A method for quality assessment of multi-source monitoring data of a use-energy device

Abstract

The application discloses a kind of quality evaluation methods for multi-source monitoring data of energy-using equipment, comprising: realizing comprehensive capture of abnormal data by multi-dimensional feature and space-time correlation attribute fusion, then combining space correlation strength and time sequence stability to construct dynamic correction mechanism, complete data deviation correction;Subsequently, the evaluation system is constructed by fusing space distribution and time variation characteristics, and the parameter weight is optimized by matching the dynamic adaptation mechanism of general parameters and scene-specific parameters, and finally the data quality level and detailed identification report are output.The application improves the comprehensiveness and adaptability of evaluation through the coordinated cooperation of hardware module and software strategy, solves the problems of insufficient cooperation of abnormal identification and correction and weak cross-scene adaptation capability of traditional methods, provides high-quality data support for energy-using equipment operation optimization and energy consumption management and control, and adapts to different types of equipment and complex monitoring scenarios.

Description

Technical Field

[0001] This invention relates to the field of multi-data quality assessment technology for energy-consuming equipment, and in particular to a method for quality assessment of multi-source monitoring data of energy-consuming equipment. Background Technology

[0002] The intelligent upgrading of energy-consuming equipment in the industrial sector is driving the continuous expansion of monitoring dimensions. During equipment operation, various types of related data are generated, including energy consumption, status, and environmental data. The reliability of this data directly affects the effectiveness of equipment operation control, energy efficiency improvement, and potential fault early warning. In multi-source and widely distributed monitoring scenarios, data acquisition involves multiple nodes and requires continuous time-series recording. Traditional data quality assessment models struggle to cope with the characteristics of dispersed data sources and complex spatiotemporal correlations, failing to meet the needs of precise control. Therefore, it is necessary to construct a comprehensive assessment scheme that can coordinate spatiotemporal dimensions and cross-source data collaboration to provide data assurance for the stable and efficient operation of energy-consuming equipment.

[0003] Currently, when assessing the quality of multi-source monitoring data for energy-consuming equipment, various types of data are typically collected first through conventional monitoring systems. Then, a single-angle anomaly identification method is used to filter data that does not conform to the conventional range. Subsequently, data deviations are corrected by simple adjacent data weighting or time-dimension interpolation. Finally, the quality is determined based on basic statistical indicators such as data dispersion and deviation values. Some solutions introduce simple parameter adjustment strategies to adapt to different equipment operating scenarios.

[0004] Existing technologies have two core drawbacks: First, the coordination between anomaly identification and data correction is insufficient. Anomaly screening relies on only a single feature dimension and fails to fully integrate the data's changing trends and spatiotemporal correlation characteristics, resulting in the failure to identify some latent anomalies. Furthermore, the correction process lacks a dynamic matching mechanism between spatial correlation and time series stability, making the correction accuracy susceptible to changes in the scenario. Second, the quality assessment system has weak cross-scenario adaptability. It lacks a dynamic adjustment mechanism for general parameters and scenario-specific parameters, and the parameter weight allocation lacks specificity. At the same time, the assessment process does not fully integrate spatial distribution characteristics and temporal variation characteristics, making it difficult to fully reflect the quality differences of multi-source monitoring data under different operating conditions and environmental conditions, thus limiting the reliability and applicability of the assessment results. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides a method for quality assessment of multi-source monitoring data of energy-consuming equipment.

[0006] The technical solution adopted in this invention is a method for quality assessment of multi-source monitoring data of energy-consuming equipment, comprising the following steps: S1, collecting multi-dimensional monitoring data generated during the operation of energy-consuming equipment through a cross-source monitoring data collaborative intelligent verification platform, wherein the multi-dimensional monitoring data includes energy consumption correlation parameters, equipment status parameters, and environmental correlation parameters of different monitoring nodes in a continuous time series; S2, extracting gradient features from the collected multi-dimensional monitoring data using an anomaly data gradient clustering identification algorithm, constructing a clustering sample space based on data fluctuation amplitude and change rate, and performing preliminary screening of anomaly data through iterative updates of cluster centers; S3, applying a spatiotemporal coupling correction algorithm to the energy consumption data after preliminary screening. The monitoring data is subjected to spatiotemporal coupling analysis, and data deviation is corrected by combining the spatial location correlation and time series continuity of different monitoring nodes; S4, the corrected monitoring data is fused with dual-domain features using a spatiotemporal dual-domain data quality discrimination model, and a data quality assessment index system is constructed through spatial domain consistency verification and temporal domain stability analysis; S5, the parameters in the assessment index system are optimized for cross-scenario adaptation by extracting common parameters and adapting localized parameters, and the parameter weight allocation is adjusted according to the operating characteristics of energy-consuming equipment and the differences in the monitoring environment; S6, the quality level results of multi-source monitoring data are output based on the optimized assessment index system, and detailed identification reports of data quality in each dimension are generated simultaneously.

[0007] Furthermore, the expression for the abnormal data gradient clustering identification algorithm is as follows:

[0008]

[0009]

[0010] in, The gradient feature value of the data at time j of the monitoring node. This indicates the time interval between the i-th monitoring node and the j-th monitoring node. The difference in data at different times, Indicates the time interval j and the time interval j. The time interval of time, This represents the gradient weight coefficient of the i-th monitoring node. Represents the gradient correction factor. This represents the spatial gradient vector of the data at time j of the i-th monitoring node; This represents the clustering effectiveness index of the k-th cluster. This represents the number of data samples included in the k-th cluster. This represents the set of data samples for the k-th cluster. This represents the mean gradient feature of the k-th cluster. This represents the standard deviation of the gradient features of the k-th cluster.

[0011] Furthermore, the expression for the spatiotemporal coupling correction algorithm for energy consumption data is:

[0012] ,

[0013] ;

[0014] in, This represents the spatial coupling correction value of the data from the i-th monitoring node at time j. This represents the raw data of the i-th monitoring node at time j. Represents the spatial coupling correction coefficient. This represents the total number of monitoring nodes adjacent to the i-th monitoring node. This represents the spatial distance between the i-th monitoring node and its m-th neighboring monitoring node. This represents the data of the m-th adjacent monitoring node at time j; This represents the final spatiotemporal coupling correction value of the data from the i-th monitoring node at time j. This represents the time coupling correction coefficient. Indicates the i-th monitoring node. Data at any given time Indicates the i-th monitoring node. Data at any given time This represents the time window parameter for time coupling correction.

[0015] Furthermore, the expression for the spatiotemporal dual-domain data quality discrimination model is:

[0016] ;

[0017] in, This represents the quality discrimination value of the data at time j of the i-th monitoring node. Represents the spatial domain feature weight coefficients. Indicates the spatial domain discriminant factor. This represents the spatial domain consistency index of the data at time j of the i-th monitoring node. Represents the time-domain discriminant factor. This represents the time-domain stability index of the data at time j of the i-th monitoring node.

