A method, apparatus, and computer device for analyzing abnormal hydropower conditions.
By performing residual analysis of transmission delay and orthogonal joint analysis of water energy anomaly data from hydropower systems, the accuracy problem of multi-factor coupled anomaly analysis in traditional methods is solved, and the reliability and completeness of anomaly location and maintenance arrangements are achieved.
Patent Information
- Application Number
- CN202611345517.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-01
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional methods for analyzing abnormal situations in hydropower are insufficient to comprehensively consider the temporal correlations and coupling effects among multiple influencing factors, resulting in a lack of accurate basis for anomaly location, operational adjustments, and maintenance arrangements under complex abnormal conditions.
By performing a residual analysis of the transmission delay based on the normal water energy input-electricity output conversion relationship on the abnormal water energy data of the hydropower system, the abnormal residuals are reduced, and a time-delay aligned orthogonal joint analysis is performed to construct a complete influence path and obtain the target abnormal analysis data.
It improves the accuracy of abnormal residual data, reduces the interference of multi-factor coupled anomaly analysis, ensures the reliability, integrity and interpretability of analysis results, and provides an accurate basis for anomaly location, operation adjustment and maintenance arrangement.
Smart Images

Figure CN123116084A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data analysis technology, and in particular to a method, apparatus and computer equipment for analyzing abnormal hydropower conditions. Background Technology
[0002] In traditional technologies, hydropower anomaly analysis typically involves collecting operational data such as water level, flow rate, unit power, speed, vibration, temperature, and electrical parameters. Real-time data is compared with preset thresholds, historical normal ranges, or standard operating curves to determine the presence of anomalies. Further statistical analysis and fault rule bases are used to identify the anomaly type and possible causes, thus generating hydropower anomaly analysis results. However, traditional technologies mostly focus on independent judgments of single parameters or fixed factors, making it difficult to comprehensively consider the temporal correlations and coupling effects between multiple influencing factors. This results in an inability to provide accurate basis for anomaly location, operational adjustments, and maintenance arrangements in complex anomaly situations. Summary of the Invention
[0003] Therefore, it is necessary to provide a hydropower anomaly analysis method, device, and computer equipment that can provide accurate basis for anomaly location, operation adjustment, and maintenance arrangement of hydropower systems, addressing the aforementioned technical problems.
[0004] Firstly, this application provides a method for analyzing abnormal hydropower conditions, including: A transfer delay residual analysis based on the normal water energy input-electricity output conversion relationship is performed on the abnormal water energy data of the hydropower system to obtain the current abnormal residual data; Based on the current abnormal residual data, an abnormal residual reduction analysis is performed on the preset set of hydropower influencing factors of the hydropower system to obtain the target hydropower influencing factors. A time-delay aligned orthogonal joint analysis is performed on the factor data corresponding to the target hydropower influencing factors and the current abnormal residual data to obtain joint anomaly analysis data. The joint anomaly analysis data is used as the current anomaly residual data. The process of performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current anomaly residual data is repeated until each of the target hydropower influencing factors forms a complete influence path from the hydropower input end to the power output end in the preset hydropower coupling topology, thus obtaining the target anomaly analysis data.
[0005] Secondly, this application also provides a device for analyzing abnormal hydropower conditions, comprising: The abnormal data analysis module is used to perform transmission delay residual analysis on abnormal water energy data of the hydropower system based on the normal water energy input-electricity output conversion relationship, and obtain the current abnormal residual data. The influencing factor identification module is used to perform abnormal residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current abnormal residual data, so as to obtain the target hydropower influencing factors. The anomaly joint analysis module is used to perform time-delay aligned orthogonal joint analysis on the factor data corresponding to the target hydropower influencing factors and the current anomaly residual data to obtain joint anomaly analysis data. The hydropower cycle analysis module is used to take the joint anomaly analysis data as the current anomaly residual data and return to execute the step of performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current anomaly residual data to obtain the target hydropower influencing factors, until each of the target hydropower influencing factors forms a complete influence path from the hydropower input end to the power output end in the preset hydropower coupling topology, and the target anomaly analysis data is obtained.
[0006] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of a method for analyzing abnormal hydropower conditions.
[0007] The aforementioned method, apparatus, and computer equipment for analyzing hydropower anomalies correct for transmission time delays in hydropower input and power output changes within abnormal hydropower data by utilizing the normal hydropower input-power output conversion relationship. This prevents time misalignments caused by normal hydraulic transmission and unit response from being mistakenly counted as anomalies, thereby improving the accuracy of current abnormal residual data. Furthermore, by comparing the reduction effect of each hydropower influencing factor in the preset set of hydropower influencing factors on the current abnormal residual data, the target hydropower influencing factor is determined. Factors with stable correspondences to remaining abnormal changes are prioritized, reducing interference from irrelevant factors, accidental synchronous fluctuations, and already analyzed factors. Using time-delay aligned orthogonal joint analysis of the factor data corresponding to the target hydropower influencing factor and the current abnormal residual data allows for the separation of overlapping factors at a unified time point, reducing the impact of multiple factors. The correlation between hydropower influencing factors leads to repeated attribution, and the current anomaly residual data is continuously updated with joint anomaly analysis data, allowing each influencing factor to participate in the analysis round by round according to its actual effect on unexplained anomalies. At the same time, the complete influence path formed by each target hydropower influencing factor in the preset hydropower coupling topology from the hydropower input end to the power output end is used as the cycle termination condition. This can constrain the data change relationship with the actual hydraulic transmission, mechanical conversion and power output relationship of the hydropower generation system, avoiding premature termination based solely on residual thresholds or the formation of disconnected factor combinations. Finally, the target anomaly analysis data with time consistency, factor independence and physical connectivity is obtained, which helps to ensure the reliability, completeness and interpretability of multi-factor coupled anomaly analysis, and can provide an accurate basis for anomaly location, operation adjustment and maintenance arrangement of hydropower generation system. Attached Figure Description
[0008] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0009] Figure 1 This is an application environment diagram of the hydropower anomaly analysis method in one embodiment; Figure 2 This is a flowchart illustrating a method for analyzing abnormal hydropower situations in one embodiment; Figure 3 This is a structural block diagram of a hydropower anomaly analysis device in one embodiment; Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0010] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0011] The method for analyzing abnormal hydropower conditions provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.
[0012] In one exemplary embodiment, such as Figure 2 As shown, a method for analyzing abnormal hydropower conditions is provided, and this method is applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 208. Wherein:
[0013] Step 202: Perform a transfer delay residual analysis on the abnormal water energy data of the hydropower generation system based on the normal water energy input-electricity output conversion relationship to obtain the current abnormal residual data.
[0014] Step 204: Based on the current abnormal residual data, perform abnormal residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system to obtain the target hydropower influencing factors.
[0015] Step 206: Perform time-delay aligned orthogonal joint analysis on the factor data corresponding to the target hydropower influencing factors and the current abnormal residual data to obtain joint anomaly analysis data.
[0016] Step 208: Take the joint anomaly analysis data as the current anomaly residual data, return to execute the step of performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current anomaly residual data, and obtain the target hydropower influencing factors, until each target hydropower influencing factor forms a complete influence path from the hydropower input end to the power output end in the preset hydropower coupling topology, and obtain the target anomaly analysis data.
[0017] The hydropower generation system is an operating system that consists of a water energy input terminal, hydraulic water conveyance facilities, water turbine, generator, regulating device and power output terminal, and completes the conversion of water energy into electrical energy.
[0018] Among them, abnormal water energy data refers to flow rate, head, pressure, power, and related time-series data that deviate from the normal water energy input and power output states during the operation of a hydropower system.
[0019] The normal hydropower input-electrical output conversion relationship is the correspondence between the hydropower input quantities such as water flow and head after energy conversion and transmission time delay and the electrical output quantities such as power generation during the normal operation of the hydropower generation system.
[0020] Among them, the transmission delay residual analysis is a process of moving the water energy input data according to the time required for water energy to be transmitted to electrical energy output, and calculating the difference between the electrical energy output corresponding to the normal conversion relationship and the actual electrical energy output.
[0021] Among them, the current abnormal residual data is the sequence of water energy input-electricity output difference that has not yet been explained by the selected hydropower influencing factors in the current analysis round.
[0022] Among them, the preset set of hydropower influencing factors is a set of hydrological factors, hydraulic engineering factors, unit factors, control factors and grid-side factors that may change the state of water energy transmission or water-to-electricity conversion.
[0023] Among them, the abnormal residual reduction analysis is the process of calculating the degree of reduction of the current abnormal residual data before and after each candidate hydropower influencing factor participates in the anomaly interpretation, and comparing the contribution capacity of each candidate hydropower influencing factor accordingly.
[0024] Among them, the target hydropower influencing factors are those hydropower influencing factors in the preset set of hydropower influencing factors that can significantly and effectively reduce the current abnormal residual data and meet the topology connection requirements.
[0025] Among them, factor data is time series data that records the operating status or changes of the target hydropower influencing factors according to the sampling time.
[0026] Among them, the time-delay aligned orthogonal joint analysis is the process of processing the time difference between the correction factor data and the current abnormal residual data, and separating the abnormal contributions of the target hydropower influencing factors through orthogonal projection.
