feature vector into the LSTM-GAN model to predict the power grid load and fault probability, combining the dynamic topology graph and the prediction result to generate a device fault risk index;

[0069] In this embodiment, the time series data is divided into multiple sequence segments by using a time window, and each segment is used as the input of the Transformer architecture; the Transformer architecture captures the dependency relationship between different positions of the time series in different representation subspaces through the multi-head attention mechanism, and extracts the long-term dependency features in the time series data.

[0070] In this embodiment, the local features and global features of the unstructured data are gradually extracted by using the convolutional layer and the pooling layer of the CNN network to obtain high-level semantic feature representation; the outputs of the Transformer architecture and the CNN network are spliced or fused through weighted summation, a gating mechanism is introduced, and the weights of the features are adaptively adjusted to obtain a comprehensive feature vector.

[0071] Step 106, triggering an early warning signal and generating a control strategy according to the device fault risk index and the real-time running state, and performing data collaboration of cross-regional power grids through federated learning.

[0072] In this embodiment, multi-level thresholds of the device fault risk index are set, such as mild risk, moderate risk and severe risk thresholds, the device fault risk index calculated in real time is compared with these thresholds, and the real-time running state of the device is combined, such as whether there is abnormal current fluctuation, temperature rise and the like, if the risk index reaches the mild risk threshold, the system automatically sends a prompt information to the mobile terminal of the operation and maintenance personnel, reminding to pay attention to the device state; when reaching the moderate risk threshold, in addition to sending the early warning information, the device position is highlighted on the power grid monitoring platform, and the preliminary fault cause analysis is attached; when reaching the severe risk threshold, the audible and visual alarm is triggered immediately, the relevant person in charge is notified, and the emergency control strategy is quickly generated, such as automatically adjusting the power dispatching scheme, transferring the load borne by the fault device to other healthy devices to avoid the expansion of the fault;

[0073] In the aspect of data collaboration of cross-regional power grids, the regional power grid centers serve as participants of federated learning, and conduct model training based on local power grid data without leaking local sensitive data. Only the model parameter update information obtained through training is uploaded to the central server in an encrypted manner. The central server aggregates the parameter update information uploaded by each region to generate new global model parameters, and then distributes the updated global model parameters to each region. After receiving the new model parameters, each region continues to train the model in combination with local data, and through multiple rounds of iterative training, the collaborative analysis and model optimization of cross-regional power grid data are realized, and the prediction and response capability of the entire power grid system to equipment fault risks is improved.

[0074] In this embodiment, a hybrid communication network is constructed by integrating power line carrier communication, 5G communication and optical fiber communication technology. For real-time data such as power grid fault information, 5G communication is preferred for fast transmission. For a large amount of conventional data, power line carrier communication and optical fiber communication are combined to ensure the stability and reliability of data transmission. At the same time, an adaptive data transmission protocol is designed to dynamically adjust the data transmission rate and priority according to the network bandwidth and data flow.

[0075] In this embodiment, a unified data standard and interface specification are formulated to standardize the data from different systems and devices. Through data mapping and conversion technology, different formats of data are converted into a unified format to realize seamless data connection. A data fusion model is established to analyze the collected power grid equipment operation data, user power consumption data, geographic information data, etc. to generate comprehensive and accurate power grid operation state information. For example, the location information of power grid equipment is combined with geographic information system data to intuitively display the distribution and operation state of power grid equipment.

[0076] Please refer to Figure 2 The second embodiment of the power grid data through method based on data analysis provided by the embodiment of the present application is schematically shown in the figure, and the method specifically includes the following steps:

[0077] Step 201: For the collected operation data, a state space model of Kalman filtering is established, and initial parameters are set;

[0078] Step 202: For the data collected by the sensor at each time point, the state estimation value at the current time is predicted according to the state estimation at the last time and the state transition matrix, and the error covariance at the current time is predicted by using the error covariance matrix and the process noise covariance matrix;

[0079] In step 203, the predicted state estimation value at the current time is compared with the actual observation value, the Kalman gain is calculated, the predicted state estimation value is corrected according to the Kalman gain, the optimal state estimation value at the current time is obtained, and the error covariance matrix is updated to obtain the denoised operation data.

