Data processing method for network-provincial cooperative monitoring of power dispatching system

Abstract

The invention discloses a data processing method for network-provincial cooperative monitoring of a power dispatching system, and the method comprises the steps: obtaining a change trend of a data stream, and employing an incremental learning and forgetting mechanism to adaptively adjust a compression dictionary; for non-stationary power data, realizing self-adaptive compression by continuously updating the compression dictionary; constructing a multi-scale space-time compression model by adopting autocorrelation analysis, mutual information calculation, principal component analysis and a deep learning method; combining the multi-scale space-time compression model with a graph model to realize fusion representation of a topological structure and data features of the power system; fusing the graph structure features into predictive coding and differential coding in a compression process; a unified compression algorithm selection framework is constructed to realize high-adaptability compression of the complex power data; adopting a reinforcement learning method to autonomously learn and optimize parameter configuration of each compression algorithm; and adaptively adjusting a compression ratio threshold and a reconstruction quality target according to the precision tolerance of the power application.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power dispatching, in particular to a data processing method for coordinated monitoring of power dispatching system networks and provinces. Background Art

[0002] Data transmission in power systems faces unique technical challenges. Due to the wide geographical distribution of power systems, the large number of monitoring points, and the huge amount of data collected by various sensors, traditional data transmission methods are difficult to meet the requirements of real-time and reliability. At the same time, the data in power systems has obvious temporal and spatial correlation characteristics and contains a large amount of redundant information. Direct transmission of raw data not only takes up a lot of bandwidth resources, but also increases the complexity of data processing.

[0003] Therefore, there is an urgent need for a data compression algorithm tailored to the characteristics of the power system, which can minimize the amount of data transmitted while ensuring data accuracy. However, there are many types of data in the power system, and different types of data have different statistical characteristics and importance. It is difficult to take into account the compression effects of various data using a unified compression algorithm. In addition, the power system has extremely high requirements for the real-time performance of data, and the time complexity of the compression algorithm must be low enough. How to find the best balance between data compression rate, compression speed and reconstruction quality is a key technical problem that needs to be solved urgently.

[0004] This requires in-depth analysis of the data characteristics of the power system, the use of differentiated compression strategies for different types of data, and the introduction of domain knowledge into the compression algorithm to make full use of the temporal and spatial correlation of the data to improve compression efficiency. At the same time, it is also necessary to balance the computational complexity of the compression algorithm with the hardware implementation cost to achieve the best compression performance under limited resources. Summary of the invention

[0005] The purpose of the present invention is to solve the above problems and provide a data processing method for coordinated monitoring of power dispatching system networks and provinces, which fully utilizes the temporal and spatial correlation of data, improves compression efficiency, and can achieve optimal compression performance under limited resources.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] A data processing method for coordinated monitoring of a power dispatching system network and province, comprising the following steps:

[0008] a. Obtain the topological structure and equipment parameter information of the power system and construct a compressed dictionary containing typical patterns of voltage, current and power physical quantities;

[0009] b. Obtaining domain ontology and semantic model, constructing semantic association graph across physical quantities and devices and integrating it into the compression dictionary;

[0010] c. Obtain the change trend of the data stream and adaptively adjust the compression dictionary using incremental learning and forgetting mechanisms;

[0011] d. for non-stationary power data, adaptive compression is achieved by continuously updating the compression dictionary;

[0012] e. Using the semantic association graph to guide the selection of compression algorithms, achieving collaborative compression and semantic reconstruction of multi-source heterogeneous power data;

[0013] f. Use autocorrelation analysis, mutual information calculation, principal component analysis and deep learning methods to build a multi-scale space-time compression model;

[0014] g. Combining the multi-scale space-time compression model with the graph model to realize the fusion representation of the power system topology structure and data characteristics;

[0015] h. Incorporating graph structure features into the prediction coding and differential coding of the compression process;

[0016] iBuild a unified compression algorithm selection framework to achieve highly adaptable compression of complex power data;

[0017] jUse reinforcement learning method to autonomously learn and optimize the parameter configuration of each compression algorithm;

[0018] k Adaptively adjust the compression rate threshold and reconstruction quality target according to the accuracy tolerance of the power application;

[0019] l Obtain application feedback on compressed data and dynamically adjust compression rate thresholds and reconstruction quality targets based on service quality assessment results.

[0020] Furthermore, in step a, a compression dictionary containing typical patterns of the physical quantities of voltage, current, and power is constructed, including the following steps:

[0021] Obtain the power grid topology and equipment parameter information from the power dispatching system, and extract the physical quantity information of voltage, current and power;

[0022] Use the preset compression algorithm to build a compression dictionary;

[0023] If the compression dictionary contains a typical pattern, mapping the typical pattern to the network-province collaborative monitoring system;

[0024] If the compression dictionary does not contain a typical pattern, cluster analysis is performed on the physical quantity data using a preset machine learning algorithm to generate a new typical pattern and update the compression dictionary.

[0025] Furthermore, in step b, constructing a semantic association graph across physical quantities and devices and integrating it into the compression dictionary includes the following steps:

[0026] Obtain domain ontology and semantic model in power dispatching system;

[0027] If the domain ontology contains concepts related to power dispatching, a mapping relationship between physical quantities and equipment is extracted from the semantic model;

[0028] According to the mapping relationship between the physical quantity and the device, a semantic association graph across physical quantities and across devices is constructed;

[0029] The semantic association graph is integrated into a compression dictionary, and a machine learning algorithm is used to optimize the dictionary structure.

[0030] Furthermore, in step c, the compression dictionary is adaptively adjusted using an incremental learning and forgetting mechanism, including the following steps:

[0031] Obtain the data stream of the power dispatching system network and provincial collaborative monitoring;

[0032] Adopting an incremental learning method, extracting new typical features from the data stream and updating a preset compression dictionary;

[0033] Through the forgetting mechanism, it is determined whether the data pattern in the compression dictionary is outdated. If the data pattern is outdated, the outdated data pattern is removed from the compression dictionary.

[0034] Further, step d comprises the following steps:

[0035] Acquire non-stationary power data in a power dispatching system, analyze non-stationary characteristics of the non-stationary power data, and determine time series characteristics of the non-stationary power data;

[0036] Using a preset initial compression dictionary, preliminarily compressing the non-stationary power data to obtain a first compression result;

[0037] According to the data characteristics of the first compression result, the update requirement of the initial compression dictionary is determined, and if update is required, a new compression dictionary is generated.