[0027] Among them, the joint anomaly analysis data is the updated residual data formed by subtracting the abnormal contributions caused by the target hydropower influencing factors after time delay correction from the current abnormal residual data.
[0028] Among them, the preset hydropower coupling topology is a directed topology structure that is pre-established according to the direction of hydropower transmission and includes the connection relationship between the hydropower input end, influencing factor nodes, energy conversion nodes and power output end.
[0029] The influence path is a sequence of nodes in the preset hydropower coupling topology that are continuously connected from the hydropower input end to the power output end through one or more nodes corresponding to the target hydropower influencing factors.
[0030] Among them, the target anomaly analysis data is the data formed after completing the cyclic factor screening and anomaly contribution deduction, which includes the target hydropower influencing factors, the complete influence path and the final remaining anomaly residual.
[0031] Specifically, the sampling time, flow rate, effective head, guide vane opening, unit speed, measured electrical power, and data validity status in the abnormal water energy data are used as processing objects. The calculation is based on the flow range, effective head range, unit load range, normal conversion efficiency, and normal transmission time during normal operation of the hydropower system. The sampling time of each abnormal water energy data is corrected according to a unified clock, and records with different sampling frequencies are adjusted to a preset unified sampling interval. When the time distance between adjacent valid records does not exceed the preset allowable value, weights are assigned according to the distance between the time to be calculated and two adjacent sampling times, with closer records receiving a larger weight, and corresponding values are supplemented accordingly. When the time distance exceeds the preset allowable value, the corresponding time period is set as an invalid time period. Normal operation records are divided into multiple operating intervals according to the flow range, effective head range, and unit load range. Within each operating interval, the water density, gravitational acceleration, flow rate, and effective head are multiplied to obtain the water energy input at each sampling time, and the ratio of measured electrical power to water energy input is used as the conversion efficiency at the corresponding sampling time. When multiple conversion efficiencies exist within the same operating range, weights are assigned based on the distance between each normal operating record and the center value of that operating range; the closer the distance, the greater the weight. The weighted average result is used as the normal conversion efficiency for that operating range. For adjacent operating ranges, the normal conversion efficiency is continuously adjusted according to the proximity of flow rate and effective head to avoid abrupt changes at the boundary of the operating ranges, and the adjusted normal conversion efficiency is limited to between zero and one. Subsequently, multiple candidate transfer times are set based on the length of the water diversion system, the allowable range of water flow velocity, and the unit response time. The sampling time of the water energy input is then postponed by the corresponding time. The postponed water energy input is then multiplied by the normal conversion efficiency of the corresponding operating range to obtain the calculated normal electrical power value at each electrical power sampling time. For each candidate transmission duration, the difference between the measured electrical power and the calculated normal electrical power is calculated. Each difference is squared and weighted according to the data validity status and operational stability. All weighted squares are then summed. The candidate transmission duration with the smallest total weighted square value is taken as the normal transmission duration for the corresponding operating interval, thus determining the normal water energy input-electrical energy output conversion relationship. Next, the normal electrical power is calculated according to the operating interval to which the abnormal water energy data belongs. The measured electrical power is subtracted from the normal electrical power, and the median of the normal power deviation for the corresponding operating interval is also subtracted. This is then divided by the larger of the absolute median difference of the normal power deviation and the preset minimum scale value. Finally, the processed residual value, sampling time, residual positive and negative direction, operating interval, and data validity status are combined to form the current abnormal residual data.
[0032] The sampling time, residual value, positive and negative direction of the residual, operating interval, and data validity status of the current abnormal residual data are used as the content to be processed. The factor number, factor data, valid value range, allowable direction of action, earliest action time, latest action time, and status of participation in joint analysis from the preset set of hydropower influencing factors are used as comparison conditions. Hydropower influencing factors whose status of participation in joint analysis is "yes" are excluded to prevent the same hydropower influencing factor from being repeatedly selected in the loop. For each remaining hydropower influencing factor, multiple candidate time differences are set according to the sampling interval between its earliest and latest action times. The relative time position between the factor data and the current abnormal residual data is adjusted accordingly, and calculations are continued only using sampling times where both have valid records. Then, the median of the factor data and the current abnormal residual data are used to correct their center positions, and their absolute median differences are used to unify the numerical scale. When the absolute median difference is less than the preset minimum scale value, the preset minimum scale value is used to avoid unstable calculation results due to excessively small numerical fluctuations. For each candidate time difference, calculation parameters are set to adjust the magnitude of changes in factor data. The product of the calculation parameter and the corresponding factor data value is subtracted from the residual value of the current abnormal residual data at each sampling time. The squared difference is then weighted according to the data validity status. The calculation parameters are changed successively until the sum of all weighted squares no longer decreases. Each time the calculation parameters are changed, the weights are reset based on the distance between the difference at each sampling time and the median of all differences. The greater the deviation of the difference, the smaller the weight is used to limit the impact of a single abrupt change on the calculation parameters. The weighted square sum of the original residual values of the current abnormal residual data is subtracted from the adjusted weighted square sum of the differences, and then divided by the original weighted square sum of the residual values to obtain the percentage decrease in residuals for the corresponding hydropower influencing factor at that candidate time difference. Subsequently, the current abnormal residual data is divided into multiple overlapping calculation periods according to continuous time. Within each calculation period, the residual decrease ratio and the direction of the calculation parameters are calculated. When the direction of the calculation parameters in most calculation periods does not conform to the permissible direction of the corresponding hydropower influencing factor, the comparison value of that hydropower influencing factor is set to zero. When the residual decrease direction is consistent across calculation periods, the median of the residual decrease ratios for each calculation period is subtracted from its absolute median difference to obtain the comparison value of that hydropower influencing factor. The remaining hydropower influencing factors are arranged in descending order of comparison value. The hydropower influencing factor with the largest comparison value exceeding the mismatch limit under normal operating conditions is identified as the target hydropower influencing factor. When multiple hydropower influencing factors have the same comparison value, their effective sampling time coverage ratio, the degree of fluctuation between different calculation periods, and the candidate time difference are compared sequentially. The target hydropower influencing factor with a larger coverage ratio, smaller fluctuation, and shorter candidate time difference is prioritized, and its factor number, corresponding candidate time difference, permissible direction of action, and residual decrease ratio are recorded.
[0033] The calculation uses the sampling time, factor value, and data validity status of the factor data corresponding to the target hydropower influencing factors, as well as the sampling time, residual value, positive / negative direction of the residual, and data validity status of the current abnormal residual data, as the joint calculation content. The earliest action time, latest action time, and allowable action direction corresponding to the target hydropower influencing factors are used as constraints. The factor data and current abnormal residual data are arranged according to a uniform sampling interval. The time range within which both have valid records is used as the calculation range, and multiple candidate time differences are set according to the sampling interval between the earliest and latest action times. For each candidate time difference, the sampling time of the factor data is adjusted as a whole to ensure that the factor data corresponds to the current abnormal residual data hourly. When the adjusted sampling time is between two valid records, weights are allocated according to the time distance between the sampling time and the two valid records, with closer valid records receiving a larger weight. When the adjusted sampling time exceeds the range of valid records, that sampling time is not used in the calculation. The time-series factor data and current outlier residual data are divided into multiple calculation periods of equal length with overlapping adjacent parts. A calculation parameter is set for each calculation period. The product of the calculation parameter and the corresponding factor data value is subtracted from the residual value of the current outlier residual data. The calculation parameter is adjusted successively until the weighted sum of the resulting differences multiplied by the corresponding factor data values is zero. When the number of valid sampling times within a calculation period is insufficient, a separate calculation parameter is not determined; instead, the calculation parameters of adjacent valid calculation periods are used as a weighted supplement based on time distance. For each candidate time difference, the weighted squares of the differences obtained in each calculation period are summed, and a continuity cost is added based on the magnitude of the change in calculation parameters between adjacent calculation periods. The greater the change in calculation parameters between adjacent calculation periods, the higher the added continuity cost; when the calculation parameters change continuously, no continuity cost is added, or only a small one is added. The total computational cost of each candidate time difference is obtained by summing all weighted squared values with the continuity cost. The candidate time difference with the lowest total computational cost, whose computational parameters in each computational period conform to the permissible direction of action, and whose effective sampling time coverage reaches a preset ratio is determined as the actual time difference. When the total computational cost of multiple candidate time differences is the same, the candidate time difference with the smaller absolute time difference and the slower change of computational parameters in adjacent computational periods is preferred. Finally, the factor data is adjusted according to the actual time difference by subtracting the product of the computational parameters and the corresponding factor data values in the computational period from the residual values of the current abnormal residual data at each sampling time. When the same sampling time belongs to multiple overlapping computational periods, the multiple calculation results are weighted and merged according to the effective weight of the sampling time in each computational period. The merged values, sampling times, data validity status, actual time differences, and corresponding target hydropower influencing factors are combined to form joint anomaly analysis data.
