A water and electricity equipment defect data credibility weighted fusion method fusing working condition adaptability and multi-source consistency

Abstract

The application discloses a water and electricity equipment defect data credibility weighted fusion method fusing working condition adaptability and multi-source consistency, and aims at defects that existing technologies do not consider dynamic changes of working conditions, lack cross-system physical verification, have weak conflict resolution capability and fusion results have no credibility marking and the like. Six function modules are constructed, i.e. a multi-source data access and alignment module, a data source dynamic credibility modeling module, a mechanical-electrical-hydraulic cross-system cross verification module, a multi-source conflict identification and intelligent resolution module, a credibility weighted fusion module and a fusion quality evaluation and output module. The application adjusts data source weights in real time through a four-dimensional dynamic credibility model, cross checks by using mechanical-electrical-hydraulic physical coupling relations, intelligently resolves four types of conflicts, i.e. numerical, state, existence and time sequence, and outputs a unified defect state vector carrying fusion confidence and conflict traces.

Description

Technical Field

[0001] This invention relates to the field of multi-source data processing technology, specifically to a weighted fusion method for the reliability of hydropower equipment defect data that integrates working condition adaptability and multi-source consistency. Background Technology

[0002] Equipment status data in large hydropower stations originates from multiple heterogeneous data sources, including process parameter acquisition from distributed control systems (DCS / SCADA), specialized monitoring data such as vibration, temperature, and oil pressure from online monitoring systems, shift inspection records, defect work orders, periodic maintenance reports, and protection action records. These data sources differ significantly in terms of acquisition frequency, data type, recording granularity, time accuracy, and reliability. For example, SCADA systems automatically acquire numerical data at a frequency of seconds or minutes, while inspection records document subjective judgments in free text format on a daily basis.

[0003] In hydropower station scenarios, conflicts between multi-source data are particularly prominent. First, numerical conflicts: different sensors provide different readings for the same measuring point, possibly due to sensor drift, range differences, or different installation locations. Second, status conflicts: SCADA shows a device parameter within the normal range, but inspection records indicate an anomaly. Third, existence conflicts: one data source records a defect event, but other data sources have no related records for the same period. Fourth, timing conflicts: different data sources show significant discrepancies in the timestamps of the same event, such as work order records lagging behind the actual occurrence time.

[0004] More importantly, the reliability of multi-source data from hydropower units is strongly dependent on operating conditions. Data collected by the same SCADA sensor under steady-state operating conditions is generally highly reliable, but its reliability may decrease under start-up and shutdown transition conditions due to large signal fluctuations. Manual inspections are relatively brief during normal operation, but are often more detailed and reliable when equipment malfunctions. In addition, there are physical coupling relationships between the mechanical, electrical, and hydraulic systems of hydropower units. This coupling relationship provides the possibility of cross-system verification. For example, the guide vane opening of the speed control system should be physically consistent with the active power of the generator.

[0005] In existing technologies, patent CN121659206A provides a method for fusing multi-source heterogeneous energy data. This patent quantifies the credibility of data sources in real time from four dimensions: historical accuracy, instantaneous stability, temporal completeness, and conformity to physical laws. It then filters low-credit data and weights and fuses high-credit data. The fused data is used for digital twins and cross-domain analysis, and the results are used to correct the credit model. It is mainly used in multi-source heterogeneous data fusion scenarios of integrated energy systems to solve problems such as energy data conflicts, static weights, lack of physical constraints, and lack of traceability, thereby improving the reliability of data fusion.

[0006] Patent CN121350619A provides a multimodal power equipment defect sample and cross-validation evaluation system. This patent generates a multimodal defect sample library using Generative Adversarial Networks (GANs) and physical constraints; then, it decouples the samples using a variational autoencoder, fuses them with graph attention, and finally completes cross-modal feature fusion by separating identity-related features and defect-related features. This system primarily addresses the problems of scarce power equipment defect samples, modal feature confusion, and low evaluation reliability.

[0007] However, the existing technology has the following drawbacks: CN121350619A only addresses defect sample generation, modal feature fusion, and model evaluation, without considering conflict fusion of multi-source heterogeneous original defect data (DCS, inspection, work orders), thus failing to meet the actual data processing needs of hydropower operation and maintenance. It directly performs feature fusion without dynamic data source reliability assessment, resulting in inconsistent hydropower data quality, lack of pre-filtering of low-quality data, and an unreliable fusion foundation. Furthermore, it fails to address multi-source data conflict identification, arbitration, and resolution, leading to frequent hydropower defect data conflicts—a core pain point completely unaddressed. CN121659206A only covers general energy scenarios, without designing dynamic reliability for differences in hydropower start-up / shutdown, steady-state, and transitional operating conditions. Hydropower operating conditions have a significant impact on data quality, making it unsuitable for hydropower scenarios.