[0038] Further, step e comprises the following steps:

[0039] Acquire multi-source heterogeneous power data in the power dispatching system, and perform classification preprocessing based on the type and semantic features of the power data;

[0040] Searching for nodes matching the features of the power data in a pre-built semantic association graph, and determining the position and association relationship of the power data in the semantic association graph;

[0041] According to the relevance of the power data in the semantic association graph, a suitable compression algorithm is selected.

[0042] Further, step f comprises the following steps:

[0043] Obtain the time series of the coordinated monitoring data of the power dispatching system network and province, calculate the time scale characteristics of the time series by using the autocorrelation analysis method, and obtain the data fluctuation law under different time scales;

[0044] For the monitoring data of different spatial granularities, a mutual information calculation method is used to perform correlation calculation to obtain the correlation strength in the spatial dimension;

[0045] According to the time scale feature and the spatial correlation feature, a principal component analysis method is used to perform feature dimension reduction to obtain a key feature vector after dimension reduction;

[0046] If the correlation between the time scale feature and the spatial correlation feature is greater than a preset threshold, a weighted fusion method is used to perform feature integration on the key feature vector to obtain an integrated feature vector; otherwise, the key feature vector is retained;

[0047] The integrated feature vector is used as input, and a multi-scale spatiotemporal compression model is obtained through deep learning model training to obtain a time-space compression representation of the monitoring data.

[0048] Further, step h comprises the following steps:

[0049] Obtaining original data, extracting graph structure features of the original data, and determining associations between data nodes;

[0050] For the differential data, determining whether the dynamic range of the differential data exceeds a preset threshold, and if it exceeds the preset threshold, adjusting the dynamic range to obtain adjusted differential data;

[0051] Using a target compression algorithm, compressing the adjusted differential data to generate a compressed data block;

[0052] The compressed data blocks are integrated with the graph structure features to obtain optimized data.

[0053] Further, step k comprises the following steps:

[0054] Obtaining the original data of the power dispatching system network-province collaborative monitoring;

[0055] According to the data size and accuracy tolerance of the original data, a preset algorithm is used to determine an initial compression rate threshold;

[0056] If the data size of the original data exceeds a preset threshold, the reconstruction quality target is adjusted, and an adaptive optimization algorithm is used to dynamically calculate the compression rate threshold;

[0057] Obtaining compressed monitoring data according to the reconstruction quality target and the compression rate threshold;

[0058] The compressed monitoring data is analyzed using a preset performance evaluation model to determine whether a balance between compression performance and resource overhead is achieved;

[0059] If the equilibrium state does not meet the preset standard, the compression rate threshold and the reconstruction quality target are readjusted to iteratively optimize the compression process;

[0060] The optimized compressed data is applied to the power dispatching system network-province collaborative monitoring to obtain the final monitoring result.

[0061] The beneficial effects of the present invention are:

[0062] 1. The present invention realizes a unified representation of the topological structure of the power system and the characteristics of multi-source heterogeneous data by constructing a compression dictionary that integrates semantic association graphs. The present invention utilizes incremental learning and forgetting mechanisms to continuously update the compression dictionary to adapt to the changing trend of non-stationary data. Through the combination of a multi-scale spatiotemporal compression model and a graph model, the present invention realizes the fusion representation of the topology of the power system and the data characteristics, and integrates the graph structure characteristics into the compression process, thereby improving the compression efficiency. The present invention also adopts a reinforcement learning method to autonomously optimize the parameter configuration of the compression algorithm to achieve global optimization of the compression performance. In addition, the present invention dynamically adjusts the compression strategy according to application feedback, so that the compression process is coordinated with business needs. This method significantly improves the adaptability and efficiency of power data compression, and provides strong support for the efficient operation and analysis of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] Figure 1 It is a flow chart of data processing for coordinated monitoring of power dispatching system network and province of the present invention. DETAILED DESCRIPTION

[0065] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0066] like Figure 1 As shown, a data processing method for coordinated monitoring of power dispatching system network and province includes the following steps:

[0067] a. Obtain the topological structure and equipment parameter information of the power system and construct a compressed dictionary containing typical patterns of voltage, current and power physical quantities;

[0068] b. Obtaining domain ontology and semantic model, constructing semantic association graph across physical quantities and devices and integrating it into the compression dictionary;

[0069] c. Obtain the change trend of the data stream and adaptively adjust the compression dictionary using incremental learning and forgetting mechanisms;

[0070] d. for non-stationary power data, adaptive compression is achieved by continuously updating the compression dictionary;

[0071] e. Using the semantic association graph to guide the selection of compression algorithms, achieving collaborative compression and semantic reconstruction of multi-source heterogeneous power data;

[0072] f. Use autocorrelation analysis, mutual information calculation, principal component analysis and deep learning methods to build a multi-scale space-time compression model;

[0073] g. Combining the multi-scale space-time compression model with the graph model to realize the fusion representation of the power system topology structure and data characteristics;

[0074] h. Incorporating graph structure features into the prediction coding and differential coding of the compression process;

[0075] iBuild a unified compression algorithm selection framework to achieve highly adaptable compression of complex power data;

[0076] jUse reinforcement learning method to autonomously learn and optimize the parameter configuration of each compression algorithm;

[0077] k Adaptively adjust the compression rate threshold and reconstruction quality target according to the accuracy tolerance of the power application;

[0078] l Obtain application feedback on compressed data and dynamically adjust compression rate thresholds and reconstruction quality targets based on service quality assessment results.

[0079] Step a includes the following steps: obtaining the power grid topology and equipment parameter information from the power dispatching system, extracting the physical quantity information of voltage, current and power; using a preset compression algorithm to construct a compression dictionary; if the compression dictionary contains a typical pattern, mapping the typical pattern to the network-province collaborative monitoring system; if the compression dictionary does not contain a typical pattern, clustering analysis is performed on the physical quantity data through a preset machine learning algorithm to generate a new typical pattern and update the compression dictionary. The power dispatching data is compressed in real time according to the updated compression dictionary to obtain compressed power data. The compressed data is input into the network-province collaborative monitoring system for real-time analysis and display.

[0080] The grid topology is an important carrier that reflects the structure of the power system, which includes the connection relationship between substations, transmission lines and other equipment. The voltage values ​​obtained from the dispatching system include the main bus voltage of the provincial substation is 500 kV, the branch line voltage is 220 kV; the current value is 2,000A for the main transformer load current; the power value is 400 MW for active power and 100 MW for reactive power.

[0081] The compression dictionary uses key-value pairs to store typical patterns. For example, in normal operation mode, the load rate of the main transformer is maintained at 60% to 70%, and the voltage qualification rate is higher than 98%.