[0034] The data processing involves iteratively processing the sampling time, merged values, data validity status, actual time difference, and target hydropower influencing factors from the joint anomaly analysis data, as well as the hydropower input end, power output end, factor nodes, node connection directions, and node connection relationships in the preset hydropower coupling topology. After each round of joint analysis, the merged values from the joint anomaly analysis data are written item by item into the current anomaly residual data according to the sampling time, allowing them to directly participate in the next round of anomaly residual reduction analysis. The factor nodes corresponding to the target hydropower influencing factors in this round are added to the selected node combination, and the participation status of the target hydropower influencing factor in the joint analysis is set to "yes." When a selected node combination is first formed, the upstream and downstream connection nodes of that factor node are recorded. In the next round, only hydropower influencing factors that have not yet participated in the joint analysis and can be directly connected to the upstream or downstream end of the selected node combination along the preset connection direction are allowed to participate in the comparison. This ensures that the newly added target hydropower influencing factors in each round can extend the existing node combination and avoid the formation of factor nodes that are disconnected from the existing node combination. For each new target hydropower influencing factor, the hydropower input end is used as the starting point for access. Selected node combinations are checked sequentially according to the node connection direction in the preset hydropower coupling topology. If a factor node has an allowed connection with an already visited node and corresponds to the target hydropower influencing factor in the current or previous round, that factor node is included in the visited range, and the downstream connected nodes of that factor node are checked until no new visited nodes can be added. If the power output end is not within the visited range, the current round's joint anomaly analysis data is used as the current anomaly residual data to re-determine the target hydropower influencing factor and complete the next round of joint analysis. If the power output end is within the visited range, the process returns from the power output end to the hydropower input end step-by-step according to the adopted node connection relationships, checking whether the factor nodes corresponding to the target hydropower influencing factors in each round are all within the same continuous connection relationship. For cases where the power output end is reachable but there are still factor nodes corresponding to the target hydropower influencing factors outside the continuous connection relationship, the loop continues, and hydropower influencing factors that can connect the outside factor node with existing node combinations are included in the next round's comparison range. The loop stops only when the hydropower input can continuously reach the power output according to the node connection direction, all adjacent factor nodes have a preset connection, and the factor nodes corresponding to each target hydropower influencing factor are all within this continuous connection relationship. The target anomaly analysis data is composed of the last round of joint anomaly analysis data, the change results of the current anomaly residual data in each round, the determination order of each target hydropower influencing factor, the connection order of each factor node, and the complete influence path from the hydropower input to the power output.
[0035] In the aforementioned method for analyzing hydropower anomalies, the transmission time delay correction of changes in water energy input and power output in the abnormal hydropower data is performed by using the normal water energy input-power output conversion relationship. This prevents time misalignments caused by normal hydraulic transmission and unit response from being mistakenly counted as anomalies, thereby improving the accuracy of the current abnormal residual data. Furthermore, by comparing the degree of reduction of the current abnormal residual data by each hydropower influencing factor in the preset set of hydropower influencing factors, the target hydropower influencing factor is determined. Factors with stable correspondences to the remaining abnormal changes are prioritized, reducing interference from irrelevant factors, accidental synchronous fluctuations, and already analyzed factors. Using time-delay aligned orthogonal joint analysis of the factor data corresponding to the target hydropower influencing factor and the current abnormal residual data, overlapping factors can be separated at a unified time point, reducing the impact of multiple hydropower factors. Repeated attribution caused by the correlation between factors is addressed by continuously updating the current anomaly residual data with joint anomaly analysis data. This allows each influencing factor to participate in the analysis round by round according to its actual effect on unexplained anomalies. At the same time, the complete influence path formed by each target hydropower influencing factor in the preset hydropower coupling topology, connecting the hydropower input end to the power output end, is used as the loop termination condition. This can constrain the data change relationship with the actual hydraulic transmission, mechanical conversion, and power output relationship of the hydropower generation system, avoiding premature termination based solely on residual thresholds or the formation of disconnected factor combinations. Ultimately, the target anomaly analysis data with temporal consistency, factor independence, and physical connectivity is obtained, which helps to ensure the reliability, completeness, and interpretability of multi-factor coupled anomaly analysis and can provide an accurate basis for anomaly location, operation adjustment, and maintenance arrangement of hydropower generation systems.
[0036] In an exemplary embodiment, time-delay aligned orthogonal joint analysis is performed on the factor data corresponding to the target hydropower influencing factors and the current abnormal residual data to obtain joint anomaly analysis data, including steps 302 to 306. Wherein:
[0037] Step 302: Based on the transmission path of the target hydropower influencing factors to the power output end in the preset hydropower coupling topology, perform step-by-step gain transfer calculation on the factor data to obtain equivalent power disturbance data.
[0038] Step 304: Based on the cumulative transmission time of the transmission path, perform energy centroid time delay alignment between the equivalent power disturbance data and the current abnormal residual data to obtain time delay aligned disturbance data.
[0039] Step 306: Based on the inertial duration of the water flow in the hydropower system and the rotational inertial duration of the unit, perform a dual-inertial orthogonal joint analysis on the time delay alignment disturbance data and the current abnormal residual data to obtain joint anomaly analysis data.
[0040] The transmission path is a directed connection sequence in which the nodes corresponding to the target hydropower influencing factors pass through nodes such as hydraulic facilities, turbines and generators in a preset hydropower coupling topology to reach the power output end.
[0041] Among them, the step-by-step gain transfer calculation is a calculation process that uses the input-output gain of each node in the order of the transfer path to perform amplitude conversion on the factor data and transfer the conversion result to the next node.
[0042] Among them, the equivalent power disturbance data is the power disturbance time series formed at the power output end after the factor data has been converted by the gain at each level in the transmission path.
[0043] Among them, energy centroid time delay alignment is a process that uses the energy magnitude of data at each sampling time as a time weight to determine the energy concentration time, and moves the corresponding data according to the difference between two energy concentration times.
[0044] Among them, the time-delay aligned disturbance data is a disturbance sequence formed by shifting the equivalent power disturbance data according to the cumulative transmission time of the transmission path and the time difference of energy concentration.
[0045] Among them, the inertial duration of water flow is the characteristic time required for the water flow state to transition from the input change to the corresponding steady state due to inertial action in the water intake pipe, pressure pipe and turbine flow passage components.
[0046] Among them, the rotational inertia time of the unit is the characteristic time required for the rotor of the hydro-generator unit to transition from mechanical power change to the corresponding speed or electrical power state under the action of rotational inertia.
[0047] Among them, the dual-inertial orthogonal joint analysis is an analysis process that processes the time delay alignment disturbance data and the current abnormal residual data according to the water flow inertial duration and the unit rotation inertial duration, respectively, and calculates the factor contribution and updates the abnormal residuals through orthogonal relationship.
[0048] Specifically, the transmission constraints are defined as follows: the factor nodes corresponding to the target hydropower influencing factors in the preset hydropower coupling topology; the transmission path from the factor nodes to the power output terminals; the node arrangement order in the transmission path; the allowable transmission direction between adjacent nodes; the operating range; the amplitude conversion relationship; and the allowable output range. The calculation objects are the sampling time, factor values, normal reference values, and data validity status in the factor data. The factor nodes, intermediate nodes, and power output terminals corresponding to the target hydropower influencing factors are arranged sequentially according to the transmission direction. Each pair of adjacent nodes is checked for a valid connection, and the corresponding connection is checked for availability under the current unit operating mode. If a connection is not allowed to be transmitted in the current operating direction, the transmission path is not used for further calculation. For each level of connection in the transmission path, the amplitude conversion relationship is determined according to the operating range to which the data at both ends of that level of connection belong. When the change amplitude is small and has not reached the equipment adjustment boundary, the stable ratio between the output change and the input change within that operating range is used as the conversion value. When the change amplitude spans multiple operating ranges, calculations are performed according to the intervals through which the factor values pass, and the remaining change after the boundary of the previous interval is passed to the next interval for further processing. Then, the factor values at each sampling time are subtracted from the corresponding normal baseline values, and the resulting changes are assigned to the first factor node in the transmission path. These changes are then processed level by level according to the node arrangement. For each level of connection, the output change of the previous level node is multiplied by the converted value of that level connection, and the result is used as the input change of the next level node. When a level connection has a positive interaction, the direction of change is maintained; when a level connection has a negative interaction, the direction of change is changed. When the calculation result exceeds the allowable output range of that level node, it is limited to the corresponding upper or lower limit, and the excess is recorded as a change that cannot be further transmitted by that level connection. For nodes in the transmission path with flow or power allocation, the change is allocated to that node according to the flow or power proportion of the corresponding branch in the current operating interval. When multiple branches converge at subsequent nodes, only the portion corresponding to the transmission path of the target hydropower influencing factor is accumulated, avoiding duplicate calculations of changes from other branches. After processing all adjacent nodes in the transmission path in sequence, the change in the amount of power reaching the power output terminal is taken as the power change corresponding to the factor data at the corresponding sampling time. The sampling time, power change, direction of change, order of nodes passed, conversion results at each level, limiting status and data validity status are combined to form the equivalent power disturbance data.