[0008] Therefore, a weighted fusion method for the reliability of hydropower equipment defect data is needed that can integrate dynamic reliability assessment with multi-source consistency, support cross-system cross-validation and intelligent conflict resolution, and provide a high-quality and reliable unified data foundation for subsequent heterogeneous graph construction and defect coupling analysis. Summary of the Invention

[0009] To address the aforementioned issues, this invention proposes a weighted fusion method for the reliability of hydropower equipment defect data that integrates operational condition adaptability and multi-source consistency. The method includes the following steps: S1. Multi-source defect data access and alignment The purpose of step S1 is to align defect data from different data sources according to a unified device identifier and time window, forming an aligned dataset of "same device - same time period - multi-source observation", specifically including: S1.1 Access the raw data from each data source and label the source system identifier for each data entry. Data sources include at least DCS / SCADA, online monitoring systems, inspection records, defect work orders, maintenance reports, and protection action records.

[0010] S1.2 Use a unified defect code to align data from different data sources according to device identifiers, so that observations of the same device or component from different sources can be associated with the same entity.

[0011] S1.3 sets a time alignment window to align data from different acquisition frequencies to a uniform time granularity. During time alignment, the time deviation between each source data and the alignment time window is recorded and used as input for the time synchronization evaluation in S2.

[0012] S1.4 connects to the current operating condition layer label, which serves as the input for the S2 operating condition adaptability assessment.

[0013] S2. Dynamic Reliability Modeling of Data Sources The purpose of step S2 is to calculate a dynamic reliability score for each data source within each time window. This score changes dynamically with the operating conditions, specifically including: S2.1 Assess the health of the data source H(s) j Health status reflects the operational status of the data source itself, including indicators such as sensor drift, communication interruption rate, and historical false alarm rate. For automated data acquisition systems, it can be assessed through sensor self-test information; for manually recorded systems, it can be assessed through the completeness and standardization of records.

[0014] S2.2 Evaluate the time synchronization degree T(s) j Time synchronization reflects the degree of deviation between the recording time of the data source and the actual event time. Real-time acquisition systems have high time synchronization, while manual work orders have lower time synchronization.

[0015] S2.3 Assess the adaptability of operating conditions A(s) j ,g t Operating condition adaptability reflects the reliability of the data source under current operating conditions. For example, vibration sensors become less reliable during start-up and shutdown transitions due to high background vibration; SCADA temperature acquisition is reliable during steady-state operation but lags behind actual temperature changes during rapid load changes. Operating condition adaptability can be pre-configured by hydropower experts based on operational experience or obtained from historical data.

[0016] S2.4 Assessing Multi-Source Consistency C(s) j Multi-source consistency reflects the degree to which a data source matches the descriptions of the same event from other data sources. A data source has a high consistency score when it is consistent with a majority of other sources; a data source has a low consistency score when it contradicts other sources.

[0017] The dynamic credibility score of a data source can be defined as: ; Among them, W(s) j ,g t ) indicates data source s j Under operating condition g t Dynamic credibility score; H(s) j T(s) represents the health of the data source;j ) represents the degree of time synchronization; A(s) j ,g t C(s) represents the adaptability to operating conditions; j ) indicates multi-source consistency; For each dimension's weight coefficients, satisfying The above formula demonstrates that the reliability of the same data source can vary under different operating conditions, reflecting the condition-dependent nature of hydropower scenarios.

[0018] S3. Cross-system verification of electromechanical and hydraulic systems The purpose of step S3 is to utilize the physical coupling relationship between the mechanical, electrical, and hydraulic systems of the hydropower unit to perform cross-system consistency checks on multi-source data and identify abnormal data that violates physical laws. Specifically, this includes: S3.1 Establish a cross-system physical constraint rule set for the hydropower unit. These rules describe the physical consistency relationships that parameters of different systems should satisfy. For example: the guide vane opening of the speed regulating system should be positively correlated with the generator's active power; the cooling system flow rate should be negatively correlated with the bearing temperature; and the excitation current and generator terminal voltage should conform to the excitation characteristic curve. These constraint rules can be configured by hydropower experts based on the unit's design parameters and operating experience.

[0019] S3.2 Performs cross-system consistency checks on the multi-source aligned data. Relevant parameters from different systems are compared according to constraint rules to identify data points that violate physical constraints. For example, if the guide vane opening increases but the active power does not increase accordingly, there may be an anomaly in one of the data sources.