[0082] Typical patterns include equipment operation characteristics, load change patterns, etc. When it is detected that the operation data of a substation conforms to the typical pattern stored in the dictionary, it can be directly mapped to the network-province collaborative monitoring system to achieve rapid data transmission. If new operating characteristics appear, such as a sudden increase in the load rate of a substation to 85% under special weather conditions, cluster analysis is required through machine learning algorithms. Using the fuzzy clustering method, data with high similarity are classified into one category, such as classifying the operating conditions with a load rate of 80% to 90% as heavy load operation mode, and updating the mode to the compression dictionary.

[0083] During real-time compression processing, the original data is subjected to dimensionality reduction and feature extraction. For example, voltage data sampled for five minutes is compressed into a fifteen-minute average, or the main characteristic values ​​of current fluctuations are extracted. This can significantly reduce the amount of data transmission while ensuring data accuracy.

[0084] The network-province collaborative monitoring system displays compressed data through a visual interface. For example, a curve chart is used to show the load change trend of the equipment, and a heat map is used to reflect the weak links of the power grid. The system can also perform intelligent analysis to detect abnormal working conditions in a timely manner. For example, when it is detected that the load rate of the main transformer of a substation exceeds 90% continuously, the system automatically issues an early warning message to prompt the dispatcher to pay attention to the power supply situation in the area. By establishing a compression dictionary and real-time data compression, the cost of data transmission and storage can be effectively reduced. At the same time, the introduction of machine learning algorithms enables the system to have adaptive capabilities and continuously optimize the compression effect. The network-province collaborative monitoring system realizes the intuitive display and intelligent analysis of data, providing strong support for the safe and stable operation of the power grid.

[0085] Step b includes the following steps: obtaining the domain ontology and semantic model in the power dispatching system; if the domain ontology contains concepts related to power dispatching, extracting the mapping relationship between physical quantities and equipment from the semantic model; constructing a semantic association map across physical quantities and across equipment based on the mapping relationship between physical quantities and equipment; integrating the semantic association map into a compression dictionary, and optimizing the dictionary structure using a machine learning algorithm. The optimized compression dictionary is used to process the network-province collaborative monitoring data of the power dispatching system to obtain a processing result with high expressiveness. Based on the processing results, the monitoring requirements are analyzed to determine whether the semantic model needs to be further optimized. If the monitoring requirements change, the semantic association map is readjusted, the compression dictionary is updated, and an adaptively optimized dictionary is obtained.

[0086] The domain ontology in the power dispatching system contains concepts such as various types of equipment, physical quantities, and operating status. For example, transformers have attributes such as rated capacity and voltage level, and transmission lines have characteristics such as impedance and conductor type. Based on these domain ontology concepts, a mapping between physical quantities and equipment can be established through semantic models. For example, there is a transformation ratio relationship between the high-voltage side voltage and the low-voltage side voltage of the main transformer of a substation, and there is a loss relationship between the first-end power and the terminal power of a transmission line.

[0087] When constructing a semantic association graph, the relationships between different physical quantities can be connected. For example, the product of voltage and current of a line reflects the transmission power, and the power change trend can reflect the load characteristics. For cross-device association, the operating status of primary equipment such as busbars and switches in the substation can be associated with the action logic of secondary equipment such as protection devices.

[0088] In the process of optimizing the compression dictionary, the machine learning algorithm can identify typical operating modes, such as the peak and valley characteristics of the load curve, the equipment start and stop sequence, etc. Taking the substation as an example, the temperature of the main transformer is positively correlated with the load rate, and this regular feature can be included in the dictionary. Cluster analysis can be used to discover the coordinated change rules of the equipment group, such as the synchronous fluctuation characteristics of the voltage of multiple substations in the region. In practical applications, the key points that the network and provincial collaborative monitoring system needs to pay attention to include equipment status monitoring, power flow distribution analysis, etc. Taking line fault analysis as an example, the association relationship between devices contained in the compression dictionary helps to quickly locate the fault point and predict the scope of the fault. When new equipment appears in the system or the topology changes, the semantic model needs to be adjusted accordingly. For example, the connection relationship of related equipment needs to be updated after the new line is put into operation.

[0089] In the dynamic adjustment process of the semantic association graph, factors such as equipment maintenance and seasonal load changes need to be considered. For example, during equipment overhaul, the association relationship of related monitoring data will change temporarily. The adaptive optimization of the compression dictionary can reflect these dynamic changes and ensure the accuracy of data compression and analysis. Through the continuous optimization of the dictionary structure, the monitoring system can improve its ability to identify abnormal power grid conditions and provide effective support for dispatching operations.

[0090] Step c includes the following steps: obtaining the power dispatching system network-province collaborative monitoring data stream; using an incremental learning method to extract new typical features from the data stream and update the preset compression dictionary; judging whether the data pattern in the compression dictionary is outdated through a forgetting mechanism, and if the data pattern is outdated, removing the outdated data pattern from the compression dictionary. Adaptively adjust the content of the compression dictionary according to the changing trend of the data stream; adaptively adjust the storage space and update frequency of the compression dictionary according to the scale of the compression dictionary; obtain the updated compression dictionary, and use the updated compression dictionary for processing the power dispatching system network-province collaborative monitoring data; judging whether the content and scale of the compression dictionary need to be further adjusted according to the processing results of the power dispatching system network-province collaborative monitoring data, if further adjustment is required, return to the step of adaptively adjusting the content of the compression dictionary, otherwise, output the processing results of the power dispatching system network-province collaborative monitoring data.

[0091] The data stream of the provincial coordinated monitoring of the power dispatching system network contains real-time data such as the operating status of generator sets, transmission line loads, and operating parameters of substation equipment.

[0092] When using the incremental learning method, the system can identify new typical features from the data stream, such as the vibration characteristics of a generator set under specific working conditions, the overload characteristics of a transmission line under severe weather conditions, etc. After these new features are extracted, they are used to update the preset compression dictionary so that the dictionary has the ability to express new data patterns.

[0093] In the forgetting mechanism, the system evaluates the timeliness of each data pattern in the compression dictionary. After the substation is transformed, the original equipment operation characteristic pattern is no longer applicable, and the system will automatically identify and remove these outdated patterns. This mechanism ensures that the dictionary always maintains the ability to accurately describe the current system status.

[0094] The changing trend of data streams may be manifested as seasonal changes in load curves, attenuation of equipment performance over time, etc. The system dynamically adjusts the compression dictionary accordingly, such as increasing the description accuracy of load characteristics during peak power consumption periods and refining fault feature patterns during equipment aging stages.