[0018] Furthermore, the cross-source monitoring data collaborative intelligent verification platform includes an edge computing gateway module, a 5G industrial module, a high-precision time synchronization module, and distributed storage nodes. The edge computing gateway module adopts a heterogeneous multi-core processor architecture and integrates an FPGA hardware acceleration unit for parallel computing of data preprocessing. The 5G industrial module supports sliced ​​private network communication, and the data transmission latency is controlled within a preset threshold. The high-precision time synchronization module adopts a Beidou + PTP dual-mode synchronization mechanism, and the time synchronization accuracy reaches the microsecond level. The distributed storage nodes adopt a distributed file system, dynamically adjust the storage level based on data heat, and in the process of common parameter extraction and localized parameter adaptation, the common parameter features of different energy-consuming devices are mined through a deep neural network model. The localized parameter adaptation adopts a reinforcement learning algorithm to adjust the adaptation strategy in real time according to the operating conditions of the equipment, and the adaptation parameter update cycle is dynamically adjusted according to the data quality fluctuation.

[0019] Further, S2 includes the following sub-steps: S21, performing time series segmentation processing on the multi-dimensional monitoring data, dividing the data into segments according to a fixed time period, extracting the extreme value features and trend features of each data segment, and constructing an initial feature vector set; S22, solving the first-order difference and second-order difference of each data point based on the gradient calculation method, generating a gradient feature matrix by combining the feature vector set of the data segment, and eliminating feature dimension differences through matrix standardization; S23, constructing a clustering sample space, taking each row of data in the gradient feature matrix as a clustering sample, setting the initial number of cluster centers, and calculating the similarity between samples using a distance metric method; S24, iteratively updating the cluster centers until the change in cluster centers is less than a preset threshold, and marking samples far from the cluster centers as abnormal data candidate sets.

[0020] Further, step S3 includes the following sub-steps: S31, establishing a spatial topology model of monitoring nodes, calculating the spatial correlation strength between nodes based on the physical installation coordinates of the monitoring nodes, and generating a spatial correlation matrix; S32, performing correlation analysis on data from different monitoring nodes at the same time stamp, calculating the Pearson correlation coefficient, and selecting node combinations whose correlation meets preset conditions; S33, constructing a spatiotemporal coupling equation based on the spatial correlation matrix and the correlation analysis results, and substituting the original monitoring data to solve for the data deviation; S34, correcting the original data point by point according to the magnitude of the deviation, while recording the parameter adjustment trajectory during the correction process to form a correction log.

[0021] Further, step S4 includes the following sub-steps: S41, extracting spatial domain features of the corrected data, including the discrete coefficients, spatial distribution entropy, and regional consistency index of data at different nodes at the same time; S42, extracting temporal domain features of the corrected data, including the moving average deviation, trend fitting error, and periodic fluctuation amplitude of data at different times of the same node; S43, normalizing the spatial domain features and temporal domain features, fusing features according to preset weight coefficients, and generating a dual-domain fusion feature vector; S44, inputting the dual-domain fusion feature vector into the spatiotemporal dual-domain data quality discrimination model, outputting a data quality quantification score, and classifying quality levels according to the score.

[0022] Further, S5 includes the following sub-steps: S51, using a feature selection algorithm to screen out common parameters that significantly affect data quality from the evaluation index system and construct a common parameter set; S52, analyzing the operating condition parameters of energy-consuming equipment and monitoring environment parameters, establishing a localized parameter requirement model, and determining the type and value range of the adaptable parameters; S53, based on the common parameter set and the localized parameter requirement model, using a parameter mapping algorithm to convert common parameters into localized parameters and generate initial adaptable parameters; S54, adjusting the values ​​of the adaptable parameters and optimizing the parameter weight allocation ratio by real-time monitoring of data quality feedback results, forming a dynamic adaptation mechanism.

[0023] A method for quality assessment of multi-source monitoring data of energy-consuming equipment is disclosed. This method is implemented through a quality assessment system for multi-source monitoring data of energy-consuming equipment, comprising: a multi-source data acquisition and transmission unit, a gradient clustering anomaly identification unit, a spatiotemporal coupling data correction unit, a dual-domain quality discriminant analysis unit, a parameter collaborative adaptation optimization unit, and a quality assessment result output unit. The multi-source data acquisition and transmission unit acquires multi-dimensional monitoring data through a distributed sensor array and an industrial bus interface, and transmits it to the gradient clustering anomaly identification unit via an encrypted transmission protocol. After receiving the data, the gradient clustering anomaly identification unit performs anomaly data filtering and transmits the filtered data to... The spatiotemporal coupled data correction unit corrects data deviations and then sends the corrected data to the dual-domain quality discrimination analysis unit. The dual-domain quality discrimination analysis unit combines spatial and temporal domain features to output quality quantification results, which are simultaneously sent to the parameter collaborative adaptation optimization unit. Based on the quality quantification results, the parameter collaborative adaptation optimization unit adjusts parameter configurations and feeds the optimized parameters back to the preceding unit. Simultaneously, it sends the final quality assessment results and parameter configuration information to the quality assessment result output unit. The quality assessment result output unit generates a quality assessment report and data quality visualization charts in a preset format, providing multi-dimensional result display.

[0024] Beneficial Effects: This invention proposes a quality assessment method for multi-source monitoring data of energy-consuming equipment. Through multi-node collaborative acquisition and multi-dimensional feature fusion for anomaly identification, it fully integrates data change trends and spatiotemporal correlation characteristics to comprehensively capture hidden anomalies. Simultaneously, it constructs a dynamic matching and correction mechanism for spatial correlation and time series stability, significantly improving data correction accuracy in different scenarios. Addressing the weakness of the assessment system's cross-scenario adaptability, it establishes a dynamic adjustment mechanism for general parameters and scenario-specific parameters, accurately allocating parameter weights according to equipment operating conditions and environmental conditions. Furthermore, it comprehensively integrates spatial distribution characteristics and temporal variation characteristics in the quality assessment, fully reflecting the quality differences of multi-source monitoring data in different scenarios, greatly improving the reliability and applicability of the assessment results. This invention achieves accurate and comprehensive assessment of multi-source monitoring data quality, providing high-quality data support for energy-consuming equipment operation control, energy efficiency improvement, and fault early warning. It is also adaptable to different types of energy-consuming equipment and complex monitoring scenarios, reducing the scenario limitations of data quality assessment, improving the intelligence and efficiency of the assessment process, and ensuring the long-term stable and efficient operation of energy-consuming equipment. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method steps of the present invention;

[0026] Figure 2 This is a diagram showing the system unit composition for implementing the method of the present invention. Detailed Implementation