[0049] The time adjustment is based on the transmission duration and cumulative result of each adjacent node in the transmission path. Alignment is achieved using the sampling time, power change, direction of change, and data validity status in the equivalent power disturbance data, as well as the sampling time, residual value, positive / negative direction of the residual, and data validity status in the current abnormal residual data. The transmission duration of each level of connection in the transmission path is added in the order of node arrangement to obtain the cumulative transmission duration of the transmission path. Then, the cumulative transmission duration of each sampling time in the equivalent power disturbance data is adjusted backward to ensure that the expected time of the factor change reaching the power output end is within the same time range as the current abnormal residual data. Next, the adjusted equivalent power disturbance data and the current abnormal residual data are divided into multiple overlapping calculation periods according to a unified sampling interval. Within each calculation period, the power change and residual value at each sampling time are squared and multiplied by the unified sampling interval to obtain the energy weight of the corresponding sampling time within that calculation period. Subsequently, each sampling moment is multiplied by its energy weight, summed, and divided by the sum of all energy weights within the calculation period to obtain the energy centroid time of the equivalent power disturbance data and the current abnormal residual data within that calculation period. When the sum of all energy weights within a certain calculation period is lower than the preset effective energy lower limit, the energy centroid time of that period is not calculated separately; instead, the result of the calculation period closest in time and satisfying the effective energy lower limit is used. If there are no calculation periods that satisfy the conditions before or after, the time adjustment for that period is completed only using the cumulative transmission time. The energy centroid time of the current abnormal residual data is subtracted from the energy centroid time of the equivalent power disturbance data to obtain the time correction amount for that calculation period. When the time correction amount exceeds the sum of the allowable deviations of the transmission time at each level of the transmission path, it is limited to the allowable deviation range to prevent excessive movement of the equivalent power disturbance data that does not conform to the actual transmission process due to local abnormal fluctuations. For cases where the same sampling moment belongs to multiple calculation periods, weights are allocated according to the distance between the sampling moment and the center time of each calculation period, and multiple time correction amounts are weighted and merged to ensure continuous change in time adjustment between adjacent calculation periods. Finally, the sampling time of the equivalent power disturbance data is adjusted again according to the merged time correction, and the adjusted power change is configured to the sampling time corresponding to the current abnormal residual data. When the adjusted sampling time is between two residual sampling times, the power change is distributed to the two adjacent sampling times according to the time distance, while maintaining the original change direction. Finally, the adjusted sampling time, power change, change direction, cumulative transmission time, time correction, and data validity status are combined to form the time delay aligned disturbance data.
[0050] The study uses the inertial duration of water flow and the rotational inertia duration of the hydropower system as two types of time-dependent ranges. It also uses the sampling time, power change, direction of change, and data validity status from the time-delay aligned disturbance data, and the sampling time, residual value, positive / negative direction of the residual, and data validity status from the current abnormal residual data as the joint processing objects. The time-delay aligned disturbance data and the current abnormal residual data are then arranged according to a unified sampling time. For each sampling time, historical records not exceeding the water flow inertia duration and historical records not exceeding the unit rotational inertia duration are selected backwards. Within the historical records corresponding to the water flow inertia duration, weights are assigned according to the time distance between the historical record and the current sampling time; the shorter the time distance, the greater the weight. Weighted results for power change and residual value are calculated separately. Within the historical records corresponding to the unit rotational inertia duration, the corresponding weighted results are calculated again according to the same rules. To avoid the repetition of the same changes between the two types of inertial effects, the synchronization between the power changes under the water flow inertial duration and the power changes under the unit rotation inertial duration is first compared. When both have the same direction of change at the same sampling time, weights are assigned according to the effective record proportions covered by each type of inertial duration. When the directions of change are opposite, the portion of the larger absolute change exceeding the other is retained, and the mutually canceling portions are set to zero. Subsequently, an adjustment ratio is set for each continuous calculation period. The residual value of the current abnormal residual data is subtracted from the product of the adjustment ratio and the power change corresponding to the time delay alignment disturbance data, and the resulting difference is historically weighted according to the water flow inertial duration and the unit rotation inertial duration, respectively. The adjustment ratio is changed successively, and after each change, the weighted sum of the product of the difference obtained under the water flow inertial duration and the power change, and the weighted sum of the product of the difference obtained under the unit rotation inertial duration and the power change are calculated respectively. Adjustment stops when both weighted sums do not exceed the preset allowable deviation. When the two weighted sums cannot simultaneously reach the preset allowable deviation, the adjustment ratio with the smallest sum after squares that conforms to the allowable direction of the target hydropower influencing factor is taken as the final value for that calculation period. For adjacent calculation periods, the system limits the change in the adjustment ratio to no more than the preset continuous change upper limit. When there are insufficient valid records for a certain calculation period, the adjacent adjustment ratios are weighted according to the time distance between it and the valid calculation periods before and after it. Finally, using the final adjustment ratio of each calculation period, the difference between the residual value of the current abnormal residual data and the power change corresponding to the adjustment ratio multiplied by the time delay alignment disturbance data is calculated hourly. When the same sampling time belongs to multiple calculation periods, the multiple differences are weighted and merged according to the valid record ratio of the corresponding calculation period. The merged difference, sampling time, positive and negative direction of the residual, target hydropower influencing factor, data validity status, and two types of inertial duration are combined to form joint anomaly analysis data.
[0051] In this embodiment, by converting factor data step by step along the transmission path, hydropower influencing factors of different dimensions can be unified to the power output side; by combining the cumulative transmission time and the energy center of gravity completion time correction, the mismatch caused by disturbance broadening and waveform offset can be reduced; and by utilizing the joint constraint of water flow inertia and unit rotation inertia, instantaneous fluctuations can be suppressed while retaining continuous disturbances, thereby improving the stability of joint anomaly analysis data.
[0052] In an exemplary embodiment, based on the water flow inertia duration and the unit rotation inertia duration of the hydropower system, a dual-inertia orthogonal joint analysis is performed on the time delay alignment disturbance data and the current abnormal residual data to obtain joint anomaly analysis data, including steps 402 to 406. Wherein:
[0053] Step 402: Perform a dual-inertial state transition on the time delay aligned disturbance data and the current abnormal residual data to obtain the disturbance dual-inertial state trajectory and the residual dual-inertial state trajectory.
[0054] Step 404: Calculate the inner product of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory to obtain the contribution coefficient of the target factor.
[0055] Step 406: Based on the contribution coefficient of the target factor, subtract the factor contribution corresponding to the time delay alignment disturbance data from the current abnormal residual data to obtain the joint anomaly analysis data.
[0056] The dual-inertial state transition is a process of performing graded inertial response calculations on the input time series according to the inertial duration of the water flow and the rotational inertial duration of the unit, and combining the response states at each sampling time into a dual-inertial state trajectory.
[0057] Among them, the perturbation dual-inertial state trajectory is a sequence of perturbation state points formed by arranging the time-delay aligned perturbation data according to the sampling time after dual-inertial state transformation.
[0058] Among them, the residual dual-inertial state trajectory is the sequence of residual state points formed by arranging the current abnormal residual data according to the sampling time after the dual-inertial state transformation.
[0059] Among them, the dual-inertial inner product calculation is the process of calculating the directional consistency and amplitude overlap of the corresponding state displacements of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory under two time scales: water flow inertia and unit rotational inertia.
[0060] Among them, the target factor contribution coefficient is a value determined based on the calculation results of the double inertial inner product, used to limit the contribution ratio and direction of the target hydropower influencing factors to the current abnormal residual data.
[0061] Among them, the factor contribution is the disturbance component corresponding to the target hydropower influencing factor in the current abnormal residual data after the amplitude of the time delay alignment disturbance data is adjusted according to the target factor contribution coefficient.
[0062] Specifically, the sampling time, power change, direction of change, and data validity status in the time-delay aligned disturbance data, and the sampling time, residual value, positive / negative direction of the residual, and data validity status in the current abnormal residual data are used as the conversion objects. The water flow inertia duration and the unit rotation inertia duration are used as time recursion parameters. The time-delay aligned disturbance data and the current abnormal residual data are arranged according to a uniform sampling interval. Only records with the same sampling time and both in a valid state are matched, and consecutive valid records are divided into independent calculation segments. When the interval between adjacent valid records exceeds the larger of the water flow inertia duration and the unit rotation inertia duration, the calculation restarts at the next valid record to avoid unfounded continuity of values due to long periods of missing data. For each calculation segment, the first power change in the time-delay aligned disturbance data and the first residual value in the current abnormal residual data are used as the initial values of their respective first components. In subsequent sampling times, the first update ratio is obtained by dividing the actual sampling interval by the water flow inertia duration. When the first update ratio is greater than one, it is set to one. The current value is then retained according to the first update ratio, and the first component of the previous sampling time is retained by subtracting the first update ratio from one. This ensures that the closer the sampling interval is to the water flow inertia duration, the greater the influence of the current value on the first component. The first component is then used as the initial value of the second component. The second update ratio is obtained by dividing the actual sampling interval by the unit rotation inertia duration. When the second update ratio is greater than one, it is set to one. The first component of the current sampling time is used according to the second update ratio, and the second component of the previous sampling time is used by subtracting the second update ratio from one. This ensures that changes in the first component gradually enter the second component after the unit rotation inertia duration. When the water flow inertia duration or the unit rotation inertia duration is less than the unified sampling interval, the corresponding update ratio is set to one, allowing the corresponding component to directly use the current input. When the data at a certain sampling time is invalid but the missing duration does not exceed the larger of the above values, the two components of the previous sampling time are maintained, and that time is set to invalid and not used in subsequent calculations. Finally, the first and second components corresponding to the time delay aligned disturbance data are arranged sequentially according to the sampling time to obtain the disturbance dual-inertial state trajectory. The first and second components corresponding to the current abnormal residual data are arranged according to the same rules to obtain the residual dual-inertial state trajectory. The two trajectories use the same sampling time, calculation segment number and data validity state to ensure the consistency of subsequent corresponding calculations.