[0095] In view of the size of the compression dictionary, the system performs hierarchical storage according to the importance of the data. Key equipment data such as the main transformer operating parameters adopt a high-frequency update strategy, while secondary data such as ordinary line status adopts a lower update frequency. The storage space is also adjusted accordingly, and important data takes up more space to ensure accuracy.

[0096] When the updated compression dictionary is applied to actual data processing, abnormal conditions can be quickly identified. The system continuously evaluates the accuracy and timeliness of processing results. If the recognition rate of a certain type of fault mode is found to decrease, or the characteristics of the newly added equipment cannot be accurately expressed, the dictionary adjustment process is initiated. Adjustments may include adding new feature patterns, optimizing feature extraction methods, reallocating storage resources, etc., until the expected processing effect is achieved. This continuous optimization mechanism ensures that the system always maintains a high level of monitoring efficiency.

[0097] Step d includes the following steps: obtaining non-stationary power data in the power dispatching system, analyzing the non-stationary characteristics of the non-stationary power data, and determining the time series characteristics of the non-stationary power data; using a preset initial compression dictionary to preliminarily compress the non-stationary power data to obtain a first compression result; judging the update requirements of the initial compression dictionary based on the data characteristics of the first compression result, and generating a new compression dictionary if an update is required; applying the new compression dictionary to the compression of subsequent non-stationary power data to obtain a second compression result; inputting the second compression result into the network-province collaborative monitoring system for collaborative processing to generate monitoring data; determining the optimization requirements of the new compression dictionary based on the processing results of the monitoring data to obtain an optimized compression dictionary; and using the optimized compression dictionary to adaptively compress new non-stationary power data.

[0098] The non-stationary power data in the power dispatching system has the dynamic characteristics of time series, which is manifested as the volatility and seasonal changes of the load curve. This non-stationarity is reflected in the change of the mean and variance of the data over time.

[0099] The initial compression dictionary can use a codebook built based on the statistical characteristics of historical data, including common load patterns and typical fluctuation characteristics. For example, the voltage fluctuation range and frequency distribution law of a substation in a certain area at different times can be used as the basic elements of the dictionary. The initial compression of data is achieved by matching and encoding the voltage data collected in real time.

[0100] The need to update the compression dictionary is usually triggered by changes in compression effects and data features. For example, when new electricity usage patterns differ greatly from existing patterns in the dictionary, these new features need to be included in the dictionary.

[0101] When the network-province collaborative monitoring system processes compressed data, it needs to consider the real-time and accuracy of the data. When a sudden power change is detected on a transmission line, the system can quickly identify and locate the anomaly, which requires the compression process to retain enough detailed information.

[0102] The optimized compression dictionary needs to balance compression efficiency and data fidelity. For example, in the power grid dispatching automation system, for the sampling data of the protection device of the high-voltage transmission line, differentiated compression strategies can be adopted according to the importance of different time periods. A higher compression ratio can be used during normal operation, while the compression ratio can be reduced to retain more details when disturbances occur.

[0103] During the adaptive compression process, the dynamic adjustment mechanism of the dictionary should take into account the periodic characteristics of the power system operation. For example, for the output characteristics of photovoltaic power generation, more detailed data sampling and compression strategies are adopted during sunrise and sunset, while the compression accuracy can be appropriately reduced at night. This adaptive mechanism can optimize the use efficiency of storage and transmission resources while ensuring data quality. At the same time, for the key parameters of important nodes such as ultra-high voltage substations, the compression process must ensure rapid response capabilities and be able to adjust the compression strategy in time when emergencies occur.

[0104] Step e includes the following steps: acquiring multi-source heterogeneous power data in the power dispatching system, and performing classification preprocessing according to the type and semantic features of the power data; searching for nodes matching the features of the power data in a pre-constructed semantic association graph, and determining the position and association relationship of the power data in the semantic association graph; selecting a suitable compression algorithm according to the association of the power data in the semantic association graph, wherein if the power data has a high association, a dictionary-based compression algorithm is adopted, and if the power data has a low association, a transformation-based compression algorithm is adopted; using the selected compression algorithm to perform collaborative compression processing on the multi-source heterogeneous power data to obtain a compressed data set; searching for semantic nodes corresponding to the compressed data in the semantic association graph to obtain the semantic information required for data reconstruction; reconstructing the compressed data according to the semantic information to generate a reconstructed data set with semantic consistency; transmitting the reconstructed data set to the network-province collaborative monitoring system to complete the data processing flow.

[0105] The multi-source heterogeneous data in the power dispatching system include equipment status data, load data, voltage data, etc. Taking the transformer as an example, its status data involves parameters such as temperature, oil level, and air pressure. These data have different collection cycles and value ranges and need to be standardized and semantically labeled respectively.

[0106] When constructing the semantic association graph, the transformer is used as the core node to connect its various monitoring data nodes. For example, if there is a strong correlation between the transformer oil temperature and the load rate, the two nodes can be closely associated. However, the transformer alarm information has a low correlation with attributes such as geographic location, and the connection strength between these nodes is relatively weak. For highly correlated data such as transformer oil temperature and load rate, it is more appropriate to use a dictionary-based compression algorithm. A compression dictionary containing data features under common working conditions can be established, such as the correspondence between oil temperature and load rate under overload conditions. For low-correlation data such as the geographic location information of the transformer, a compression method based on wavelet transform is used to extract the main features for compression.

[0107] In the collaborative compression process, the same compression dictionary is used for highly correlated data groups to ensure that the correlation characteristics between the data are preserved. For example, when the transformer oil temperature and load rate data are compressed at the same time, a joint dictionary that reflects their correlation is used to improve the compression efficiency and maintain the correlation of the data.

[0108] During the semantic reconstruction process, the semantic information stored in the graph is used to guide data reconstruction. For example, when reconstructing transformer operating data, the operating conditions of the period are known through the graph, and the corresponding reconstruction parameters are selected. If it is in a heavy load condition, joint reconstruction is performed according to the corresponding relationship between the load rate and the oil temperature to ensure that the reconstructed data reflects the true equipment status. The data set finally transmitted to the network-province collaborative monitoring system has complete semantic information. For example, the reconstructed data of a transformer not only contains specific measurement values, but also contains semantic annotations such as equipment operating status and potential risks, which facilitates unified monitoring and analysis by the system. This semantic-based collaborative compression method can effectively reduce the amount of data transmission, while ensuring the comprehensibility and availability of the data, and improving the efficiency of power monitoring.