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1As shown, a method for quality assessment of multi-source monitoring data of energy-consuming equipment includes the following steps: S1, collecting multi-dimensional monitoring data generated during the operation of energy-consuming equipment through a cross-source monitoring data collaborative intelligent verification platform, wherein the multi-dimensional monitoring data includes energy consumption correlation parameters, equipment status parameters, and environmental correlation parameters of different monitoring nodes in a continuous time series; S2, extracting gradient features from the collected multi-dimensional monitoring data using an anomaly data gradient clustering identification algorithm, constructing a clustering sample space based on data fluctuation amplitude and change rate, and performing preliminary screening of anomaly data through iterative updates of cluster centers; S3, applying an energy consumption data spatiotemporal coupling correction algorithm to the preliminarily screened monitoring data. The process involves: S4, performing spatiotemporal coupling analysis to correct data deviations by combining the spatial location correlation and temporal series continuity of different monitoring nodes; S5, using a spatiotemporal dual-domain data quality discrimination model to perform dual-domain feature fusion on the corrected monitoring data, and constructing a data quality assessment index system through spatial domain consistency verification and temporal domain stability analysis; S6, optimizing the parameters in the assessment index system across scenarios by extracting common parameters and adapting localized parameters, and adjusting parameter weight allocation based on the operating characteristics of energy-consuming equipment and differences in the monitoring environment; and S7, outputting the quality level results of multi-source monitoring data based on the optimized assessment index system, and simultaneously generating detailed identification reports of data quality in each dimension.

[0029] Step S1 executes multi-dimensional monitoring data collection operations for energy-consuming equipment through a cross-source monitoring data collaborative intelligent verification platform. The edge computing gateway module integrated into the platform adopts a heterogeneous multi-core processor architecture, and is equipped with an FPGA hardware acceleration unit to ensure the efficiency of parallel processing of data collection. The 5G industrial module supports sliced ​​private network communication, and the data transmission latency is controlled within 50. The high-precision time synchronization module adopts a Beidou + PTP dual-mode synchronization mechanism, and the time synchronization accuracy reaches 10 microseconds. The distributed storage node adopts a distributed file system and dynamically adjusts the storage level based on the data heat. The collected multi-dimensional monitoring data specifically includes energy consumption-related parameters, equipment status parameters, and environmental-related parameters from 32 monitoring nodes over a continuous time series. Energy consumption-related parameters include 16 core parameters such as equipment operating power and cumulative energy consumption; equipment status parameters include 12 key parameters such as operating temperature and vibration frequency; and environmental-related parameters include 8 auxiliary parameters such as ambient temperature and humidity. The data acquisition interval is set to 1 second, and the amount of data acquired in a single session is controlled to 1024 bytes. During the acquisition process, data is acquired in real time through a distributed sensor array and an industrial bus interface. The data is then transmitted to a distributed storage node via an encrypted transmission protocol for temporary storage, ensuring data integrity and traceability in subsequent processing. This step provides comprehensive and accurate raw data support for the entire quality assessment process, and its implementation quality directly affects the accuracy of subsequent anomaly identification, data correction, and quality assessment.

[0030] Step S2 employs an anomaly data gradient clustering identification algorithm to extract gradient features from the collected multi-dimensional monitoring data. First, the data segmentation time period is set to 60 seconds, dividing the continuous monitoring data into several data segments. For each data segment, extreme value features such as maximum, minimum, and average values, and trend features such as linear and fluctuating trends are extracted, constructing an initial feature vector set with 24 dimensions. Then, the first-order and second-order differencing methods are used to solve for the gradient change value of each data point. The first-order differencing step size is set to 1, and the second-order differencing step size is set to 2. Combined with the initial feature vector set, a 128×64-dimensional gradient feature matrix is ​​generated. The gradient feature matrix is ​​processed using the min-max normalization method to eliminate dimensional differences between different features. Next, a clustering sample space is constructed, with each row of data in the gradient feature matrix treated as an independent clustering sample. The initial number of cluster centers is set to 8, and Euclidean distance is used as the similarity metric between samples to calculate the distance between each sample and each cluster center. Finally, the cluster centers are updated iteratively, with the number of iterations set to 50. The iteration terminates when the change in cluster centers is less than 0.001. Samples whose distance from all cluster centers is greater than 3 times the standard deviation are marked as candidate sets of outliers. This step, through refined gradient feature extraction and cluster analysis, achieves comprehensive screening of explicit and implicit outliers, laying the foundation for subsequent data correction steps.

[0031] Step S3 employs a spatiotemporal coupling correction algorithm for energy consumption data to perform spatiotemporal coupling analysis on the initially screened monitoring data. First, a spatial topology model of the monitoring nodes is established. The spatial distance between nodes is calculated based on their physical installation coordinates. The spatial correlation strength between nodes is determined based on this distance, using a distance-inverse weighted method to generate a 32×32 spatial correlation matrix. Subsequently, correlation analysis is performed on data from different monitoring nodes at the same time stamp, calculating the Pearson correlation coefficient. A threshold of 0.8 is set, and node combinations with correlation coefficients greater than this threshold are selected to form a set of effective correlated node pairs. Next, a spatiotemporal coupling equation is constructed based on the spatial correlation matrix and the set of effective correlated node pairs. The original monitoring data is then substituted to solve for the deviation at each data point, with the deviation calculation precision retained to six decimal places. Finally, the original data is corrected point by point according to the magnitude of the deviation, with the correction coefficient set to 0.95. At the same time, the parameter adjustment trajectory during the correction process is recorded, including the changes in key parameters such as spatial correlation strength, correlation coefficient, and deviation, forming a correction log. This step fully explores the spatial correlation and time series continuity of the data to achieve accurate correction of data deviation, improve the accuracy and consistency of monitoring data, and provide a reliable data foundation for subsequent quality judgment.

[0032] Step S4 utilizes a spatiotemporal dual-domain data quality discrimination model to perform dual-domain feature fusion on the corrected monitoring data. First, spatial domain features are extracted, calculating the coefficient of variation, spatial distribution entropy, and regional consistency index for data at different nodes at the same time. The coefficient of variation is calculated using the ratio of standard deviation to mean. Spatial distribution entropy is calculated based on information entropy theory. The regional consistency index is obtained by calculating the degree of deviation between the data within the region and the regional mean. Each spatial domain feature has three levels of classification. Next, temporal domain features are extracted, calculating the moving average deviation, trend fitting error, and periodic fluctuation amplitude for data at different times at the same node. The moving average window size is set to 30 data points. A linear regression model is used for trend fitting, and the periodic fluctuation amplitude is extracted using Fourier transform to obtain the amplitude value corresponding to the dominant frequency. Then, the spatial and temporal features are normalized, with a normalization range of 0 to 1. Feature fusion is performed with a weighting of 0.6 for spatial domain features and 0.4 for temporal domain features, generating a dual-domain fused feature vector with six dimensions. Finally, the dual-domain fusion feature vector is input into the spatiotemporal dual-domain data quality discrimination model, and the output data quality quantification score is set from 0 to 100. Based on the score, four quality levels are divided: excellent, good, qualified, and unqualified. A score greater than 90 is excellent, 80 to 90 is good, 60 to 80 is qualified, and less than 60 is unqualified. This step achieves a scientific and accurate assessment of data quality by comprehensively fusing spatiotemporal dual-domain features.