[0063] The calculation content includes the sampling time, calculation segment number, first component, second component, and valid data state in the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory. The calculation weights of the two types of components are allocated according to the proportion of the water flow inertial duration and the unit rotation inertial duration in the sum of their durations. For each calculation segment, the first component of the two trajectories at the same sampling time is multiplied by the first component of the disturbed dual-inertial state trajectory and the first component of the residual dual-inertial state trajectory, and weighted according to the calculation weights corresponding to the valid data state and the water flow inertial duration. Then, the second component of the two trajectories at the same sampling time is multiplied by the second component of the two trajectories, and weighted according to the calculation weights corresponding to the valid data state and the unit rotation inertial duration. Finally, the two weighted products of all valid sampling times are added together to obtain the cumulative value of the common change part of the two trajectories. Simultaneously, the first component of the disturbed dual-inertial state trajectory at each effective sampling time is squared and accumulated according to the calculation weight corresponding to the water flow inertia duration. The second component is squared and accumulated according to the calculation weight corresponding to the unit rotation inertia duration. The two accumulated results are added together to obtain the cumulative reference value of the disturbed dual-inertial state trajectory. At the same time, the cumulative value of the common change part is divided by the cumulative reference value to obtain the proportional result corresponding to the calculation segment. When the cumulative reference value is less than the preset minimum effective value, the calculation segment is set to invalid to prevent the proportional result from being abnormally amplified when the disturbance amplitude is too small. When there are multiple effective calculation segments, weights are allocated according to the number of effective sampling times and the cumulative reference value of each calculation segment, and the weighted average of each proportional result is calculated. When the difference between the proportional result of a certain calculation segment and the median of the proportional results of the other calculation segments exceeds the preset allowable range, the weight of the calculation segment is reduced, and the weighted average is recalculated. The final weighted average value is compared with the permissible direction of the target hydropower influencing factors. When the two directions are consistent, the final weighted average value is retained. When the two directions are opposite and the reverse amplitude does not exceed the permissible fluctuation range under normal operating conditions, it is set to zero. When the reverse amplitude exceeds the permissible fluctuation range, its actual direction is retained and the state of inconsistency is recorded. Finally, the contribution coefficient of the target factor is obtained.
[0064] The time-by-time processing involves using the target factor contribution coefficient, sampling time, power change, and data validity status in the time-delay alignment disturbance data, as well as the sampling time, residual value, positive / negative direction, and data validity status in the current abnormal residual data. The time-delay alignment disturbance data is mapped one-to-one with the current abnormal residual data according to the sampling time. When both are valid at the same sampling time, the target factor contribution coefficient is multiplied by the power change at that sampling time to obtain the factor contribution of the time-delay alignment disturbance data at that sampling time. Then, the factor contribution is subtracted from the residual value of the current abnormal residual data to obtain the joint residual value at that sampling time. When the current abnormal residual data is valid but the time-delay alignment disturbance data is invalid, the original residual value is maintained and the factor contribution is set to invalid. When the current abnormal residual data is invalid, the joint residual value is not calculated. When the target factor contribution coefficient is zero, the residual value of the current abnormal residual data is directly used. For sampling moments located between two effective power changes with a time interval not exceeding a preset allowable value, weights are assigned according to the time distance between this sampling moment and the two preceding and following effective sampling moments. The corresponding power change is then calculated, and the factor contribution is determined. When the time interval exceeds the preset allowable value, no numerical padding is performed. After processing all sampling moments, the weighted sum of the power changes obtained by multiplying the joint residual value and the time-delay aligned disturbance data item by item within the continuous effective time period is checked according to the calculation weights corresponding to the water flow inertia duration and the unit rotation inertia duration. When the weighted sum does not fall within the preset allowable deviation due to numerical accuracy, its last value is gradually adjusted without changing the direction of the target factor contribution coefficient until the weighted sum falls within the preset allowable deviation. Finally, the joint residual value, sampling moment, positive and negative direction of the joint residual, factor contribution, target factor contribution coefficient, target hydropower influencing factors, and data validity status are combined to form the joint anomaly analysis data.
[0065] In this embodiment, the historical response of water flow and unit is preserved by using dual inertial state trajectories, and the contribution of factors is determined by using the dual inertial inner product. This can reduce over-attribution caused by instantaneous correlation and enable joint anomaly analysis data to more accurately preserve unexplained anomaly components.
[0066] In an exemplary embodiment, a dual-inertial state transition is performed on the time-delay aligned perturbation data and the current abnormal residual data to obtain the perturbation dual-inertial state trajectory and the residual dual-inertial state trajectory, including steps 502 to 508. Wherein:
[0067] Step 502: Based on the inertial duration of the water flow, perform first-order inertial recursion on the time-delay aligned disturbance data and the current abnormal residual data to obtain the disturbed water flow state sequence and the residual water flow state sequence.
[0068] Step 504: Based on the unit's rotational inertia duration, perform first-order inertial recursion on the disturbed water flow state sequence and the residual water flow state sequence to obtain the disturbed rotational state sequence and the residual rotational state sequence.
[0069] Step 506: Perform hourly directed difference calculations on the disturbed flow state sequence, the disturbed rotation state sequence, the residual flow state sequence, and the residual rotation state sequence to obtain the disturbed inertial conversion difference sequence and the residual inertial conversion difference sequence.
[0070] Step 508: Based on the inertial duration of the water flow and the rotational inertial duration of the unit, perform dual-duration-delay coordinate embedding on the disturbance inertial conversion difference sequence and the residual inertial conversion difference sequence to obtain the disturbance dual-inertial state trajectory and the residual dual-inertial state trajectory.
[0071] The first-order inertial recursion is a process of determining the update ratio of new and old data according to the sampling interval and the corresponding inertial duration, and calculating the current state value hourly using the state value of the previous sampling moment and the current sample value.
[0072] Among them, the disturbed water flow state sequence is the water flow side response sequence formed by first-order inertial recursion of time-delay aligned disturbance data according to the water flow inertial duration.
[0073] Among them, the residual flow state sequence is the flow-side response sequence formed by first-order inertial recursion of the current abnormal residual data according to the flow inertia duration.
[0074] Among them, the disturbance rotation state sequence is the unit rotation side response sequence formed by continuing the disturbance water flow state sequence according to the first-order inertial recursion of the unit rotation inertia duration.
[0075] Among them, the residual rotational state sequence is the unit rotational side response sequence formed by continuing the residual water flow state sequence with first-order inertial recursion according to the unit rotational inertia duration.
[0076] Among them, the time-by-time directional difference calculation is the process of subtracting the corresponding water flow state value from the rotation state value according to the direction from the water flow state to the rotation state at each same sampling time.
[0077] Among them, the disturbance inertial conversion difference sequence is a time series formed by calculating the time-by-time directional difference between the disturbance rotation state sequence and the disturbance water flow state sequence.
[0078] Among them, the residual inertia transformation difference sequence is a time series formed by the time-by-time directed difference calculation of the residual rotation state sequence and the residual flow state sequence.
[0079] Among them, the dual-duration delay coordinate embedding is the process of converting the water flow inertia duration and the unit rotation inertia duration into sampling delay steps respectively, and combining the sequence values of the current time and the two corresponding delay times into state coordinates.
[0080] Specifically, the sampling time, power change, direction of change, and data validity status in the time-delay aligned perturbation data, and the sampling time, residual value, positive and negative direction of the residual, and data validity status in the current abnormal residual data are used as the recursive objects. The water flow inertia duration is used as the control parameter for the update speed. The time-delay aligned perturbation data and the current abnormal residual data are adjusted to the same sampling interval, and continuous valid segments are divided according to the sampling time. When the time interval between adjacent valid records is greater than the water flow inertia duration, the recursion restarts at the next valid record to avoid the continuity of calculation results caused by long-term data loss. For each continuous valid segment, the first power change and the first residual value are used as the initial water flow state value of the corresponding sequence. For each subsequent sampling time, the time interval between the sampling time and the previous sampling time is divided by the sum of the time interval and the water flow inertia duration to obtain the current update ratio. The current input value is then included in this recursion according to the current update ratio, and the water flow state value of the previous sampling time is included in this recursion by subtracting the current update ratio from one. The two are added together to obtain the water flow state value at the current sampling time. The longer the flow inertia duration, the higher the proportion of the previous flow state value retained; when the flow inertia duration is zero, the current flow state value directly adopts the current input value. When the data at a certain sampling time is invalid and the missing duration does not exceed the flow inertia duration, the previous valid flow state value is maintained, and that sampling time is set to invalid; when the missing duration exceeds the flow inertia duration, the initial value is reset at the next valid sampling time. The flow state values corresponding to the disturbance data are arranged according to the sampling time and time delay to obtain the disturbance flow state sequence, and the flow state values corresponding to the current abnormal residual data are arranged according to the same rules to obtain the residual flow state sequence; both sequences record the sampling time, flow state value, continuous valid segment number, and data validity status.