[0109] Step f includes the following steps: obtaining the time series of the coordinated monitoring data of the power dispatching system network and province, using the autocorrelation analysis method to calculate the time scale characteristics of the time series, and obtaining the data fluctuation law at different time scales; for the monitoring data of different spatial granularities, using the mutual information calculation method to calculate the correlation, and obtaining the correlation strength in the spatial dimension; according to the time scale characteristics and the spatial correlation characteristics, using the principal component analysis method to reduce the dimension of the features, and obtaining the key feature vector after the dimension reduction; if the correlation between the time scale characteristics and the spatial correlation characteristics is greater than the preset threshold, using the weighted fusion method to integrate the key feature vectors to obtain the integrated feature vectors, otherwise, retaining the key feature vectors; constructing a deep learning model, taking the integrated feature vectors as input, training a multi-scale space-time compression model, and obtaining the time-space compression representation of the monitoring data; according to the time-space compression representation, determining the time-space distribution law of the coordinated monitoring data of the power dispatching system network and province; based on the multi-scale space-time compression model, generating a multi-scale space-time feature map of the coordinated monitoring data of the power dispatching system network and province.

[0110] In the power dispatching system, the time series feature analysis of the network-province collaborative monitoring data is crucial. Taking the substation operation data as an example, through the autocorrelation analysis of the time series of the voltage level, it is found that it has obvious hourly scale fluctuations and daily scale periodicity. When the voltage deviation exceeds the standard range, the system will produce significant time scale characteristics, such as characteristic fluctuations during the daily peak and trough periods. In the spatial dimension, the correlation strength between different substations can be found through mutual information calculation. For example, substation equipment in adjacent areas often shows strong correlation. When the load of a substation changes significantly, the operating parameters of the adjacent substation will also fluctuate accordingly. This spatial correlation is particularly evident in densely populated urban areas.

[0111] For dimensionality reduction analysis, the voltage, current and other parameters of multiple monitoring points can be combined into high-dimensional feature vectors. Through principal component analysis, the original data of hundreds of dimensions can be reduced to key features of tens of dimensions, while retaining the main information of the data. For example, if there are fifty monitoring points in a certain area, each point has ten monitoring parameters, and through dimensionality reduction, twenty key features that can reflect the overall operating status can be obtained.

[0112] In the feature integration stage, when the correlation between time and space features exceeds the preset threshold, weighted fusion is required. For example, in power grid fault analysis, a sudden change in the current of a transmission line will cause a chain reaction in adjacent lines. At this time, the correlation between time and space features is significantly enhanced, and the features of these two dimensions need to be fused to more accurately characterize the fault propagation law.

[0113] The deep learning model can be constructed using a multi-layer convolutional neural network structure. The input layer receives the integrated feature vector, extracts spatiotemporal features through multiple convolutional layers and pooling layers, and finally outputs a compressed representation. This compressed representation can effectively capture key modes of power grid operation, such as load transfer paths, voltage fluctuation propagation, and other phenomena.

[0114] In the analysis of spatiotemporal distribution patterns, typical spatiotemporal coupling patterns can be identified. For example, in commercial areas, the power load during working days shows obvious spatial aggregation and temporal periodicity, which can guide the dispatching and operation of the power grid. Through the multi-scale spatiotemporal compression model, characteristic maps at different spatiotemporal scales can be generated, effectively supporting the analysis and decision-making of safe and stable operation of the power grid.

[0115] Step g includes the following steps: obtaining the original data of the power system, using a multi-scale space-time compression model to perform dimensionality reduction processing on the original data to obtain space-time feature data after dimensionality reduction; combining the space-time feature data with a preset power system topology structure to construct a power system graph model; for the power system graph model, using a graph matching algorithm to identify similar substructures to obtain at least one key area; based on the key areas, using a graph segmentation algorithm to divide the similar substructures into several data-related areas; extracting data features in each of the data-related areas to obtain a regional feature set; if there is abnormal data in the regional feature set, using a collaborative monitoring algorithm to perform anomaly detection to obtain anomaly detection results; generating a power system monitoring report based on the anomaly detection results and the operating status of the power system.

[0116] The original data of the power system is processed by multi-scale space-time compression model to reduce the dimension, which can effectively extract key features. The time series is decomposed at multiple scales by wavelet transform and other methods, and then combined with principal component analysis to reduce the dimension, the original data can be reduced to about 20% of the original dimension, while retaining key feature information.

[0117] When combining the reduced spatiotemporal feature data with the power system topology to construct a power system graph model, nodes can be used to represent substations and lines to represent transmission channels. The node attributes include static information such as voltage level and equipment type, as well as dynamic operation data such as voltage and current.

[0118] When using the graph matching algorithm to identify similar substructures, matching can be performed based on node attributes and topological features. For example, it can identify areas with similar power supply ranges and similar load characteristics, or discover line groups with similar voltage fluctuation characteristics.

[0119] Similar substructures are divided into data-related regions through graph segmentation algorithms, which can be based on indicators such as electrical distance and data correlation.

[0120] When extracting regional features, we focus on key indicators such as voltage stability and power transmission. In the anomaly detection phase, the collaborative monitoring algorithm comprehensively considers the data of each monitoring point in the region.

[0121] The monitoring report generated finally includes the description of abnormal events, analysis of the impact range and disposal suggestions, which can provide support for the dispatcher's decision-making. If a sign of cascading failure is found in a data-related area, it can provide timely warning and recommend preventive measures such as load shedding.

[0122] Step h includes the following steps: obtaining the original data in the power dispatching system network-province collaborative monitoring scenario, extracting the graph structure features of the original data, and determining the correlation relationship between the data nodes in the original data; establishing a prediction model based on the graph structure features, and generating a prediction value for the correlation relationship between the data nodes; using differential coding technology to calculate the difference between the original data and the prediction value to obtain differential data; for the differential data, analyzing the dynamic range of the differential data to determine whether the dynamic range exceeds a preset threshold, and if it exceeds the preset threshold, adjusting the dynamic range to obtain adjusted differential data; based on the adjusted differential data, calculating the entropy value of the adjusted differential data, and selecting a target compression algorithm in combination with a preset compression rate target; using the target compression algorithm to compress the adjusted differential data to generate a compressed data block; integrating the compressed data block with the graph structure features to generate optimized data.

[0123] The scenario of coordinated monitoring of power dispatching system network and province involves the collection and analysis of multi-level data. The original data includes substation operation data, line load data, equipment status information, etc.

[0124] The structural characteristics of the graph are reflected in the topological relationship between data nodes, such as the transmission line connection between substations, the connection relationship between the main transformer and the busbar, etc. The association relationship between data nodes can be determined through correlation analysis. For example, there is a significant correlation between the main transformer load of a substation and the power of the adjacent line.