[0033] Step S5 optimizes the parameters in the evaluation index system across scenarios through common parameter extraction and localized parameter adaptation. First, the ReliefF feature selection algorithm is used to select 12 common parameters that significantly impact data quality from the 24 parameters in the evaluation index system, constructing a common parameter set. The feature importance threshold is set to 0.7 during the selection process. Then, 16 types of operating condition parameters and 8 types of monitoring environment parameters of energy-consuming equipment are analyzed to establish a localized parameter requirement model. The types of adaptation parameters are determined to include 6 categories, such as weight coefficients, discrimination thresholds, and correction coefficients. Each adaptation parameter has a defined value range: weight coefficients range from 0.1 to 0.9, discrimination thresholds range from 0.2 to 0.8, and correction coefficients range from 0.8 to 1.0. Next, based on the common parameter set and the localized parameter requirement model, a linear mapping algorithm is used to convert common parameters to localized parameters, generating initial adaptation parameters. The mapping coefficient is determined according to the equipment type: 0.85 for industrial boilers and 0.90 for central air conditioning equipment. Finally, by monitoring the data quality feedback results in real time, the values ​​of the adaptation parameters are adjusted with an adjustment step size of 0.01, and the parameter weight allocation ratio is optimized to form a dynamic adaptation mechanism. The update cycle of the adaptation parameters is determined according to the fluctuation of data quality. When the fluctuation range is less than 5, the update cycle is 3600 seconds, and when the fluctuation range is greater than or equal to 5, the update cycle is 1800 seconds. This step improves the cross-scenario applicability of the evaluation method through precise parameter adaptation and dynamic optimization.

[0034] Step S6 outputs the quality level results of multi-source monitoring data based on the optimized evaluation index system, and simultaneously generates detailed identification reports of data quality for each dimension. First, the quality quantification score and quality level of each monitoring node and each timestamp data output in step S4 are obtained. The percentages of excellent, good, qualified, and unqualified data for each node are statistically analyzed according to the monitoring node group. The statistical time granularity is set to 1 hour, and one set of statistical results is generated for each node, resulting in 32 sets of statistical data for 32 monitoring nodes. Subsequently, a detailed analysis of the data quality for each dimension is conducted, including the quality distribution of energy consumption-related parameters, equipment status parameters, and environmentally related parameters. The distribution of anomaly types in different dimensions of data is analyzed, including the proportion of anomalies such as gradient anomalies and spatiotemporal inconsistencies. Next, a detailed identification report is generated, comprising six parts: data collection overview, anomaly identification results, data correction details, quality assessment results, parameter adaptation status, and optimization suggestions. The data collection overview includes key information such as collection duration, total data volume, and effective data volume. The anomaly identification results include the number of anomalies, the distribution of anomaly types, and the distribution of anomaly nodes. The data correction details include parameters such as the amount of corrected data and correction accuracy. The quality assessment results include information such as the overall quality level and the quality levels of each dimension. The parameter adaptation status includes the specific values ​​of common and localized parameters. The optimization suggestions propose targeted improvement measures based on the quality assessment results. Finally, the quality level results and the detailed identification report are stored on a distributed storage node and displayed through the visualization interface of the cross-source monitoring data collaborative intelligent verification platform. This step provides an intuitive and comprehensive data quality reference for the operation optimization, energy consumption control, and fault early warning of energy-consuming equipment.

[0035] Preferably, the expression for the anomaly data gradient clustering identification algorithm is:

[0036]

[0037]

[0038] in, The gradient feature value of the data at time j of the monitoring node. This indicates the time interval between the i-th monitoring node and the j-th monitoring node. The difference in data at different times, Indicates the time interval j and the time interval j. The time interval of time, This represents the gradient weight coefficient of the i-th monitoring node. Represents the gradient correction factor. This represents the spatial gradient vector of the data at time j of the i-th monitoring node; This represents the clustering effectiveness index of the k-th cluster. This represents the number of data samples included in the k-th cluster. This represents the set of data samples for the k-th cluster. This represents the mean gradient feature of the k-th cluster. This represents the standard deviation of the gradient features of the k-th cluster.

[0039] Specifically, the anomaly data gradient clustering identification algorithm is based on a progressive logic of data gradient features and clustering effectiveness analysis. Targeting the temporal volatility and spatial distribution correlation of multi-source monitoring data, it constructs an initial gradient feature calculation formula by combining the data change rate and gradient vector. The basic gradient value is obtained by the ratio of the difference between adjacent data points to the time interval. Then, a monitoring node weight coefficient is introduced to balance the importance of data from different nodes, and a gradient correction factor is added to optimize the accuracy of gradient calculation. Finally, a gradient feature value calculation formula that comprehensively reflects the spatiotemporal change characteristics of the data is formed. This formula is based on the premise that data anomalies are often accompanied by gradient abrupt changes, and multi-dimensional gradient fusion can accurately capture anomaly precursors. The subsequent clustering effectiveness index calculation formula is based on the preceding gradient feature value calculation. It calculates the sum of squared deviations of the gradient feature values ​​of all samples within a cluster from the cluster mean, and normalizes this by combining the cluster standard deviation to eliminate the influence of differences in data distribution between different clusters. Clustering effectiveness is reflected by the compactness of samples within the cluster; this compactness can be quantified by the combination of the sum of squared deviations and the standard deviation. The gradient weight coefficient for monitoring nodes is set to 0.6 to 0.9 based on the node monitoring accuracy, the gradient correction factor is set to 0.1 to 0.3, and the number of clusters is set to 6 to 10. During iteration, the cluster mean and standard deviation are updated in real time. In implementation, the gradient feature value of each data point is first calculated using the initial formula, and then cluster samples are constructed based on this value. The cluster centers are optimized through iterative iteration until the clustering effectiveness index converges. This algorithm, by combining gradient features with cluster analysis, achieves accurate identification of abnormal data, providing a reliable screening basis for subsequent data processing.

[0040] Preferably, the expression for the spatiotemporal coupling correction algorithm for energy consumption data is:

[0041] ,

[0042] ;

[0043] in, This represents the spatial coupling correction value of the data from the i-th monitoring node at time j. This represents the raw data of the i-th monitoring node at time j. Represents the spatial coupling correction coefficient. This represents the total number of monitoring nodes adjacent to the i-th monitoring node. This represents the spatial distance between the i-th monitoring node and its m-th neighboring monitoring node. This represents the data of the m-th adjacent monitoring node at time j; This represents the final spatiotemporal coupling correction value of the data from the i-th monitoring node at time j. This represents the time coupling correction coefficient. Indicates the i-th monitoring node. Data at any given time Indicates the i-th monitoring node. Data at any given time This represents the time window parameter for time coupling correction.