[0081] The sampling time, flow state value, continuous valid segment number, and data validity status in the disturbed flow state sequence and residual flow state sequence are used as the recursive objects, and the unit rotational inertia duration is used as the control parameter for the update rate of this round. Two flow state sequences are arranged according to the sampling time, and it is checked whether any adjacent sampling time belongs to the same continuous valid segment. For each continuous valid segment, the first disturbed flow state value and the first residual flow state value are used as the initial rotational state value of the corresponding sequence. For each subsequent sampling time, the current sampling interval is divided by the sum of the sampling interval and the unit rotational inertia duration to obtain the current rotational update ratio. The current flow state value is then included in this recursion according to the current rotational update ratio, and the rotational state value of the previous sampling time is subtracted from the current rotational update ratio and included in this recursion. The two are added together to obtain the current rotational state value. The larger the unit rotational inertia duration, the slower the rotational state value changes with the flow state value; when the unit rotational inertia duration is zero, the rotational state value directly adopts the flow state value of the same sampling time. When a sampling moment in the flow state sequence is invalid but the missing duration does not exceed the unit's rotational inertia duration, the previous valid rotational state value is maintained, and that sampling moment is set to invalid. If the missing duration exceeds the unit's rotational inertia duration, the recursion restarts at the next valid flow state value. The rotational state values corresponding to the disturbed flow state sequence are arranged according to the sampling moments to obtain the disturbed rotational state sequence. The rotational state values corresponding to the residual flow state sequence are arranged according to the same rules to obtain the residual rotational state sequence. Both rotational state sequences use the consecutive valid segment numbers of the corresponding flow state sequences to ensure that the flow state and rotational state correspond hourly.
[0082] The sampling times, state values, and valid data states in the disturbed flow state sequence and the disturbed rotation state sequence, as well as the sampling times, state values, and valid data states in the residual flow state sequence and the residual rotation state sequence, are used as the time-by-time calculation objects. The direction from the flow state to the rotation state is set as the fixed calculation direction for the directional difference. For the disturbance part, the disturbance flow state sequence and the disturbance rotation state sequence are matched one by one according to the sampling time. The calculation continues only when both state sequences are in a valid state and belong to the same continuous valid segment. The disturbance rotation state value at the current sampling time is subtracted from the disturbance flow state value to obtain the disturbance inertial conversion difference at the corresponding sampling time. When the obtained value is positive, the rotation state is recorded as higher than the flow state; when the obtained value is negative, the rotation state is recorded as lower than the flow state. When the absolute value does not exceed the minimum effective change value of the corresponding monitored quantity, it is set to zero. For the residual part, the residual rotation state value is subtracted from the residual flow state value according to the same sampling time and the same calculation direction to obtain the residual inertial conversion difference. The same minimum effective change value rule is used to process results close to zero. When any corresponding state sequence is invalid or two state values do not belong to the same continuous valid segment, the inertial conversion difference at that sampling time is not calculated, and its data validity state is set to invalid. Finally, the perturbation inertial conversion difference, the positive and negative directions of the difference, and the continuous valid segment number are arranged in chronological order for each valid sampling time to obtain the perturbation inertial conversion difference sequence; the residual inertial conversion difference, the positive and negative directions of the difference, and the continuous valid segment number are arranged in the same way to obtain the residual inertial conversion difference sequence.
[0083] The sampling time, inertial conversion difference, positive and negative directions of the difference, continuous valid segment number, and data validity status in the disturbance inertial conversion difference sequence and residual inertial conversion difference sequence are used as coordinate construction objects, and two delay positions are determined by the water flow inertial duration and the unit rotation inertial duration, respectively. For each current sampling time, the inertial conversion difference of that sampling time is used as the current coordinate, the inertial conversion difference of the time corresponding to the water flow inertial duration forward of the current sampling time is used as the water flow delay coordinate, and the inertial conversion difference of the time corresponding to the time corresponding to the unit rotation inertial duration forward of the current sampling time is used as the rotation delay coordinate. When the aforementioned delay time is completely consistent with the existing sampling time, the corresponding inertial conversion difference is directly used; when the delay time is between two valid sampling times and the two sampling times belong to the same continuous valid segment, the weight is assigned according to the time distance between the delay time and the two sampling times, and the closer the distance, the greater the weight of the inertial conversion difference; when the delay time exceeds the current continuous valid segment or there are invalid states in the preceding and following records, the complete coordinates of the current sampling time are not constructed. When the inertia duration of the water flow or the rotational inertia duration of the unit is less than the uniform sampling interval, the valid record closest to the corresponding delay time before the current sampling time is used. When the two inertia durations are the same, they are still recorded separately according to the water flow delay coordinates and the rotational delay coordinates to maintain a fixed order of the two time dimensions in subsequent calculations. The current coordinates, water flow delay coordinates, and rotational delay coordinates corresponding to each valid sampling time of the disturbance inertia conversion difference sequence are combined in a fixed order and connected sequentially along the sampling time to obtain the disturbance dual-inertia state trajectory; the residual inertia conversion difference sequence is combined according to the same rule to obtain the residual dual-inertia state trajectory.
[0084] In this embodiment, by using two-stage inertial recursion and time-by-time directional difference calculation, the dynamic differences in the process of water flow response to unit rotation response can be amplified, and a comparable trajectory structure can be formed by using dual-time-delay coordinates, which makes it easier to distinguish abnormal changes in different inertial stages.
[0085] In an exemplary embodiment, the inner product of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory is calculated to obtain the target factor contribution coefficient, including steps 602 to 606. Wherein:
[0086] Step 602: Calculate the local inner product of the state displacements of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory in each of the same sampling intervals to obtain the local contribution sequence.
[0087] Step 604: Based on the inertial duration of the water flow, perform forward inertial recursion on the local contribution sequence to obtain the water flow side contribution sequence.
[0088] Step 606: Based on the unit's rotational inertia duration, perform reverse inertial recursion on the local contribution sequence to obtain the unit-side contribution sequence.
[0089] Step 608: Perform a directed maximum-minimum bottleneck convergence calculation on the water flow side contribution sequence and the unit side contribution sequence under the constraint of continuous interval with the same sign to obtain the target factor contribution coefficient.
[0090] The sampling interval is a time range defined by two adjacent sampling times and their corresponding state points.
[0091] Among them, the state displacement is the change in each coordinate component between the final state point and the initial state point of the same state trajectory within a sampling interval.
[0092] The local inner product calculation is a process of multiplying the state displacement components corresponding to the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory within the same sampling interval and then summing them.
[0093] Among them, the local contribution sequence is a numerical sequence formed by arranging the local inner product results corresponding to each sampling interval in chronological order of sampling time.
[0094] Among them, forward inertial recursion is a process of updating the current contribution value hourly by using the recursion result of the previous moment and the current local contribution value according to the sampling time from early to late.
[0095] Among them, the water flow side contribution sequence is the contribution sequence formed by forward inertial recursion of the local contribution sequence according to the water flow inertial duration.
[0096] Among them, the reverse inertial recursion is a process of updating the current contribution value hourly by using the recursion result of the next moment and the current local contribution value according to the sampling time from late to early.
[0097] Among them, the unit-side contribution sequence is the contribution sequence formed by recursively extrapolating the local contribution sequence in the reverse inertia time according to the unit rotation inertia duration.
[0098] Among them, the same sign continuous interval constraint is to only include the corresponding positions of the water flow side contribution sequence and the unit side contribution sequence with the same sign and continuous along the sampling time into the same calculation interval, and to break the interval at the position where the sign changes or zero value appears.
[0099] The directed maximum-minimum bottleneck convergence calculation involves hourly selecting the smaller absolute value between the flow-side contribution and the generator-side contribution within each consecutive interval of the same sign as the bottleneck value, and then selecting the bottleneck value with the largest absolute value while retaining its positive and negative directions. Specifically, the sampling time, current coordinates, flow delay coordinates, rotation delay coordinates, continuous effective segment numbers, and data validity status in the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory are used as the calculation objects, and the time range between any two adjacent sampling times is defined as a sampling interval. For each sampling interval, it is first checked whether the two trajectories are both in a valid state at the start and end of the interval and belong to the same continuous effective segment. If either endpoint is invalid or the continuous effective segment numbers are inconsistent, the sampling interval is not calculated. For sampling intervals that meet the conditions, the corresponding coordinates of the start of the interval are subtracted from the current coordinates, flow delay coordinates, and rotation delay coordinates of the disturbed dual-inertial state trajectory at the end of the interval to obtain three disturbance state displacements; then, the three residual state displacements of the residual dual-inertial state trajectory are calculated in the same way. Multiply the disturbance state displacement at the same coordinate position by the residual state displacement, and sum the three products to obtain the directed cumulative value of the common change of the two trajectories within the sampling interval. Simultaneously, square the three disturbance state displacements and sum them to obtain the disturbance change baseline value for the sampling interval. Divide the directed cumulative value by the disturbance change baseline value to obtain the local contribution value for the sampling interval. When the disturbance change baseline value is lower than a preset minimum effective value, the corresponding local contribution value is set to invalid to avoid abnormal amplification of the value due to excessively small disturbance state displacements. A positive local contribution value indicates that the two trajectories change in the same direction within the sampling interval; a negative local contribution value indicates that they change in opposite directions; and a zero local contribution value indicates that there is no effective correspondence between the residual state displacement and the disturbance state displacement direction. Arrange the local contribution values, contribution directions, disturbance change baseline values, start and end times of the sampling interval, and data validity status in chronological order of the sampling interval to obtain the local contribution sequence.