[0125] In terms of forecasting model establishment, a time series forecasting model can be constructed based on the characteristics of historical data. Taking substation load forecasting as an example, by analyzing the periodic characteristics of the load curve and combining factors such as weather and holidays, the future load change trend can be predicted.

[0126] Differential coding technology can effectively reduce data redundancy by calculating the difference between the actual value and the predicted value. For example, the actual load of a substation is 120MW, the predicted value is 118MW, and the difference is 2MW.

[0127] Dynamic range analysis focuses on the fluctuation range of differential data. Taking voltage data as an example, if the dynamic range of differential data exceeds the preset threshold (such as ±5%), the range needs to be adjusted. Adjustment methods include interval mapping, normalization, etc., to make the data distribution more concentrated. The entropy value of differential data reflects the uncertainty of the data. A lower entropy value indicates that the data has better compressibility.

[0128] The selection of compression algorithm needs to consider data characteristics and compression targets. For differential data with small fluctuations, run-length encoding can be used. For example, if the switch state of a line remains unchanged for a period of time, compression can be achieved by recording the state value and duration. For data with large fluctuations, dictionary-based compression methods can be used, such as pattern matching and encoding of multiple similar load curves.

[0129] The integration of data blocks and graph structure features reflects the spatial correlation of data. Taking the regional power grid as an example, the data of adjacent substations are often similar, and these data blocks can be organized into a hierarchical structure. This integration optimization not only retains the temporal characteristics of the data, but also reflects the spatial topological relationship, which is helpful for subsequent data analysis and application.

[0130] The decompression and restoration process of compressed data needs to be combined with graph structure characteristics to ensure the integrity and spatiotemporal relevance of the data. Data processed in this way not only saves storage space but also maintains key information, providing strong support for the efficient operation of the power dispatching system.

[0131] Step i includes the following steps: obtaining the original data set of the power dispatching system network and provincial collaborative monitoring data, performing feature extraction on the original data set to obtain data feature description information; analyzing the data transmission path and node distribution according to the network topology structure, and determining the key nodes and link characteristics of the data transmission; judging the compression requirements of the current data set and determining the compression target according to the data feature description information and the key nodes and link characteristics; using a preset compression algorithm library, preliminarily screening out a candidate compression algorithm set according to the compression target and the data feature description information; dynamically analyzing the changes in the data feature description information, determining the triggering conditions for algorithm switching, and if the data feature description information changes, selecting the optimal algorithm from the candidate compression algorithm set; adjusting the parameter configuration of the optimal algorithm according to the key nodes and link characteristics and the compression requirements to obtain the optimized compression algorithm; applying the optimized compression algorithm to the original data set to generate a compressed data set.

[0132] The original data of the coordinated monitoring of the power dispatching system network and the province include voltage, current, active power and other measurement quantities. These data can be used to obtain descriptive information such as data distribution characteristics and change trends through feature extraction.

[0133] In network topology analysis, key nodes usually include major substations, important distribution stations, etc., and link characteristics are reflected in indicators such as data transmission bandwidth and latency.

[0134] The compression demand analysis needs to consider data characteristics and network conditions. For example, for key measurement data with high real-time requirements, the compression rate target can be set to 70%, and the allowed compression delay is no more than 0.1% of a second. For archiving historical data, a higher compression rate can be used, such as more than 90%.

[0135] The preset compression algorithm library includes two categories: lossless compression and lossy compression. The appropriate algorithm is selected according to different data characteristics. For example, for fault recording data with strong mutation, lossless compression is used to ensure data integrity. For continuously changing analog measurement values, lossy compression can be used to improve compression efficiency.

[0136] The triggering condition for algorithm switching is related to the change of data characteristics. For example, if the measured data of a substation fluctuates violently, it means that a fault may have occurred and it is necessary to switch to a lossless compression algorithm. During normal operation, the data changes steadily and a lossy compression algorithm can be used.

[0137] The tuning of compression algorithm parameters needs to be combined with specific application scenarios. For example, for important protection measurement data, the compression error limit should be less than one thousandth. For ordinary telemetry data, the limit can be appropriately relaxed to one percent. Parameter configuration also needs to consider network bandwidth limitations and balance compression rate and data quality under limited bandwidth.

[0138] In the process of generating compressed data sets, the time sequence and relevance of the data must be ensured. For example, the data of different measuring points in the same substation are packaged and compressed for subsequent analysis. The quality of the compression results is evaluated to ensure that they meet business requirements. This method ensures data compression efficiency while maintaining data availability.

[0139] Step k includes the following steps: obtaining data for coordinated monitoring of the power dispatching system network and province, the data having preset characteristics; initializing the parameter configuration of the compression algorithm according to the preset characteristics; using a reinforcement learning algorithm to explore in the parameter configuration space to obtain the compression rate, compression time and reconstruction quality evaluation results under the current parameter configuration; if the evaluation result under the current parameter configuration is better than the preset threshold, the current parameter configuration is used as a candidate optimal solution; if the evaluation result under the current parameter configuration does not reach the preset threshold, the parameter configuration is adjusted and the exploration is continued until at least one candidate optimal solution is obtained; according to the candidate optimal solution set, a multi-objective optimization algorithm is used for screening to determine the optimal parameter configuration scheme in terms of compression rate, compression time and reconstruction quality; applying the optimal parameter configuration scheme to the data processing flow for coordinated monitoring of the power dispatching system network and province, and monitoring the operating status of the compression algorithm in real time; for performance fluctuations occurring during operation, an adaptive adjustment mechanism is used to dynamically optimize the parameter configuration to maintain the global optimal performance of the compression algorithm; a parameter configuration optimization record library is established to store historical optimal parameter configuration schemes and their corresponding performance evaluation results to provide data support for subsequent algorithm optimization.

[0140] The coordinated monitoring data of the power dispatching system usually contains multiple measurement values ​​such as voltage, current, active power, etc. These data have different characteristics, such as fluctuation range, sampling frequency and data accuracy. According to these preset characteristics, the initial parameters of the compression algorithm can be set to different configurations.

[0141] During the reinforcement learning process, the algorithm will explore different parameter combinations. For example, for the voltage data of a substation, it may try multiple compression ratios from 2:1 to 10:1, while adjusting the data block size from 4KB to 64KB. Through this exploration, the system can find that when the compression ratio is 6:1 and the data block size is 16KB, it can achieve a higher compression rate while ensuring that the data reconstruction quality is within the allowable error range.