[0080] In this embodiment, for the collected operation data, a state space model of Kalman filtering is established, the power grid equipment operation parameters are regarded as state variables, and the sensor collected data are regarded as observation variables. For each parameter to be processed, the state equation and the observation equation are defined respectively. The state equation describes the evolution law of the system state over time, and the observation equation represents the relationship between the observation data and the system state. At the same time, the process noise and the observation noise are set for the state equation and the observation equation respectively, for describing the system uncertainty and the measurement error; the initial parameters of the Kalman filtering algorithm are set, including the initial state estimation value, the initial error covariance matrix, the state transition matrix, the observation matrix, the process noise covariance matrix and the observation noise covariance matrix. The initial state estimation value can be set based on the historical data or the field knowledge, and the error covariance matrix reflects the uncertainty of the state estimation. The state transition matrix and the observation matrix are determined according to the system characteristics and the physical law, and the process noise covariance matrix and the observation noise covariance matrix are estimated according to the statistical characteristics of the noise. The prediction and updating process is repeated to process the entire time series data, and the denoised operation data is obtained.

[0081] Figure 3 A structure schematic diagram of a power grid data through equipment based on data analysis provided by the embodiment of the application, the power grid data through equipment based on data analysis 300 can have a large difference due to different configurations or performances, and can include one or more than one processor (central processing units, CPU) 310 (for example, one or more than one processor) and a memory 320, one or more than one storage medium 330 (for example, one or more than one mass storage device) storing an application program 333 or data 332. Wherein, the memory 320 and the storage medium 330 can be temporary storage or persistent storage. The program stored in the storage medium 330 can include one or more than one module (not shown in the figure), each module can include a series of instruction operations in the power grid data through equipment based on data analysis 300. Further, the processor 310 can be arranged to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the power grid data through equipment based on data analysis 300, so as to realize the method provided by the above embodiment.

[0082] The data analysis based power grid data through equipment 300 can also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating devices 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that, Figure 3 The data analysis based power grid data through equipment structure shown is not a limitation of the computer equipment provided by the present application, and can include more or fewer components than shown, or combine certain components, or different component arrangements.

[0083] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, or a volatile computer readable storage medium, and the computer readable storage medium stores instructions, which, when executed on a computer, cause the computer to perform the steps of the data analysis based power grid data through method provided by each embodiment.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described device or apparatus, unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0085] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0086] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for integrating power grid data based on data analysis, characterized in that, The method includes the following steps: Distributed sensor networks are used to collect operating data of power grid equipment. Kalman filtering is used to denoise the operating data, and the DTW algorithm is used to unify the timestamps to obtain power grid data. Establish a blockchain-based distributed database to classify and store power grid data according to different types and levels; Construct a dynamic topology map of the power grid, and use a GNN network to encode the dynamic topology map to capture the spatial dependencies between devices; The Transformer architecture is used to extract features from time-series data in power grid data, and the CNN network is used to encode features from unstructured data in power grid data. The outputs of the Transformer architecture and the CNN network are fused to obtain a comprehensive feature vector. The comprehensive feature vector is input into the LSTM-GAN model to predict the power grid load and fault probability. The dynamic topology map and prediction results are combined to generate the equipment fault risk index. Based on the equipment failure risk index and real-time operating status, early warning signals are triggered and control strategies are generated, enabling cross-regional power grid data collaboration through federated learning.

2. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The process of collecting operational data from power grid equipment using a distributed sensor network and then denoising the data using Kalman filtering includes: Based on the collected operational data, a state-space model of Kalman filtering is established, and initial parameters are set; For the data collected by the sensor at each time point, the state estimate at the current time is predicted based on the state estimate and state transition matrix of the previous time point. The error covariance matrix and the process noise covariance matrix are used to predict the error covariance at the current time point. The predicted state estimate at the current moment is compared with the actual observation, the Kalman gain is calculated, the predicted state estimate is corrected based on the Kalman gain, the optimal state estimate at the current moment is obtained, and the error covariance matrix is ​​updated to obtain the denoised running data.

3. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The process of unifying timestamps using the DTW algorithm to obtain power grid data includes: The time series distance matrix is ​​constructed from the running data after Kalman filtering and denoising, where the distance matrix stores the distance between each data point in the two time series; Initialize the cumulative distance matrix, calculate the cumulative distance value for each element in the cumulative distance matrix, and gradually fill the cumulative distance matrix starting from the starting point until all elements have been calculated. The cumulative distance matrix is ​​the same size as the time series distance matrix. Starting from the bottom right element of the cumulative distance matrix, backtrack to find the optimal alignment path. Based on the optimal alignment path, unify the timestamps of the two time series to finally obtain grid data with unified timestamps.

4. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The construction of the dynamic power grid topology map, which utilizes a GNN network to encode the dynamic topology map and capture the spatial dependencies between devices, includes: A dynamic power grid topology is constructed by using power grid equipment as nodes in the topology graph and the physical connections between equipment as edges. For each node in the dynamic topology graph, its corresponding running data and attribute information are extracted as node features, and then encoded and transformed to form the feature vector of each node. The feature vector of each node is input into the GNN network. Starting from each node in the dynamic topology graph, the node propagates its own feature information along the edge to its neighboring nodes. After receiving the information from the neighboring nodes, each node aggregates it with its original features to obtain a new intermediate feature representation. The intermediate features are processed by an activation function to update the feature vectors of the nodes. After multiple iterations through the GNN network, the spatial dependencies between devices are captured.

5. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The method of using the Transformer architecture to extract features from time-series data in power grid data includes: The time series data is divided into multiple sequence segments using time windows, and each segment serves as the input to the Transformer architecture; The Transformer architecture uses a multi-head attention mechanism to capture the dependencies between different positions in a time series within different representation subspaces, extracting long-term dependency features from the time series data.

6. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The process involves using a CNN network to encode features from unstructured data in the power grid data, and then fusing the outputs of the Transformer architecture and the CNN network to obtain a comprehensive feature vector, including: By using the convolutional and pooling layers of a CNN network to progressively extract local and global features from unstructured data, a high-level semantic feature representation is obtained. The outputs of the Transformer architecture and the CNN network are concatenated or fused through weighted summation, and a gating mechanism is introduced to adaptively adjust the weights of the features to obtain a comprehensive feature vector.

7. The power grid data integration method based on data analysis as described in claim 1, characterized in that, The process involves inputting a comprehensive feature vector into an LSTM-GAN model to predict power grid load and fault probability. Combining the dynamic topology graph and prediction results, an equipment fault risk index is generated, including: By inputting the comprehensive feature vector into the LSTM-GAN model, the LSTM network learns the dynamic changes of the time series through memory units and gating mechanisms. GAN networks consist of generators and discriminators, and the fitting of data distribution is handled through adversarial training of the generators and discriminators. The LSTM-GAN model generates predicted values ​​of future loads and, based on equipment operating data and comprehensive feature vectors, predicts the probability of equipment failure within a certain period of time. Combining the dynamic topology map and the prediction results, it generates an equipment failure risk index.

8. A system for implementing the power grid data interconnection method based on data analysis as described in claim 1, characterized in that, The system includes: The data acquisition module is used to collect operating data of power grid equipment using a distributed sensor network, denoise the operating data using Kalman filtering, and unify the timestamps using the DTW algorithm to obtain power grid data. The storage module is used to establish a blockchain-based distributed database to classify and store power grid data according to different types and levels. The encoding module is used to construct a dynamic topology map of the power grid. It uses a GNN network to encode the dynamic topology map and capture the spatial dependencies between devices. The fusion module is used to extract features from time-series data in power grid data using the Transformer architecture, encode features from unstructured data in power grid data using a CNN network, and fuse the outputs of the Transformer architecture and the CNN network to obtain a comprehensive feature vector. The prediction module is used to input the comprehensive feature vector into the LSTM-GAN model to predict the power grid load and fault probability, and generate the equipment fault risk index by combining the dynamic topology map and the prediction results. The data collaboration module is used to trigger early warning signals and generate control strategies based on the equipment failure risk index and real-time operating status, and to conduct cross-regional power grid data collaboration through federated learning.

9. A power grid data interconnection device based on data analysis, characterized in that, The power grid data interconnection device based on data analysis includes a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the power grid data interconnection device based on data analysis to perform each step of the power grid data interconnection method based on data analysis as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the power grid data interconnection method based on data analysis as described in any one of claims 1-7.

Images