[0020] S3.3 imposes credibility penalties on data sources that violate cross-system physical constraints. The cross-system verification results are used as a credibility correction factor to reduce the weight of data sources that violate physical constraints in subsequent fusion processes.

[0021] The reliability of the correction after cross-system cross-validation can be defined as: ; in, W(s) represents the confidence level after cross-system cross-validation correction; j ,g t V(s) represents the original dynamic credibility; j V represents the cross-system validation correction factor. V=1 when the data source passes the cross-system validation, and V<1 when the data source violates the cross-system physical constraints. The degree of penalty is proportional to the degree of violation.

[0022] S4. Multi-source conflict identification and intelligent resolution The purpose of step S4 is to perform typological identification of conflicts in multi-source aligned data and to automatically learn a conflict arbitration strategy using a learning algorithm, specifically including: S4.1 performs conflict detection on multi-source aligned data and classifies conflicts into four categories: numerical conflict (different sources give different values ​​for the same measurement point), state conflict (one source judges it as normal while another source judges it as abnormal), existence conflict (one source records the event while other sources do not record it), and timing conflict (different sources have inconsistent timestamps for the same event).

[0023] S4.2 Establish a conflict arbitration learning model. This model takes conflict type, source credibility scores, operational condition labels, cross-system verification results, and conflict context information as input, and historically verified conflict resolution results as supervision signals. It automatically learns conflict arbitration strategies through a learning algorithm. The learning algorithm can employ classification models, ranking learning models, or attention-based fusion networks, etc.

[0024] For scenarios with insufficient historical data, S4.3 provides a rule-based default arbitration strategy as a baseline. Default rules can be configured by hydropower experts, such as "the reliability of protection action records is always higher than that of manual inspection records" and "when most automatic data collection sources are consistent, the reliability of a few inconsistent sources is reduced."

[0025] S4.4 As the system runs, new human confirmation results can be fed back as training samples to continuously optimize the accuracy of the conflict arbitration learning model.

[0026] The conflict arbitration process can be formalized as follows: ; in, This indicates the outcome of the conflict arbitration, i.e., which source of data was adopted or how it was synthesized; F arb This represents the conflict arbitration function, which can be implemented by a learning algorithm or by rules; type represents the conflict type. Indicates the corrected confidence level of each source; V cross This indicates the cross-system verification result; ctx represents the conflict context information.

[0027] S5. Credibility-Weighted Fusion The purpose of step S5 is to weight and fuse multi-source data into a unified defect state vector based on the dynamic credibility weights of each data source. Specifically, this includes: S5.1 For conflict-free multi-source data, weighted fusion is performed using the corrected credibility of each source as the weight. Sources with higher credibility contribute more, while sources with lower credibility contribute less.

[0028] S5.2 For data that has already undergone conflict resolution in S4, fusion is performed according to the arbitration result. If the arbitration result is to adopt a certain source, then the data from that source is directly used; if the arbitration result is to weight data by confidence, then fusion is performed using the weights adjusted by the arbitration.

[0029] S5.3 For data points with only a single data source, the data from that source is used directly, and the fusion confidence level is equal to the confidence level of that source.

[0030] The weighted fusion process can be defined as: ; in, ; Represents the unified state value after fusion; x j This represents the observation value from the j-th data source; This represents the normalized fusion weights; This represents the corrected dynamic credibility; m represents the number of data sources participating in the fusion.

[0031] S6. Integrating Quality Assessment and Standardized Output The purpose of step S6 is to calculate the fusion confidence score for each fused data point and output a unified defect state vector in a standardized format, specifically including: S6.1 calculates the fusion confidence score using a multi-factor weighted product model. The fusion confidence score comprehensively reflects the number of data sources involved in the fusion, the dynamic confidence level of each source, the degree of consistency among multiple sources, and whether there are any resolved conflicts.

[0032] The quantitative calculation formula for fusion confidence (conf) is as follows: ; in, The arithmetic mean of the corrected credibility of all data sources participating in the fusion is calculated using the following formula: ; in, The total number of data sources participating in the integration. For the first The credibility of the data source after cross-system verification.

[0033] The number of data sources is a factor reflecting the reliability of multi-source validation, calculated using a marginal decrease formula: , It is the adjustment coefficient; As a multi-source consistency factor, the numerical defect calculation method is as follows: ( The standard deviation of the observed values. If the mean is... Then take 1), the calculation method for state-type defects is as follows: ( (Number of data sources to support the i-th state). As a conflict correction factor, when there is no conflict Minor conflict when the conflict has been resolved Moderate conflict Severe conflict When the conflict is not resolved .