[0100] The starting and ending times of the sampling intervals, local contribution values, contribution directions, disturbance change baseline values, and data validity states in the local contribution sequence are used as the forward recursion objects, and the water flow inertia duration controls the degree to which previous local contributions are retained in the current result. Sampling intervals are arranged from earliest to latest time, and the first valid local contribution value is used as the initial water flow-side contribution value. For each subsequent valid sampling interval, the duration of the current sampling interval is divided by the sum of that duration and the water flow inertia duration to obtain the current update ratio. The current local contribution value is then included in this recursion according to the current update ratio, and the water flow-side contribution value of the previous valid sampling interval is included in this recursion by subtracting the current update ratio from one. The two are added together to obtain the water flow-side contribution value of the current sampling interval. The larger the water flow inertia duration, the higher the retention ratio of previous water flow-side contributions. When the water flow inertia duration is zero, the current water flow-side contribution value directly adopts the current local contribution value. When an invalid interval exists between two consecutive valid sampling intervals and the total duration of the invalid interval does not exceed the inertial duration of the water flow, the previous valid water flow-side contribution value is maintained, and the invalid interval is marked as not participating in subsequent merging. When the total duration of the invalid interval exceeds the inertial duration of the water flow, the recursion restarts in the next valid sampling interval with the corresponding local contribution value. For adjacent sampling intervals where the positive and negative directions change, the recursion is still completed according to the above proportion, so that the water flow-side contribution value passes through zero before changing direction, avoiding a single abrupt change in the local contribution value that immediately alters the previous accumulated result. The water flow-side contribution value, contribution direction, sampling interval start and end time, disturbance change baseline value, and data validity status of each valid sampling interval are arranged in chronological order to obtain the water flow-side contribution sequence.
[0101] The back-recursive approach uses the start and end times of the sampling intervals, local contribution values, contribution directions, disturbance change baseline values, and data validity states in the local contribution sequence as the objects of back-recursion. The degree to which subsequent local contributions retain the current result is controlled by the unit rotation inertia duration. The sampling intervals are then arranged from latest to earliest time, with the last valid local contribution value used as the initial unit-side contribution value. For each preceding valid sampling interval, the time distance between the current and subsequent valid sampling intervals is divided by the sum of that time distance and the unit rotation inertia duration to obtain the back-update ratio. The current local contribution value is then included in this recursion according to the back-update ratio, and the unit-side contribution value already obtained from the subsequent valid sampling interval is included in this recursion by subtracting the back-update ratio. The two are then added together to obtain the unit-side contribution value for the current sampling interval. The longer the unit rotation inertia duration, the higher the proportion of subsequent unit-side contribution values retained forward; when the unit rotation inertia duration is zero, the current unit-side contribution value directly adopts the current local contribution value. When an invalid interval is encountered during the reverse recursion process, and the total duration of the invalid interval does not exceed the unit's rotational inertia duration, the next valid unit-side contribution value is maintained, and the invalid interval is set not to participate in subsequent merging. When the total duration of the invalid interval exceeds the unit's rotational inertia duration, the recursion restarts with the corresponding local contribution value in the first valid sampling interval before the invalid interval. For adjacent sampling intervals where the positive and negative directions change, the unit-side contribution value is allowed to gradually pass through the zero value according to the reverse update ratio, thus preserving the gradual change process when subsequent changes are transmitted to the current sampling interval. After completing all reverse recursions, the original time order of each sampling interval is restored, and the unit-side contribution value, contribution direction, sampling interval start and end time, disturbance change baseline value, and data validity status are arranged sequentially to obtain the unit-side contribution sequence.
[0102] The contribution value, contribution direction, start and end times of the sampling interval, and data validity status corresponding to the same sampling interval in the water flow-side contribution sequence and the unit-side contribution sequence are used as the merging objects. Each sampling interval is matched with one of the two contribution sequences, and only sampling intervals where both are valid and have the same contribution direction are retained. If both contribution values are greater than zero, they are determined to be positive and have the same sign; if both contribution values are less than zero, they are determined to be negative and have the same sign. The current continuous interval is interrupted if either contribution value is zero or if the two signs are different. Adjacent sampling intervals with the same sign are connected in chronological order. When there is no time gap between two adjacent sampling intervals and their contribution directions are consistent, they are included in the same continuous interval with the same sign. When there is an invalid interval, a time gap, or a change in contribution direction, the current continuous interval with the same sign ends and restarts. For each consecutive interval with the same sign, the absolute values of the flow-side contribution and the unit-side contribution are compared within each sampling interval. The smaller absolute value is taken as the bilateral support value for that sampling interval, maintaining the original same sign direction. Then, all bilateral support values within the consecutive interval with the same sign are compared, and the smallest bilateral support value is taken as the bottleneck value for that consecutive interval, ensuring that the bottleneck value does not exceed the flow-side support and unit-side support that can be simultaneously obtained at any sampling position within the interval. Consecutive intervals with the same sign whose duration is less than the smaller of the flow inertia duration and the unit rotational inertia duration are not considered valid candidate intervals. For the remaining consecutive intervals with the same sign, the absolute values of each bottleneck value are compared, and the bottleneck value with the largest absolute value is determined as the merged result, retaining the positive and negative directions of the corresponding consecutive interval with the same sign. When multiple bottleneck values have the same absolute value, the consecutive interval with the longer duration is preferred. If the durations are still the same, the consecutive interval with the larger cumulative disturbance change baseline value is preferred. If there is no consecutive interval with the same sign that meets the duration requirement, the merged result is set to zero, and the value and direction of the merged result are ultimately determined as the contribution coefficient of the target factor.
[0103] In this embodiment, residual reduction relationships are constructed by time-by-time displacement, and a continuous alignment process with monotonically maximum-minimum path constraints is adopted. This avoids time delay jumps caused by isolated high matching values, and ensures that orthogonal processing only applies to factor changes with overall consistency, thereby improving the continuity of joint anomaly analysis under variable time delay conditions.
[0104] In an exemplary embodiment, based on the current abnormal residual data, an abnormal residual reduction analysis is performed on the preset set of hydropower influencing factors of the hydropower system to obtain the target hydropower influencing factors, including steps 702 to 706. Wherein:
[0105] Step 702: Perform time-by-time displacement residual reduction calculation on the factor data based on the current abnormal residual data to obtain the time delay reduction field.
[0106] Step 704: Solve the monotonically maximum-minimum path for the delay reduction field to obtain the delay alignment path.
[0107] Step 706: Based on the time delay alignment path, perform path-constrained orthogonal projection on the factor data and the current abnormal residual data to obtain joint anomaly analysis data.
[0108] Among them, the time-by-time displacement residual reduction calculation is a process of moving factor data one by one according to different candidate time delays and calculating the degree of reduction of the current abnormal residual data before and after the deduction of the sampled value of each factor data.
[0109] Among them, the delay reduction field is a two-dimensional data structure that uses the sampling time of factor data and the sampling time of current abnormal residual data as two time dimensions, and records the residual reduction amount and corresponding delay at each corresponding position.
[0110] Among them, the monotonic maximum-minimum path solution is the process of selecting the continuous connection path with the largest minimum residual reduction along the way, under the condition that the time order of factor data and current abnormal residual data does not regress.
[0111] Among them, the time delay alignment path is the path composed of the correspondence between the sampling time of factor data arranged continuously in time order and the sampling time of the current abnormal residual data, as well as their time delay.
[0112] Among them, path-constrained orthogonal projection is a process of projecting the current abnormal residual data onto the direction of factor data and subtracting the corresponding projection components based solely on the sampling correspondence determined by the time-delay alignment path.