[0044] Specifically, the spatiotemporal coupling correction algorithm for energy consumption data is based on the logic of spatial correction first and spatiotemporal fusion optimization. The spatial coupling correction expression is based on the spatial correlation of monitoring nodes. First, the difference between adjacent node data reflects the deviation of target node data. Then, a weighted coefficient of the inverse of spatial distance is introduced to balance the influence weight of different adjacent nodes. The correction coefficient is used to adjust the strength of spatial coupling. The basis for its establishment is that the data of adjacent monitoring nodes in the same area have strong correlation, and local deviations can be corrected through spatial collaboration. The final spatiotemporal coupling correction expression is calculated based on the spatial correction results of the first expression. The second-order difference in the time domain is introduced to reflect the smoothness of the data time series. Combining the temporal coupling correction coefficient and the time window parameter, the time dimension of the spatially corrected data is optimized. Data deviation not only comes from spatial differences but is also affected by time series fluctuations. Comprehensive correction is achieved through spatiotemporal dual-domain fusion. The spatial coupling correction coefficient is set to 0.7 to 0.9, the temporal coupling correction coefficient is set to 0.2 to 0.4, and the time window parameter is set to 3 to 5 acquisition cycles according to the data acquisition interval. The spatial distance between adjacent monitoring nodes is accurately calculated using physical coordinates. During implementation, the spatial dimension data is first corrected using the primary formula, and then substituted into the secondary formula for time dimension optimization. During the correction process, changes in parameters such as spatial correlation strength and correlation coefficient are recorded in real time to form a correction trajectory. Through the spatiotemporal coupling mechanism, this algorithm effectively reduces data deviation, improves the consistency and accuracy of monitoring data, and provides a high-quality data foundation for quality assessment.

[0045] Preferably, the expression for the spatiotemporal dual-domain data quality discrimination model is:

[0046] ;

[0047] in, This represents the quality discrimination value of the data at time j of the i-th monitoring node. Represents the spatial domain feature weight coefficients. Indicates the spatial domain discriminant factor. This represents the spatial domain consistency index of the data at time j of the i-th monitoring node. Represents the time-domain discriminant factor. This represents the time-domain stability index of the data at time j of the i-th monitoring node.

[0048] Specifically, the spatiotemporal dual-domain data quality discrimination model is based on the collaborative evaluation logic of spatial and temporal domain features. By analyzing the core influencing factors of data quality, it identifies spatial domain consistency and temporal domain stability as key evaluation dimensions. The Sigmoid function is used to normalize the spatial domain consistency and temporal stability indices, mapping their values ​​to the 0-1 range to eliminate dimensional differences between different dimensions. A spatial domain feature weight coefficient is then introduced to balance the importance of dual-domain features. The model is based on the premise that the quality of multi-source monitoring data must simultaneously satisfy both spatial distribution rationality and temporal stability. Single-dimensional evaluation is one-sided, and dual-domain fusion enables comprehensive judgment. The spatial domain feature weight coefficient is set to 0.5-0.7 based on the monitoring scenario, the spatial domain discrimination factor is set to 2-4, and the temporal domain discrimination factor is set to 1.5-3. The spatial domain consistency index is calculated based on the dispersion of data from different nodes at the same time, and the temporal domain stability index is obtained based on the fluctuation amplitude of data from the same node at different times. During implementation, the spatial and temporal features of the corrected data are first extracted, and the corresponding consistency and stability indices are calculated. These indices are then substituted into the model expression to obtain the data quality discrimination value. The closer the discrimination value is to 1, the better the data quality. This model achieves accurate quantitative evaluation of the quality of multi-source monitoring data through deep fusion of dual-domain features, providing a scientific basis for parameter adaptation and result output.

[0049] Preferably, the cross-source monitoring data collaborative intelligent verification platform includes an edge computing gateway module, a 5G industrial module, a high-precision time synchronization module, and distributed storage nodes. The edge computing gateway module adopts a heterogeneous multi-core processor architecture and integrates an FPGA hardware acceleration unit for parallel computing of data preprocessing. The 5G industrial module supports sliced ​​private network communication, and the data transmission latency is controlled within a preset threshold. The high-precision time synchronization module adopts a Beidou + PTP dual-mode synchronization mechanism, and the time synchronization accuracy reaches the microsecond level. The distributed storage nodes adopt a distributed file system, dynamically adjust the storage level based on data heat, and in the process of common parameter extraction and localized parameter adaptation, the common parameter features of different energy-consuming devices are mined through a deep neural network model. The localized parameter adaptation adopts a reinforcement learning algorithm to adjust the adaptation strategy in real time according to the operating conditions of the equipment, and the adaptation parameter update cycle is dynamically adjusted according to the data quality fluctuation.

[0050] Preferably, step S2 includes the following sub-steps: S21, performing time series segmentation processing on the multi-dimensional monitoring data, dividing the data into segments according to a fixed time period, extracting the extreme value features and trend features of each data segment, and constructing an initial feature vector set; S22, solving the first-order difference and second-order difference of each data point based on the gradient calculation method, generating a gradient feature matrix by combining the feature vector set of the data segment, and eliminating feature dimension differences through matrix standardization; S23, constructing a clustering sample space, taking each row of data in the gradient feature matrix as a clustering sample, setting the initial number of cluster centers, and calculating the similarity between samples using a distance metric method; S24, iteratively updating the cluster centers until the change in cluster centers is less than a preset threshold, and marking samples far from the cluster centers as abnormal data candidate sets.

[0051] Specifically, step S2 includes S21, which first performs time series segmentation processing on the multi-dimensional monitoring data, setting a fixed time period of 60 seconds. The continuously collected data stream is divided into several independent data segments according to this period. For each segment, four types of extreme value features (maximum, minimum, range, and average) and three types of trend features (linear rate of change, fluctuation amplitude, and number of trend inflection points) are extracted simultaneously. These are integrated to form an initial feature vector set with seven dimensions, providing basic data support for subsequent gradient analysis. S22 uses a difference calculation method to solve for the first and second differences of each data point. The first difference is based on adjacent timestamp data, and the second difference is further derived based on the first difference result. The obtained difference data is concatenated column-wise with the initial feature vector set constructed in S21 to generate a 128-row, 64-column gradient feature matrix. Subsequently, the elements in the matrix are scaled using max-min standardization to ensure that all feature values ​​are in the range of 0 to 1, completely eliminating dimensional differences between different features. S23 constructs a clustering sample space, treating each row of data in the gradient feature matrix as a single independent clustering sample. Based on the complexity of the monitored data, the initial number of cluster centers is set to 8. Euclidean distance is used as the metric for sample similarity, and the distance between each sample and each of the 8 cluster centers is calculated to form a sample-center distance matrix. S24 initiates an iterative update process for cluster centers, setting a maximum of 50 iterations. The iteration terminates when the change in all cluster centers between two consecutive iterations is less than 0.001. Sample affiliation is recalculated after each iteration until the termination condition is met. Finally, samples whose distance to all cluster centers is greater than 3 times the global standard deviation are marked as candidate abnormal data. This step, through refined, phased processing, achieves comprehensive screening of abnormal data, laying the foundation for subsequent data correction.