[0034] The fusion confidence level is divided into 5 standardized levels: very high confidence (conf∈[0.9,1.0]), high confidence (conf∈[0.7,0.9)), medium confidence (conf∈[0.5,0.7)), low confidence (conf∈[0.3,0.5)), and very low confidence (conf∈[0,0.3)).

[0035] S6.2 Retain conflict traces. For data points that have conflicts and have undergone arbitration, record the conflict type, the data sources involved in the conflict, the arbitration result, and the arbitration basis as traceable information of the fusion result.

[0036] S6.3 outputs standardized unified defect status vectors. Each status vector contains at least the device identifier, time window, unified status value after fusion, fusion confidence level, set of data sources participating in the fusion, and conflict traces.

[0037] The standardized output result can be defined as: ; Among them, O i This represents the i-th standardized output result; dev i Indicates device identifier; t_win i Indicates a time window; x i Represents the unified state value after merging; conf i Indicates the fusion confidence level; sources i Indicates the set of data sources participating in the fusion; conflict_trace i This indicates conflict trace information.

[0038] Based on the above method, a weighted fusion system for reliability of defect data of hydropower equipment that integrates working condition adaptability and multi-source consistency is provided, including: a multi-source data access and alignment module, a data source dynamic reliability modeling module, a cross-validation module for electromechanical and hydraulic systems, a multi-source conflict identification and intelligent resolution module, a reliability weighted fusion module, and a fusion quality assessment and output module.

[0039] The multi-source data access and alignment module is responsible for the unified access, standardized coding, time dimension alignment and field unification of all heterogeneous defect data of hydropower equipment, and constructs a regular dataset of "same equipment - same time period - multi-source observation", laying the data foundation for subsequent processing.

[0040] The dynamic credibility modeling module for data sources calculates the dynamic credibility score of each data source in real time under the current operating conditions from four dimensions: data quality, timeliness, consistency, and source reliability, so as to realize the credibility adaptive adjustment according to the operating conditions.

[0041] The electromechanical-hydraulic cross-system cross-verification module is based on the physical coupling constraint rules of the mechanical, electrical and hydraulic systems of the hydropower unit. It performs cross-system consistency verification on the aligned data, identifies abnormal data and corrects the credibility of the data source, and strengthens the physical rationality of the data.

[0042] The multi-source conflict identification and intelligent resolution module automatically identifies four types of conflicts among multi-source data: numerical, state, existence, and time sequence. It integrates learning algorithms and expert rules to achieve intelligent arbitration of conflicts, supports historical data feedback to optimize the arbitration model, and outputs conflict resolution results.

[0043] The credibility weighted fusion module performs weighted fusion on multi-source data based on the corrected dynamic credibility weight and the conflict resolution results, generating a unified defect status value and distinguishing between three types of scenarios: no conflict, conflict resolved, and single-source data, to adapt the fusion logic.

[0044] The fusion quality assessment and output module quantifies the quality of the fused data, calculates the fusion confidence level, retains conflict tracing traces, and standardizes the output of a unified defect state vector carrying a confidence level label to support subsequent defect coupling graph construction and graph learning.

[0045] A computer electronic device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a weighted fusion method for reliability of hydropower equipment defect data with fusion working condition adaptability and multi-source consistency.

[0046] A computer-readable storage medium storing a computer program / instructions, which, when executed by a processor, implements the steps of a weighted fusion method for reliability of hydropower equipment defect data with fusion operating condition adaptability and multi-source consistency.

[0047] Compared with the prior art, the beneficial effects of the present invention include: (1) This invention integrates a dynamic credibility scoring model that combines four dimensions: data source health, time synchronization, working condition adaptability, and multi-source consistency. This model can reflect the characteristics of hydropower scenarios where the credibility of the same data source varies under different working conditions.

[0048] (2) This invention utilizes the physical coupling relationship between the mechanical, electrical and hydraulic systems of a hydropower unit to perform cross-system verification, which can identify abnormal data that violate physical laws. This is a unique verification method in the field of hydropower, which cannot be achieved by ordinary multi-source fusion methods.

[0049] (3) The present invention uses a learning algorithm to automatically learn the conflict arbitration strategy and supports the optimization of historical confirmation results. As the system uses the accuracy of conflict resolution, it continues to improve.