[0142] When evaluating candidate optimal solutions, the system will consider multiple performance indicators at the same time. Assuming that a set of parameter configurations achieves a compression rate of 85%, a compression time of less than 10 milliseconds, and a reconstructed data error of no more than 0.1%, such a configuration may be identified as a candidate optimal solution. If it is found that the compression time exceeds the threshold of 50 milliseconds, even if the compression rate is very high, it will be rejected by the system and trigger parameter adjustments.

[0143] In actual operation, the system continuously monitors the performance of the compression algorithm. When it is found that the power data of a substation fluctuates violently, the adaptive adjustment mechanism will automatically reduce the compression ratio to ensure the accurate transmission of important data. For example, when the power fluctuation exceeds 20% of the rated value, the system will adjust the compression ratio from 8:1 to 4:1 and increase the number of sampling points. The management of historical optimization records is also an important part of the system. By recording the optimal parameter configuration under different scenarios, the system can respond quickly to similar situations. Based on the newly emerging data features, the system will create a new parameter configuration scheme and store it in the record library, gradually expanding the optimization experience. This method can not only ensure the stability of the algorithm performance, but also adapt to the dynamic changes in the operation status of the power grid.

[0144] Step k includes the following steps: obtaining the original data of the power dispatching system network-province collaborative monitoring; using a preset algorithm to calculate the initial compression rate threshold according to the data size and accuracy tolerance of the original data; if the data size of the original data exceeds the preset threshold, adjusting the reconstruction quality target, and using an adaptive optimization algorithm to dynamically calculate the compression rate threshold; generating compressed monitoring data according to the reconstruction quality target and the compression rate threshold; using a preset performance evaluation model to analyze the compressed monitoring data to determine whether the compression performance and resource overhead have reached a balance; if the balance does not meet the preset standard, readjusting the compression rate threshold and the reconstruction quality target, and iteratively optimizing the compression process; applying the optimized compressed data to the power dispatching system network-province collaborative monitoring to obtain the final monitoring result.

[0145] The coordinated monitoring of power dispatching systems usually involves large-scale real-time data collection and transmission. The raw data mainly includes key parameters such as voltage, current, active power, reactive power, etc. When the preset algorithm calculates the initial compression rate threshold, the time-varying characteristics of the data need to be considered. For example, for voltage data, when the system is running stably, the changes between adjacent sampling points are small, and a higher compression rate can be used; under fault or fluctuating conditions, the compression rate needs to be reduced to ensure data quality.

[0146] In practical applications, the initial compression rate is usually set between four and six times. When the data size exceeds the preset threshold, such as when a large-scale fault causes the sampling frequency to increase, the system will intelligently adjust the reconstruction quality target. For example, when frequency fluctuations occur in a certain area, the reconstruction quality of the voltage data in that area can be reduced to 99.5%, while the reconstruction quality of the frequency data can be maintained above 99.9%.

[0147] The compressed monitoring data needs to meet business needs. Taking frequency monitoring as an example, the compressed data should accurately reflect the frequency change trend, and the error should be controlled within 0.01 Hz. The system uses indicators such as root mean square error and peak signal-to-noise ratio to evaluate compression performance, while monitoring resource overhead indicators such as processor occupancy and memory usage.

[0148] The performance evaluation model is based on a multi-index weighted scoring mechanism. For example, during the data compression process of a substation, when the compression rate is five times, if the processor occupancy rate exceeds 80%, the compression strategy needs to be adjusted even if the reconstruction quality meets the standard. By reducing the compression rate to four times, the processor occupancy rate can be reduced to about 60%, achieving a balance in resource utilization while ensuring data quality. The final monitoring results are used for scheduling decision support.

[0149] Practice has shown that the optimized compression scheme can reduce the amount of data transmission by 80 percent, while ensuring that the accuracy of power monitoring is not less than 99.8 percent, effectively supporting dispatchers in safety verification and early warning.

[0150] Step 1 includes the following steps: obtaining data for collaborative monitoring of the power dispatching system network and province; using a preset compression algorithm to process the data for collaborative monitoring of the power dispatching system network and province to obtain a preliminary compression result; according to a pre-established business quality evaluation index system, quantitatively evaluating the preliminary compression result to obtain an evaluation value; judging whether the evaluation value is lower than a preset threshold; if the evaluation value is lower than the preset threshold, calling a dynamic adjustment module to recalculate the compression rate threshold and reconstruction quality parameters; using the updated compression rate threshold and reconstruction quality parameters, re-executing the preset compression algorithm to obtain optimized compressed data; according to the update rules of the power data semantic association map, writing the optimized compressed data features into the power data semantic association map to obtain a new semantic map; using a collaborative evaluation algorithm to judge the degree of matching between the compressed data corresponding to the new semantic map and the business requirements to obtain an optimization degree; according to the optimization degree, determining the final compressed data version to obtain the compression optimization result of the data for collaborative monitoring of the power dispatching system network and province.

[0151] The acquisition of data for coordinated monitoring of power dispatching system and provincial power grid is the starting point of the entire compression optimization process. These data usually include multi-dimensional information such as power grid operation status, equipment parameters, load curves, etc. The preset compression algorithm may use methods such as wavelet transform or discrete cosine transform to perform preliminary compression processing for different types of data features.

[0152] The service quality evaluation index system is the key to measuring the compression effect. For example, for voltage data, the error tolerance can be set to ±0.5%; for power data, a deviation of ±1% is acceptable. These indicators will be converted into quantitative evaluation values, such as a comprehensive score of 90 points (out of 100 points). If the evaluation value is lower than the preset threshold (such as 85 points), the dynamic adjustment mechanism is triggered.

[0153] The core of the dynamic adjustment module is to balance the compression rate and data quality. Assume that the initial compression rate is 60% and the reconstruction quality parameter is set to 0.95. If the evaluation value does not meet the standard, the system may adjust the compression rate to 50% and increase the reconstruction quality parameter to 0.98 to ensure higher data accuracy. This adjustment is based on the sensitivity of the power system to different types of data. For example, frequency data may require higher accuracy.

[0154] The power data semantic association graph is a high-level data structure used to represent the logical relationship between various types of power data. For example, there is a close correlation between the output of a generator set and the system frequency. Writing optimized compressed data features into the graph may involve updating node attributes (such as data accuracy) or adjusting edge weights (such as correlation strength). This process helps maintain the semantic integrity of the data and ensures that the compressed data still accurately reflects the system status.

[0155] The synergy evaluation algorithm is designed to verify whether the compressed data meets actual business needs. For example, for power grid security and stability analysis, it may be necessary to check whether the voltage margin of key nodes remains within the allowable range. If the optimization degree reaches the expected level (such as more than 95% of business needs are met), the current version is confirmed as the final compressed data.