[0052] Preferably, step S3 includes the following sub-steps: S31, establishing a spatial topology model of monitoring nodes, calculating the spatial correlation strength between nodes based on the physical installation coordinates of the monitoring nodes, and generating a spatial correlation matrix; S32, performing correlation analysis on data from different monitoring nodes at the same time stamp, calculating the Pearson correlation coefficient, and selecting node combinations whose correlation meets preset conditions; S33, constructing a spatiotemporal coupling equation based on the spatial correlation matrix and the correlation analysis results, and substituting the original monitoring data to solve for the data deviation; S34, correcting the original data point by point according to the magnitude of the deviation, while recording the parameter adjustment trajectory during the correction process to form a correction log.

[0053] Specifically, step S3 includes S31, first establishing a spatial topology model of the monitoring nodes. By obtaining the physical installation coordinates of 32 monitoring nodes, the spatial distance between any two nodes is calculated using the Euclidean distance formula. Based on the distance values, a spatial correlation strength matrix is ​​constructed. The matrix elements range from 0 to 1, with closer distances resulting in stronger correlation values. Finally, a 32-row, 32-column spatial correlation matrix is ​​generated, clearly presenting the spatial correlation characteristics between nodes. S32, for all monitoring node data at the same timestamp, the Pearson correlation coefficient calculation method is used to calculate the correlation coefficient between any two node data points. A correlation coefficient threshold of 0.8 is set, and node combinations with correlation coefficients greater than this threshold are selected to form a set of effective correlated node pairs. Each node has at least 5 effective correlated nodes to ensure the reliability of spatial correction. S33, based on the spatial correlation matrix generated in S31 and the set of effective correlation node pairs obtained in S32, constructs a spatiotemporal coupling correction equation. This equation incorporates spatial correlation strength weights and correlation coefficient weights. The original monitoring data is substituted into the equation, and the deviation for each data point is obtained by solving the equation. The deviation calculation is retained to six decimal places to ensure the accuracy of the correction. S34, based on the calculated deviation, performs point-by-point correction on the original data. A correction coefficient of 0.95 is set, and the product of the deviation and the correction coefficient is used as the correction increment to adjust the original data. Simultaneously, the changes in key parameters during the correction process are recorded in real time, including spatial correlation strength values, correlation coefficient values, deviation magnitudes, and correction increments, forming a complete correction log. This step, through the phased implementation of spatial correlation analysis and spatiotemporal coupling correction, effectively corrects data deviations and improves the consistency and accuracy of the monitoring data.

[0054] Preferably, step S4 includes the following sub-steps: S41, extracting spatial domain features of the corrected data, including the discrete coefficients, spatial distribution entropy, and regional consistency index of data at different nodes at the same time; S42, extracting temporal domain features of the corrected data, including the moving average deviation, trend fitting error, and periodic fluctuation amplitude of data at different times at the same node; S43, normalizing the spatial domain features and temporal domain features, fusing features according to preset weight coefficients, and generating a dual-domain fusion feature vector; S44, inputting the dual-domain fusion feature vector into the spatiotemporal dual-domain data quality discrimination model, outputting a data quality quantification score, and classifying quality levels according to the score.

[0055] Specifically, step S4 includes S41, firstly extracting the spatial domain features of the corrected data. For similar data from 32 monitoring nodes at the same time, the coefficient of variation, spatial distribution entropy, and regional consistency index are calculated. The coefficient of variation is calculated as the ratio of the standard deviation to the mean. The spatial distribution entropy is calculated based on the information entropy theory. The regional consistency index is obtained by calculating the average absolute deviation of all data within the region from the regional mean. Each spatial domain feature is divided into four levels: excellent, good, qualified, and unqualified, with corresponding value ranges of 0 to 0.2, 0.2 to 0.4, 0.4 to 0.6, and 0.6 to 1.0, respectively. S42, extracting the temporal domain features of the data. For data from 3600 consecutive timestamps at the same node, the moving average deviation, trend fitting error, and periodic fluctuation amplitude are calculated. The moving average window size is set to 30 data points. The trend fitting uses a linear regression model, and the fitting parameters are solved using the least squares method. The periodic fluctuation amplitude is extracted by Fourier transform to obtain the amplitude value corresponding to the main frequency. The temporal domain features are also divided into four levels, with value ranges consistent with the spatial domain features. S43 normalizes the extracted three spatial domain features and three temporal domain features respectively. After normalization, all feature values ​​are mapped to the interval between 0 and 1. A weighted sum is then generated using a weighting ratio of 0.6 for spatial domain features and 0.4 for temporal domain features, resulting in a dual-domain fusion feature vector with six dimensions. Each element in the vector corresponds to a fused feature value. S44 inputs the dual-domain fusion feature vector into a spatiotemporal dual-domain data quality discrimination model. The model outputs a data quality quantification score, ranging from 0 to 100. Quality levels are assigned based on the score: 90-100 is excellent, 80-90 is good, 60-80 is acceptable, and 0-60 is unacceptable. This step, through the comprehensive extraction and fusion of spatial and temporal domain features, achieves a scientific and accurate assessment of data quality.

[0056] Preferably, step S5 includes the following sub-steps: S51, using a feature selection algorithm to screen out common parameters that significantly affect data quality from the evaluation index system and construct a common parameter set; S52, analyzing the operating condition parameters of energy-consuming equipment and monitoring environment parameters, establishing a localized parameter requirement model, and determining the type and value range of the adaptable parameters; S53, based on the common parameter set and the localized parameter requirement model, using a parameter mapping algorithm to convert common parameters into localized parameters and generate initial adaptable parameters; S54, adjusting the values ​​of the adaptable parameters and optimizing the parameter weight allocation ratio by monitoring the data quality feedback results in real time, thus forming a dynamic adaptation mechanism.