[0050] (4) The fusion results of the present invention carry fusion confidence labels and conflict trace information, providing a quantitative basis for the reliability of data for subsequent graph construction and graph learning, so that subsequent analysis can treat data with different confidence levels differently. Attached Figure Description

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a system architecture diagram of the method of the present invention; Figure 3 This is an example diagram illustrating the multi-source conflict identification and resolution method of the present invention. Detailed Implementation

[0053] The present invention discloses a weighted fusion method for the reliability of hydropower equipment defect data that integrates operational condition adaptability and multi-source consistency. The overall flowchart is shown below. Figure 1 As shown, the specific implementation is as follows: Example 1 A weighted fusion method for assessing the reliability of defect data in hydropower equipment by integrating operational condition adaptability and multi-source consistency, the method comprising the following steps: S1. Multi-source defect data access and alignment The purpose of step S1 is to align defect data from different data sources according to a unified device identifier and time window, forming an aligned dataset of "same device - same time period - multi-source observation", specifically including: S1.1 Access the raw data from each data source and label the source system identifier for each data entry. Data sources include at least DCS / SCADA, online monitoring systems, inspection records, defect work orders, maintenance reports, and protection action records.

[0054] S1.2 Use a unified defect code to align data from different data sources according to device identifiers, so that observations of the same device or component from different sources can be associated with the same entity.

[0055] S1.3 sets a time alignment window to align data from different acquisition frequencies to a uniform time granularity. During time alignment, the time deviation between each source data and the alignment time window is recorded and used as input for the time synchronization evaluation in S2.

[0056] S1.4 connects to the current operating condition layer label, which serves as the input for the S2 operating condition adaptability assessment.

[0057] S2. Dynamic Reliability Modeling of Data Sources The purpose of step S2 is to calculate a dynamic reliability score for each data source within each time window. This score changes dynamically with the operating conditions, specifically including: S2.1 Assess the health of the data source H(s) j Health status reflects the operational status of the data source itself, including indicators such as sensor drift, communication interruption rate, and historical false alarm rate. For automated data acquisition systems, it can be assessed through sensor self-test information; for manually recorded systems, it can be assessed through the completeness and standardization of records.

[0058] S2.2 Evaluate the time synchronization degree T(s) j Time synchronization reflects the degree of deviation between the recording time of the data source and the actual event time. Real-time acquisition systems have high time synchronization, while manual work orders have lower time synchronization.

[0059] S2.3 Assess the adaptability of operating conditions A(s) j ,g t Operating condition adaptability reflects the reliability of the data source under current operating conditions. For example, vibration sensors become less reliable during start-up and shutdown transitions due to high background vibration; SCADA temperature acquisition is reliable during steady-state operation but lags behind actual temperature changes during rapid load changes. Operating condition adaptability can be pre-configured by hydropower experts based on operational experience or obtained from historical data.

[0060] S2.4 Assessing Multi-Source Consistency C(s) j Multi-source consistency reflects the degree to which a data source matches the descriptions of the same event from other data sources. A data source has a high consistency score when it is consistent with a majority of other sources; a data source has a low consistency score when it contradicts other sources.

[0061] The dynamic credibility score of a data source can be defined as: ; Among them, W(s) j ,g t ) indicates data source s j Under operating condition g t Dynamic credibility score; H(s) j T(s) represents the health of the data source; j ) represents the degree of time synchronization; A(s) j ,g t C(s) represents the adaptability to operating conditions; j ) indicates multi-source consistency; For each dimension's weight coefficients, satisfying The above formula demonstrates that the reliability of the same data source can vary under different operating conditions, reflecting the condition-dependent nature of hydropower scenarios.

[0062] S3. Cross-system verification of electromechanical and hydraulic systems The purpose of step S3 is to utilize the physical coupling relationship between the mechanical, electrical, and hydraulic systems of the hydropower unit to perform cross-system consistency checks on multi-source data and identify abnormal data that violates physical laws. Specifically, this includes: S3.1 Establish a cross-system physical constraint rule set for the hydropower unit. These rules describe the physical consistency relationships that parameters of different systems should satisfy. For example: the guide vane opening of the speed regulating system should be positively correlated with the generator's active power; the cooling system flow rate should be negatively correlated with the bearing temperature; and the excitation current and generator terminal voltage should conform to the excitation characteristic curve. These constraint rules can be configured by hydropower experts based on the unit's design parameters and operating experience.

[0063] S3.2 Performs cross-system consistency checks on the multi-source aligned data. Relevant parameters from different systems are compared according to constraint rules to identify data points that violate physical constraints. For example, if the guide vane opening increases but the active power does not increase accordingly, there may be an anomaly in one of the data sources.

[0064] S3.3 imposes credibility penalties on data sources that violate cross-system physical constraints. The cross-system verification results are used as a credibility correction factor to reduce the weight of data sources that violate physical constraints in subsequent fusion processes.