[0156] This multi-level, iterative compression optimization method can flexibly adjust the compression strategy according to the characteristics and real-time needs of the power system. It not only ensures data transmission and storage efficiency, but also guarantees the data quality required for power dispatch decisions. By combining data compression with semantic understanding, this method can significantly reduce data redundancy while retaining key information, providing strong support for the efficient operation of large-scale power systems.

[0157] In the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "left", "right", "up", "down", etc. are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0158] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection, it can be a direct connection, it can be an indirect connection through an intermediate medium, and it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

Claims

1. A data processing method for coordinated monitoring of power dispatching system network and province, characterized in that: The following steps are involved: a. Obtain the topological structure and equipment parameter information of the power system and construct a compressed dictionary containing typical patterns of voltage, current and power physical quantities; b. Obtaining domain ontology and semantic model, constructing semantic association graph across physical quantities and devices and integrating it into the compression dictionary; c. Obtain the change trend of the data stream and adaptively adjust the compression dictionary using incremental learning and forgetting mechanisms; d. for non-stationary power data, adaptive compression is achieved by continuously updating the compression dictionary; e. Using the semantic association graph to guide the selection of compression algorithms, achieving collaborative compression and semantic reconstruction of multi-source heterogeneous power data; f. Use autocorrelation analysis, mutual information calculation, principal component analysis and deep learning methods to build a multi-scale space-time compression model; g. Combining the multi-scale space-time compression model with the graph model to realize the fusion representation of the power system topology structure and data characteristics; h. Incorporating graph structure features into the prediction coding and differential coding of the compression process; iBuild a unified compression algorithm selection framework to achieve highly adaptable compression of complex power data; jUse reinforcement learning method to autonomously learn and optimize the parameter configuration of each compression algorithm; k Adaptively adjust the compression rate threshold and reconstruction quality target according to the accuracy tolerance of the power application; l Obtain application feedback on compressed data and dynamically adjust compression rate thresholds and reconstruction quality targets based on service quality assessment results.

2. A data processing method for coordinated monitoring of power dispatching system network and province according to claim 1, characterized in that: In step a, a compression dictionary containing typical patterns of the physical quantities of voltage, current, and power is constructed, including the following steps: Obtain the power grid topology and equipment parameter information from the power dispatching system, and extract the physical quantity information of voltage, current and power; Use the preset compression algorithm to build a compression dictionary; If the compression dictionary contains a typical pattern, mapping the typical pattern to the network-province collaborative monitoring system; If the compression dictionary does not contain a typical pattern, cluster analysis is performed on the physical quantity data using a preset machine learning algorithm to generate a new typical pattern and update the compression dictionary.

3. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: In step b, constructing a semantic association graph across physical quantities and devices and integrating it into the compression dictionary includes the following steps: Obtain domain ontology and semantic model in power dispatching system; If the domain ontology contains concepts related to power dispatching, a mapping relationship between physical quantities and equipment is extracted from the semantic model; According to the mapping relationship between the physical quantity and the device, a semantic association graph across physical quantities and across devices is constructed; The semantic association graph is integrated into a compression dictionary, and a machine learning algorithm is used to optimize the dictionary structure.

4. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: In step c, the compression dictionary is adaptively adjusted using an incremental learning and forgetting mechanism, including the following steps: Obtain the data stream of the power dispatching system network and provincial collaborative monitoring; Adopting an incremental learning method, extracting new typical features from the data stream and updating a preset compression dictionary; Through the forgetting mechanism, it is determined whether the data pattern in the compression dictionary is outdated. If the data pattern is outdated, the outdated data pattern is removed from the compression dictionary.

5. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: In step d, for non-stationary power data, adaptive compression is achieved by continuously updating the compression dictionary, including the following steps: Acquire non-stationary power data in a power dispatching system, analyze non-stationary characteristics of the non-stationary power data, and determine time series characteristics of the non-stationary power data; Using a preset initial compression dictionary, preliminarily compressing the non-stationary power data to obtain a first compression result; According to the data characteristics of the first compression result, the update requirement of the initial compression dictionary is determined, and if update is required, a new compression dictionary is generated.

6. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: Step e includes: Acquire multi-source heterogeneous power data in the power dispatching system, and perform classification preprocessing based on the type and semantic features of the power data; Searching for nodes matching the features of the power data in a pre-built semantic association graph, and determining the position and association relationship of the power data in the semantic association graph; According to the relevance of the power data in the semantic association graph, a suitable compression algorithm is selected.

7. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: Step f comprises: Obtain the time series of the coordinated monitoring data of the power dispatching system network and province, calculate the time scale characteristics of the time series by using the autocorrelation analysis method, and obtain the data fluctuation law under different time scales; For the monitoring data of different spatial granularities, a mutual information calculation method is used to perform correlation calculation to obtain the correlation strength in the spatial dimension; According to the time scale feature and the spatial correlation feature, a principal component analysis method is used to perform feature dimension reduction to obtain a key feature vector after dimension reduction; If the correlation between the time scale feature and the spatial correlation feature is greater than a preset threshold, a weighted fusion method is used to perform feature integration on the key feature vector to obtain an integrated feature vector; otherwise, the key feature vector is retained; The integrated feature vector is used as input, and a multi-scale spatiotemporal compression model is obtained through deep learning model training to obtain a time-space compression representation of the monitoring data.

8. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: Step h comprises: Obtaining original data, extracting graph structure features of the original data, and determining associations between data nodes; For the differential data, determining whether the dynamic range of the differential data exceeds a preset threshold, and if it exceeds the preset threshold, adjusting the dynamic range to obtain adjusted differential data; Using a target compression algorithm, compressing the adjusted differential data to generate a compressed data block; The compressed data blocks are integrated with the graph structure features to obtain optimized data.

9. A data processing method for coordinated monitoring of power dispatching system network and province as claimed in claim 1, characterized in that: Step k comprises: Obtaining the original data of the power dispatching system network-province collaborative monitoring; According to the data size and accuracy tolerance of the original data, a preset algorithm is used to determine an initial compression rate threshold; If the data size of the original data exceeds a preset threshold, the reconstruction quality target is adjusted, and an adaptive optimization algorithm is used to dynamically calculate the compression rate threshold; Obtaining compressed monitoring data according to the reconstruction quality target and the compression rate threshold; The compressed monitoring data is analyzed using a preset performance evaluation model to determine whether a balance between compression performance and resource overhead is achieved; If the equilibrium state does not meet the preset standard, the compression rate threshold and the reconstruction quality target are readjusted to iteratively optimize the compression process; The optimized compressed data is applied to the power dispatching system network-province collaborative monitoring to obtain the final monitoring result.

Images