[0057] Specifically, step S5 includes S51 using the ReliefF feature selection algorithm to rank the importance of the 24 parameters in the evaluation index system, setting the feature importance threshold to 0.7, and selecting 12 parameters with importance scores higher than the threshold as common parameters to construct a common parameter set. These common parameters include key parameters for the entire process such as data collection, anomaly identification, data correction, and quality judgment, ensuring the universality and core nature of the parameters. S52 comprehensively analyzes 16 types of operating condition parameters and 8 types of monitoring environmental parameters for energy-consuming equipment. Operating condition parameters include operating power, load rate, and working duration, while monitoring environmental parameters include ambient temperature, humidity, and air pressure. Based on the analysis results, a localized parameter requirement model is established, specifying six types of adaptable parameters: weighting coefficient, discrimination threshold, correction coefficient, iteration count, sample size, and distance threshold. The value ranges for each type of adaptable parameter are also determined: weighting coefficient from 0.1 to 0.9, discrimination threshold from 0.2 to 0.8, correction coefficient from 0.8 to 1.0, iteration count from 30 to 50, sample size from 500 to 1000, and distance threshold from 0.5 to 2.0. S53, based on the common parameter set constructed in S51 and the localized parameter requirement model established in S52, uses a linear mapping algorithm to convert common parameters to localized parameters. Mapping coefficients are set according to the type of energy-consuming equipment: 0.85 for industrial boilers, 0.90 for central air conditioning equipment, and 0.88 for fans. Initial adaptation parameters are generated through mapping calculations. S54 establishes a real-time data quality feedback mechanism, collecting data quality assessment results every 1800 seconds. Based on the quality level distribution and score fluctuations in the assessment results, the adaptation parameter values ​​are adjusted with an adjustment step size of 0.01, dynamically optimizing the parameter weight allocation ratio to form a closed-loop dynamic adaptation mechanism. When the data quality fluctuation amplitude is less than 5, the adaptation parameter update cycle is extended to 3600 seconds; when the fluctuation amplitude is greater than or equal to 5, the update cycle remains at 1800 seconds. This step, through precise parameter extraction, adaptation, and dynamic optimization, significantly improves the cross-scenario applicability of the assessment method.

[0058] like Figure 2As shown, a method for quality assessment of multi-source monitoring data of energy-consuming equipment is implemented through a quality assessment system for multi-source monitoring data of energy-consuming equipment, including: a multi-source data acquisition and transmission unit, a gradient clustering anomaly identification unit, a spatiotemporal coupling data correction unit, a dual-domain quality discriminant analysis unit, a parameter collaborative adaptation optimization unit, and a quality assessment result output unit. The multi-source data acquisition and transmission unit acquires multi-dimensional monitoring data through a distributed sensor array and an industrial bus interface, and sends it to the gradient clustering anomaly identification unit via an encrypted transmission protocol. After receiving the data, the gradient clustering anomaly identification unit performs anomaly data filtering and transmits the filtered data to the system. The data is then sent to the spatiotemporal coupling data correction unit. After correcting the data deviation, the spatiotemporal coupling data correction unit sends the corrected data to the dual-domain quality discrimination analysis unit. The dual-domain quality discrimination analysis unit combines spatial and temporal domain features to output quality quantification results, which are simultaneously sent to the parameter collaborative adaptation optimization unit. The parameter collaborative adaptation optimization unit adjusts the parameter configuration based on the quality quantification results, feeds back the optimized parameters to the preceding unit, and simultaneously sends the final quality assessment results and parameter configuration information to the quality assessment result output unit. The quality assessment result output unit generates a quality assessment report and data quality visualization charts in a preset format for multi-dimensional result display.

[0059] A quality assessment method for multi-source monitoring data of energy-consuming equipment addresses the problem of insufficient coordination between anomaly identification and data correction. It utilizes a multi-node collaborative data acquisition mode to integrate multi-dimensional features and spatiotemporal correlation attributes of the data, achieving comprehensive capture of hidden anomalies and avoiding missed detections caused by single-dimensional identification. Simultaneously, a dynamically adaptable correction mechanism is constructed, deeply combining spatial correlation strength and temporal series stability. The correction strategy is adjusted in real time according to data change patterns, significantly improving data correction accuracy under different monitoring scenarios and overcoming the limitations of traditional correction methods that are restricted by specific scenarios. To address the weak cross-scenario adaptability of the assessment system, a dynamic adaptation mechanism is established between general parameters and scenario-specific parameters. Combined with differences in the operating conditions and environment of energy-consuming equipment, parameter weight allocation is precisely optimized. Furthermore, spatial distribution characteristics and temporal variation characteristics are comprehensively integrated into the quality assessment, fully presenting the quality differences of multi-source data under different scenarios, significantly improving the reliability and applicability of the assessment results.

[0060] This method offers comprehensive evaluation by integrating multi-dimensional features and employing spatiotemporal dual-domain analysis to achieve a holistic and in-depth assessment of the quality of multi-source monitoring data, avoiding the one-sidedness of traditional single-indicator evaluations. It also boasts excellent adaptability, leveraging a dynamic parameter adjustment mechanism to flexibly adapt to different types of energy-consuming equipment and complex monitoring environments, reducing evaluation costs associated with scenario switching. Furthermore, it is highly intelligent, requiring no manual intervention throughout the entire process. Multi-unit collaborative operation enables automatic data processing, correction, and evaluation, improving evaluation efficiency. Finally, the method provides reliable data support; the high-quality data, after precise correction and comprehensive evaluation, provides a solid guarantee for optimizing the operation of energy-consuming equipment, controlling energy consumption, and providing early warning of faults, thus contributing to the long-term stable and efficient operation of equipment.

[0061] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0062] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for quality assessment of multi-source monitoring data from energy-consuming equipment, characterized in that, Includes the following steps: S1. Collect multi-dimensional monitoring data generated during the operation of energy-consuming equipment through a cross-source monitoring data collaborative intelligent verification platform. The multi-dimensional monitoring data includes energy consumption correlation parameters, equipment status parameters and environmental correlation parameters of different monitoring nodes in a continuous time series. S2, using an anomaly data gradient clustering identification algorithm to extract gradient features from the collected multi-dimensional monitoring data, constructing a clustering sample space based on the data fluctuation amplitude and change rate, and performing preliminary screening of anomalies through iterative updates of cluster centers; S3 uses a spatiotemporal coupling correction algorithm for energy consumption data to perform spatiotemporal dimension coupling analysis on the initially screened monitoring data, and corrects data deviations by combining the spatial location correlation and time series continuity of different monitoring nodes. S4. The spatiotemporal dual-domain data quality discrimination model is used to perform dual-domain feature fusion on the corrected monitoring data. A data quality evaluation index system is constructed through spatial domain consistency verification and temporal domain stability analysis. S5 optimizes parameters in the evaluation index system across scenarios by extracting public parameters and adapting localized parameters, and adjusts parameter weight allocation based on the operating characteristics of energy-consuming equipment and the differences in the monitoring environment. S6 outputs the quality level results of multi-source monitoring data based on the optimized evaluation index system, and simultaneously generates detailed identification reports of data quality in each dimension.

2. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, The expression for the gradient clustering algorithm for identifying abnormal data is: ； ； in, The gradient feature value of the data at time j of the monitoring node. This indicates the time interval between the i-th monitoring node and the j-th monitoring node. The difference in data at different times, Indicates the time interval j and the time interval j. The time interval of time, This represents the gradient weight coefficient of the i-th monitoring node. Represents the gradient correction factor. This represents the spatial gradient vector of the data at time j of the i-th monitoring node; This represents the clustering effectiveness index of the k-th cluster. This represents the number of data samples included in the k-th cluster. This represents the set of data samples for the k-th cluster. This represents the mean gradient feature of the k-th cluster. This represents the standard deviation of the gradient features of the k-th cluster.

3. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, The expression for the spatiotemporal coupling correction algorithm for energy consumption data is: ， ； in, This represents the spatial coupling correction value of the data from the i-th monitoring node at time j. This represents the raw data of the i-th monitoring node at time j. Represents the spatial coupling correction coefficient. This represents the total number of monitoring nodes adjacent to the i-th monitoring node. This represents the spatial distance between the i-th monitoring node and its m-th neighboring monitoring node. This represents the data of the m-th adjacent monitoring node at time j; This represents the final spatiotemporal coupling correction value of the data from the i-th monitoring node at time j. This represents the time coupling correction coefficient. Indicates the i-th monitoring node. Data at any given time Indicates the i-th monitoring node. Data at any given time This represents the time window parameter for time coupling correction.

4. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, The expression for the spatiotemporal dual-domain data quality discrimination model is: ； in, This represents the quality discrimination value of the data at time j of the i-th monitoring node. Represents the spatial domain feature weight coefficients. Indicates the spatial domain discriminant factor. This represents the spatial domain consistency index of the data at time j of the i-th monitoring node. Represents the time-domain discriminant factor. This represents the time-domain stability index of the data at time j of the i-th monitoring node.

5. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, The cross-source monitoring data collaborative intelligent verification platform includes an edge computing gateway module, a 5G industrial module, a high-precision time synchronization module, and distributed storage nodes. The edge computing gateway module adopts a heterogeneous multi-core processor architecture and integrates an FPGA hardware acceleration unit for parallel computing of data preprocessing. The 5G industrial module supports sliced ​​private network communication, and the data transmission latency is controlled within a preset threshold. The high-precision time synchronization module adopts a Beidou + PTP dual-mode synchronization mechanism, and the time synchronization accuracy reaches the microsecond level. The distributed storage nodes adopt a distributed file system, dynamically adjust the storage level based on data heat, and in the process of common parameter extraction and localized parameter adaptation, the common parameter features of different energy-consuming devices are mined through a deep neural network model. The localized parameter adaptation adopts a reinforcement learning algorithm to adjust the adaptation strategy in real time according to the operating conditions of the equipment, and the adaptation parameter update cycle is dynamically adjusted according to the data quality fluctuation.

6. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, S2 includes the following sub-steps: S21, performing time series segmentation processing on the multi-dimensional monitoring data, dividing the data into segments according to a fixed time period, extracting the extreme value features and trend features of each data segment, and constructing an initial feature vector set; S22, Solve the first and second differences of each data point based on the gradient calculation method, generate a gradient feature matrix by combining the feature vector set of the data segment, and eliminate the difference in feature dimensions through matrix standardization; S23, Construct a clustering sample space, take each row of data in the gradient feature matrix as a clustering sample, set the initial number of cluster centers, and calculate the similarity between samples using a distance metric method; S24, Iteratively update the cluster centers until the change in cluster centers is less than a preset threshold, and mark the samples far from the cluster centers as abnormal data candidate sets.

7. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, S3 includes the following sub-steps: S31, establishing a spatial topology model of monitoring nodes, calculating the spatial correlation strength between nodes based on the physical installation coordinates of the monitoring nodes, and generating a spatial correlation matrix; S32, performing correlation analysis on data from different monitoring nodes at the same time stamp, calculating the Pearson correlation coefficient, and selecting node combinations whose correlation meets preset conditions; S33, constructing a spatiotemporal coupling equation based on the spatial correlation matrix and correlation analysis results, and substituting the original monitoring data to solve for the data deviation; S34, correcting the original data point by point according to the magnitude of the deviation, while recording the parameter adjustment trajectory during the correction process to form a correction log.

8. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, S4 includes the following sub-steps: S41, extracting the spatial domain features of the corrected data, including the discrete coefficients, spatial distribution entropy, and regional consistency index of data at different nodes at the same time; S42, extracting the temporal domain features of the corrected data, including the moving average deviation, trend fitting error, and periodic fluctuation amplitude of data at different times at the same node; S43, normalizing the spatial domain features and temporal domain features, fusing the features according to preset weight coefficients, and generating a dual-domain fused feature vector. S44 inputs the dual-domain fusion feature vector into the spatiotemporal dual-domain data quality discrimination model, outputs a data quality quantification score, and classifies the quality level according to the score.

9. The method for quality assessment of multi-source monitoring data of energy-consuming equipment according to claim 1, characterized in that, S5 includes the following steps: S51, using a feature selection algorithm to screen out common parameters that significantly affect data quality from the evaluation index system and construct a common parameter set; S52, analyzing the operating condition parameters of energy-consuming equipment and monitoring environment parameters, establishing a localized parameter requirement model, and determining the type and value range of the adaptable parameters; S53, based on the common parameter set and the localized parameter requirement model, using a parameter mapping algorithm to convert common parameters into localized parameters and generate initial adaptable parameters; S54, adjusting the values ​​of the adaptable parameters and optimizing the parameter weight allocation ratio by real-time monitoring of data quality feedback results, forming a dynamic adaptation mechanism.

10. A method for quality assessment of multi-source monitoring data of energy-consuming equipment according to any one of claims 1-9, characterized in that, This method is implemented through a quality assessment system for multi-source monitoring data of energy-consuming equipment, comprising: a multi-source data acquisition and transmission unit, a gradient clustering anomaly identification unit, a spatiotemporal coupling data correction unit, a dual-domain quality discriminant analysis unit, a parameter collaborative adaptation optimization unit, and a quality assessment result output unit. The multi-source data acquisition and transmission unit collects multi-dimensional monitoring data through a distributed sensor array and an industrial bus interface, and sends it to the gradient clustering anomaly identification unit via an encrypted transmission protocol. After receiving the data, the gradient clustering anomaly identification unit performs anomaly data filtering and transmits the filtered data to the spatiotemporal coupling data correction unit. After correcting data deviations, the spatiotemporal coupling data correction unit sends the corrected data to the dual-domain quality discriminant analysis unit. The dual-domain quality discriminant analysis unit combines spatial and temporal domain features to output quality quantification results, which are simultaneously sent to the parameter collaborative adaptation optimization unit. The parameter collaborative adaptation optimization unit adjusts parameter configurations based on the quality quantification results, feeds back the optimized parameters to the preceding units, and simultaneously sends the final quality assessment result and parameter configuration information to the quality assessment result output unit. The quality assessment result output unit generates a quality assessment report and data quality visualization charts in a preset format, displaying the results in multiple dimensions.

Images