[0065] The reliability of the correction after cross-system cross-validation can be defined as: ; in, W(s) represents the confidence level after cross-system cross-validation correction; j ,g t V(s) represents the original dynamic credibility; j V represents the cross-system validation correction factor. V=1 when the data source passes the cross-system validation, and V<1 when the data source violates the cross-system physical constraints. The degree of penalty is proportional to the degree of violation.

[0066] S4. Multi-source conflict identification and intelligent resolution The purpose of step S4 is to perform typological identification of conflicts in multi-source aligned data and to automatically learn a conflict arbitration strategy using a learning algorithm, specifically including: S4.1 performs conflict detection on multi-source aligned data and classifies conflicts into four categories: numerical conflict (different sources give different values ​​for the same measurement point), state conflict (one source judges it as normal while another source judges it as abnormal), existence conflict (one source records the event while other sources do not record it), and timing conflict (different sources have inconsistent timestamps for the same event).

[0067] S4.2 Establish a conflict arbitration learning model. This model takes conflict type, source credibility scores, operational condition labels, cross-system verification results, and conflict context information as input, and historically verified conflict resolution results as supervision signals. It automatically learns conflict arbitration strategies through a learning algorithm. The learning algorithm can employ classification models, ranking learning models, or attention-based fusion networks, etc.

[0068] For scenarios with insufficient historical data, S4.3 provides a rule-based default arbitration strategy as a baseline. Default rules can be configured by hydropower experts, such as "the reliability of protection action records is always higher than that of manual inspection records" and "when most automatic data collection sources are consistent, the reliability of a few inconsistent sources is reduced."

[0069] S4.4 As the system runs, new human confirmation results can be fed back as training samples to continuously optimize the accuracy of the conflict arbitration learning model.

[0070] The conflict arbitration process can be formalized as follows: ; in, This indicates the outcome of the conflict arbitration, i.e., which source of data was adopted or how it was synthesized; F arb This represents the conflict arbitration function, which can be implemented by a learning algorithm or by rules; type represents the conflict type. Indicates the corrected confidence level of each source; V cross This indicates the cross-system verification result; ctx represents the conflict context information.

[0071] S5. Credibility-Weighted Fusion The purpose of step S5 is to weight and fuse multi-source data into a unified defect state vector based on the dynamic credibility weights of each data source. Specifically, this includes: S5.1 For conflict-free multi-source data, weighted fusion is performed using the corrected credibility of each source as the weight. Sources with higher credibility contribute more, while sources with lower credibility contribute less.

[0072] S5.2 For data that has already undergone conflict resolution in S4, fusion is performed according to the arbitration result. If the arbitration result is to adopt a certain source, then the data from that source is directly used; if the arbitration result is to weight data by confidence, then fusion is performed using the weights adjusted by the arbitration.

[0073] S5.3 For data points with only a single data source, the data from that source is used directly, and the fusion confidence level is equal to the confidence level of that source.

[0074] The weighted fusion process can be defined as: ; in, ; Represents the unified state value after fusion; x j This represents the observation value from the j-th data source; This represents the normalized fusion weights; This represents the corrected dynamic credibility; m represents the number of data sources participating in the fusion.

[0075] S6. Integrating Quality Assessment and Standardized Output The purpose of step S6 is to calculate the fusion confidence score for each fused data point and output a unified defect state vector in a standardized format, specifically including: S6.1 calculates the fusion confidence score using a multi-factor weighted product model. The fusion confidence score comprehensively reflects the number of data sources involved in the fusion, the dynamic confidence level of each source, the degree of consistency among multiple sources, and whether there are any resolved conflicts.

[0076] The quantitative calculation formula for fusion confidence (conf) is as follows: in, The arithmetic mean of the corrected credibility of all data sources participating in the fusion is calculated using the following formula: ; in, The total number of data sources participating in the integration. For the first The credibility of the data source after cross-system verification.

[0077] The number of data sources is a factor reflecting the reliability of multi-source validation, calculated using a marginal decrease formula: , (Adjustment coefficient), m=1≈0.39, m=2≈0.63, m=3≈0.78, m≥5≈0.92; As a multi-source consistency factor, the numerical defect calculation method is as follows: ( The standard deviation of the observed values. If the mean is... Then take 1), the calculation method for state-type defects is as follows: ( (Number of data sources to support the i-th state). As a conflict correction factor, when there is no conflict Minor conflict when the conflict has been resolved Moderate conflict Severe conflict When the conflict is not resolved (Manual intervention required).

[0078] The fusion confidence level is divided into 5 standardized levels: very high confidence (conf∈[0.9,1.0]), high confidence (conf∈[0.7,0.9)), medium confidence (conf∈[0.5,0.7)), low confidence (conf∈[0.3,0.5)), and very low confidence (conf∈[0,0.3)).

[0079] S6.2 Retain conflict traces. For data points that have conflicts and have undergone arbitration, record the conflict type, the data sources involved in the conflict, the arbitration result, and the arbitration basis as traceable information of the fusion result.

[0080] S6.3 outputs standardized unified defect status vectors. Each status vector contains at least the device identifier, time window, unified status value after fusion, fusion confidence level, set of data sources participating in the fusion, and conflict traces.

[0081] The standardized output result can be defined as: ; Among them, O i This represents the i-th standardized output result; dev i Indicates device identifier; t_win i Indicates a time window; x i Represents the unified state value after merging; conf i Indicates the fusion confidence level; sources i Indicates the set of data sources participating in the fusion; conflict_trace i This indicates conflict trace information.

[0082] Example 2 Based on Example 1, such as Figure 2 As shown, a weighted fusion system for reliability of defect data of hydropower equipment that integrates working condition adaptability and multi-source consistency is provided. The system includes: a multi-source data access and alignment module, a data source dynamic reliability modeling module, a cross-validation module for electromechanical and hydraulic systems, a multi-source conflict identification and intelligent resolution module, a reliability weighted fusion module, and a fusion quality assessment and output module.

[0083] The multi-source data access and alignment module is responsible for the unified access, standardized coding, time dimension alignment and field unification of all heterogeneous defect data of hydropower equipment, and constructs a regular dataset of "same equipment - same time period - multi-source observation", laying the data foundation for subsequent processing.

[0084] The dynamic credibility modeling module for data sources calculates the dynamic credibility score of each data source in real time under the current operating conditions from four dimensions: data quality, timeliness, consistency, and source reliability, so as to realize the credibility adaptive adjustment according to the operating conditions.

[0085] The electromechanical-hydraulic cross-system cross-verification module is based on the physical coupling constraint rules of the mechanical, electrical and hydraulic systems of the hydropower unit. It performs cross-system consistency verification on the aligned data, identifies abnormal data and corrects the credibility of the data source, and strengthens the physical rationality of the data.

[0086] The multi-source conflict identification and intelligent resolution module automatically identifies four types of conflicts among multi-source data: numerical, state, existence, and time sequence. It integrates learning algorithms and expert rules to achieve intelligent arbitration of conflicts, supports historical data feedback to optimize the arbitration model, and outputs conflict resolution results.

[0087] The credibility weighted fusion module performs weighted fusion on multi-source data based on the corrected dynamic credibility weight and the conflict resolution results, generating a unified defect status value and distinguishing between three types of scenarios: no conflict, conflict resolved, and single-source data, to adapt the fusion logic.

[0088] The fusion quality assessment and output module quantifies the quality of the fused data, calculates the fusion confidence level, retains conflict tracing traces, and standardizes the output of a unified defect state vector carrying a confidence level label to support subsequent defect coupling graph construction and graph learning.

[0089] The embodiments described above are merely illustrative of implementation methods of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A weighted fusion method for the reliability of defect data in hydropower equipment that integrates operational condition adaptability and multi-source consistency, characterized in that, Includes the following steps: S1. Multi-source defect data access and alignment; S2. Dynamic reliability modeling of data sources; S3. Cross-system verification of electromechanical and hydraulic systems; S4. Multi-source conflict identification and intelligent resolution; S5. Credibility-weighted fusion; S6. Integrate quality assessment and standardized output.

2. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S1 specifically includes: S1.1 Access the raw data from each data source and label the source system identifier for each data entry. Data sources must include at least DCS / SCADA, online monitoring systems, inspection records, defect work orders, maintenance reports, and protection action records. S1.2 Using a unified defect code, data from each data source is aligned according to the device identifier, so that observations of the same device or component from different sources can be associated with the same entity; S1.3 sets a time alignment window to align data from different acquisition frequencies to a uniform time granularity. During time alignment, the time deviation between each source data and the alignment time window is recorded and used as input for the time synchronization evaluation in S2. S1.4 connects to the current operating condition layer label, which serves as the input for the S2 operating condition adaptability assessment.

3. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S2 specifically includes: S2.1 Assess the health of the data source H(s) j ); S2.2 Evaluate the time synchronization degree T(s) j ); S2.3 Assess the adaptability of operating conditions A(s) j ,g t ); S2.4 Assessing Multi-Source Consistency C(s) j ); The dynamic credibility score of a data source can be defined as: ； Among them, W(s) j ,g t ) indicates data source s j Under operating condition g t Dynamic credibility score; H(s) j T(s) represents the health of the data source; j A(s) represents the degree of time synchronization; j ,g t C(s) represents the adaptability to operating conditions; j ) indicates multi-source consistency; This represents the comprehensive confidence calculation function.

4. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S3 specifically includes: S3.1 Establish a cross-system physical constraint rule set for hydropower units; S3.2 Perform cross-system consistency checks on the data after multi-source alignment; S3.3 imposes credibility penalties on data sources that violate cross-system physical constraints.

5. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 4, characterized in that, The credibility penalty specifically involves using cross-system test results as a credibility correction factor to reduce the weight of data sources that violate physical constraints in subsequent fusion. The corrected confidence level after cross-system cross-validation is defined as: ； in, W(s) represents the confidence level after cross-system cross-validation correction; j ,g t V(s) represents the original dynamic credibility; j V represents the cross-system validation correction factor. V=1 when the data source passes the cross-system validation, and V<1 when the data source violates the cross-system physical constraints. The degree of penalty is proportional to the degree of violation.

6. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S4 specifically includes: S4.1 performs conflict detection on multi-source aligned data and classifies conflicts into four categories: numerical conflicts, state conflicts, existence conflicts, and temporal conflicts. S4.2 Establish a conflict arbitration learning model; S4.3 provides a rule-based default arbitration strategy as a baseline for scenarios with insufficient historical data; S4.4 As the system runs, new human confirmation results are fed back as training samples to continuously optimize the accuracy of the conflict arbitration learning model.

7. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S5 specifically includes: S5.1 For conflict-free multi-source data, weighted fusion is performed using the corrected credibility of each source as the weight. Sources with higher credibility contribute more, while sources with lower credibility contribute less. S5.2 For data that has already undergone conflict resolution in S4, the fusion is performed according to the arbitration result. If the arbitration result is to adopt a certain source, then the data from that source is directly adopted; if the arbitration result is to weight the data by confidence, then the fusion is performed using the weights adjusted by the arbitration. S5.3 For data points with only a single data source, the data from that source is used directly, and the fusion confidence level is equal to the confidence level of that source.

8. The weighted fusion method for reliability of hydropower equipment defect data based on fusion operating condition adaptability and multi-source consistency as described in claim 1, characterized in that, Step S6 specifically includes: S6.1 Calculate the fusion confidence score; S6.2 Preserve conflict traces; S6.3 outputs standardized unified defect status vectors. Each status vector contains at least the device identifier, time window, unified status value after fusion, fusion confidence level, set of data sources participating in the fusion, and conflict traces.

9. A system for implementing the weighted fusion method for reliability of hydropower equipment defect data as described in any one of claims 1-8, characterized in that, include: The module includes a multi-source data access and alignment module, a data source dynamic credibility modeling module, a cross-system cross-validation module for electromechanical and hydraulic systems, a multi-source conflict identification and intelligent resolution module, a credibility weighted fusion module, and a fusion quality assessment and output module. The multi-source data access and alignment module is responsible for the unified access, standardized coding, time dimension alignment and field unification of all heterogeneous defect data of hydropower equipment, and constructs a regular dataset of "same equipment - same time period - multi-source observation". The dynamic credibility modeling module for the data source calculates the dynamic credibility score of each data source in real time under the current operating conditions from four dimensions: data quality, timeliness, consistency, and source reliability, so as to realize the credibility adaptive adjustment according to the operating conditions. The electromechanical-hydraulic cross-system cross-verification module is based on the physical coupling constraint rules of the mechanical, electrical and hydraulic systems of the hydropower unit. It performs cross-system consistency verification on the aligned data, identifies abnormal data and corrects the credibility of the data source, and strengthens the physical rationality of the data. The multi-source conflict identification and intelligent resolution module automatically identifies four types of conflicts between multi-source data: numerical, state, existence, and time sequence. It integrates learning algorithms and expert rules to achieve intelligent arbitration of conflicts, supports historical data backflow to optimize the arbitration model, and outputs conflict resolution results. The credibility weighted fusion module performs weighted fusion on multi-source data based on the corrected dynamic credibility weight and the conflict resolution results, generates a unified defect status value, and distinguishes between three types of scenarios: no conflict, conflict resolved, and single-source data to adapt the fusion logic. The fusion quality assessment and output module quantifies the quality of the fused data, calculates the fusion confidence level, retains conflict tracing traces, and standardizes the output to carry a unified defect state vector with confidence level labels.

10. A computer electronic device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-8.

Images