[0113] Specifically, the processing objects are the factor data corresponding to the hydropower influencing factors currently participating in the calculation within the preset set of hydropower influencing factors, as well as the sampling time, residual value, positive and negative direction of the residual, and data validity status in the current abnormal residual data. The factor data and the current abnormal residual data are adjusted to the same sampling interval, and the center position is corrected using the median of their respective valid values. Then, the absolute median difference is used to unify the numerical scale. When the absolute median difference is less than the preset minimum scale value, the preset minimum scale value is used to avoid abnormal amplification of the calculation results due to excessively small numerical fluctuations. Then, the sampling times of the factor data are set as row positions according to time sequence, and the sampling times of the current abnormal residual data are set as column positions, so that each row and column position corresponds to a factor data sampling time and an abnormal residual data sampling time, and the difference between the two sampling times is used as the time delay of that position. For locations where the time delay exceeds the allowable range of the corresponding hydropower influencing factors, any data is invalid, or the consecutive valid segments of the two sampling times are inconsistent, these locations are set as invalid. For the remaining locations, local calculation periods of equal length are set centered on the sampling times of the corresponding factor data and the abnormal residual data, and these are paired sequentially according to their order in their respective time series. Within each local calculation period, an amplitude adjustment parameter is set. The residual value of the current abnormal residual data is subtracted from the product of the amplitude adjustment parameter and the corresponding value of the factor data, and the squared difference is then used to assign weights according to the data's validity status. The amplitude adjustment parameter is adjusted successively until the sum of all weighted squared values no longer decreases. Subsequently, the total weighted square of the original residual values within the local calculation period is subtracted from the total weighted square of the adjusted difference, and then divided by the total weighted square of the original residual values to obtain the residual reduction amount for the corresponding row and column positions. If the residual reduction amount is less than zero, it is set to zero; if the proportion of valid records within the local calculation period is insufficient, the corresponding position is set as invalid. Finally, all rows and columns are arranged according to the correspondence between factor data sampling time, abnormal residual data sampling time, time delay, residual reduction amount, amplitude adjustment parameter and data validity status to obtain the time delay reduction field.
[0114] The path calculation uses the factor data sampling time, abnormal residual data sampling time, delay, residual reduction amount, and data validity status corresponding to each row and column position in the delay reduction field as the content. The earliest common valid position of the two time series is used as the starting range, and the latest common valid position of the two time series is used as the ending range. Starting from the starting range, the system only allows moving one sampling position forward according to the factor data time, one sampling position forward according to the abnormal residual data time, or both time periods moving one sampling position forward simultaneously. It does not allow any time period to return to a sampling position that has already been passed. At the same time, it limits the number of consecutive times that only one time period can move forward to avoid the same factor data sampling time corresponding to too many abnormal residual data sampling times, or the same abnormal residual data sampling time corresponding to too many factor data sampling times. Valid positions are processed sequentially in ascending order of the sum of the two time positions, and the minimum reduction hold value required to reach each position is recorded. The minimum reduction hold value for the starting position is the residual reduction amount for that position. For the remaining positions, all previous positions that are allowed to reach that position are checked, and the minimum reduction hold value recorded at the previous position is compared with the residual reduction amount at the current position. The smaller of the two values is taken as the minimum reduction hold value for reaching the current position via the previous position. If the current position has multiple valid previous positions, the maximum value among the minimum reduction hold values is used, and the previous position corresponding to that maximum value is recorded. If the results from multiple previous positions are the same, the one with the larger cumulative residual reduction amount up to the current position is used first. If the cumulative values are still the same, the one with the smaller time delay change between adjacent positions is used. After all valid positions have been calculated, the minimum reduction hold values of each position within the final range are compared. The final position with the largest value is taken as the path endpoint, and the path returns to the starting position step by step from the path endpoint according to the previous position recorded for each position. Then, the order of the positions traversed is reversed so that the sampling times of factor data and abnormal residual data are arranged from earliest to latest. Check each adjacent position to see if it meets the conditions of no time reversal, no more than the preset limit for the number of consecutive single-sided movements, and the time delay is within the allowable range. Then, combine the sampling time of the factor data that meets the conditions, the sampling time of the abnormal residual data, the corresponding time delay and the residual reduction amount in sequence to obtain the time delay alignment path.
[0115] The pairing constraints are the sampling times of factor data, abnormal residual data, time delay, and residual reduction amount corresponding to each item in the time delay alignment path. The calculation content includes the values and validity status of the factor data and the current abnormal residual data. Following the time order of the time delay alignment path, the factor value of each factor data sampling time is assigned to its corresponding abnormal residual data sampling time. When an abnormal residual data sampling time corresponds to only one factor data sampling time, the factor value is directly used. When an abnormal residual data sampling time corresponds to multiple consecutive factor data sampling times, weights are assigned according to the residual reduction amount at each corresponding position, and a weighted average is calculated. For cases where there are unpaired abnormal residual data sampling times between two adjacent corresponding positions in the time delay alignment path, weights are assigned according to the time distance between the unpaired sampling time and the two corresponding positions only when the length of the unpaired period does not exceed a preset allowable value and the preceding and following positions belong to the same continuous valid segment. Factor values with closer distances receive a larger weight; if the distance exceeds the preset allowable value, the corresponding sampling time is set to invalid. The weighted average of the paired factor values and the residual values of the current abnormal residual data are subtracted from their respective values, and the residual reduction at each position in the time delay alignment path is used as the calculation weight. Then, the factor values at each sampling time are multiplied by the residual values and summed according to the calculation weights. Simultaneously, the factor values at each sampling time are squared and summed with the same weight, and the ratio of the two sums determines the amplitude corresponding ratio. When the sum of the squared factor values is less than a preset minimum effective value, the amplitude corresponding ratio is set to zero; otherwise, the product of the amplitude corresponding ratio and the corresponding factor value is subtracted from the residual value of the current abnormal residual data at each effective sampling time to obtain the joint residual value for that sampling time. The weighted sum of the joint residual value and the corresponding factor value is calculated again. If the weighted sum does not fall within the preset allowable range due to numerical precision issues, its last digit is adjusted without changing the direction of the amplitude corresponding ratio until it meets the allowable range. The joint residual values, sampling time, positive and negative directions of the residuals, corresponding positions of the time delay alignment path, and data validity status are combined to form joint anomaly analysis data. After performing the above calculations on each hydropower influencing factor in the preset set of hydropower influencing factors, the residual reduction ratio of each joint anomaly analysis data relative to the current anomaly residual data is compared. The hydropower influencing factor with the largest residual reduction ratio that is greater than the preset valid threshold is determined as the target hydropower influencing factor, and the joint anomaly analysis data corresponding to the target hydropower influencing factor is retained.
[0116] In this embodiment, by pairing hourly displacement with monotonic path, the sequential relationship of factors can be maintained when sampling changes are asynchronous. Furthermore, the orthogonal projection with path constraints prevents the same abnormal residual segment from being counted repeatedly, thus providing a consistent basis for comparison of the residual reduction results of different hydropower influencing factors.
[0117] In an exemplary embodiment, a monotonically maximum-minimum path solution is performed on the delay reduction field to obtain the delay alignment path, including steps 802 to 806. Wherein:
[0118] Step 802: With the constraint that neither the time order of the factor data nor the time order of the abnormal residual data should be backwards, forward connectivity is performed on the time delay reduction field to obtain the monotonic reachable domain.
[0119] Step 804: Perform a maximum-minimum bottleneck recursion on the monotonically reachable domain to obtain the path bottleneck field.
[0120] Step 806: Perform reverse backtracking on the path bottleneck field under the constraint of the maximum bottleneck value at the end to obtain the time-delay aligned path.
[0121] Forward connectivity is a process that starts from the starting position of the time delay reduction field and connects adjacent valid positions only along the direction in which neither the time of factor data nor the time of abnormal residual data goes back.
[0122] Among them, the monotonic reachable region is the region consisting of all valid positions and their connection relationships that can be continuously reached from the starting position of the delay reduction field according to the forward connectivity rule.
[0123] The maximum-minimum bottleneck recursion is a process in which the smaller value between the current position residual reduction and the bottleneck value of the predecessor path is used at each position, and the maximum value is selected from multiple predecessor corresponding results for position-by-position update.
[0124] The path bottleneck field is a two-dimensional data structure formed by recording the optimal path bottleneck value and its corresponding predecessor position at each position in the monotonic reachability domain.
[0125] Among them, the reverse backtracking under the constraint of the maximum bottleneck value at the end is the process of selecting the position with the largest bottleneck value at the end of the path bottleneck field and returning to the starting position along the corresponding predecessor position to form a time delay aligned path.
[0126] Specifically, the row positions of the delay reduction field are set from earliest to latest according to the sampling time of the factor data, and the column positions of the delay reduction field are set from earliest to latest according to the sampling time of the abnormal residual data. Each position corresponds to one factor data sample value and one abnormal residual data sample value, and the time difference, residual reduction amount, and validity status of that position are retained. Then, positions where the time difference is within a preset allowable range, the residual reduction amount reaches a preset lower limit, and both sample values are valid are designated as allowed positions. Positions with missing samples, excessive time differences, or residual reduction amounts not reaching the preset lower limit are designated as unavoidable positions. Allowed positions are processed sequentially in ascending order of row and column positions. For any current position, only positions where the factor data time remains unchanged while the abnormal residual data time advances backward, positions where the abnormal residual data time remains unchanged while the factor data time advances backward, or positions where both times advance backward simultaneously are allowed to be reached. Returning from a later position to an earlier position is not allowed. To prevent the same factor data sample value from corresponding to too many consecutive abnormal residual data sample values, or vice versa, an upper limit is set on the number of consecutive advances in only one time dimension; connections exceeding this limit are not retained. When processing the initial region, locations near the first valid sampling time of both factor data and abnormal residual data, where the time difference meets the allowable range, are set as reachable locations. When processing subsequent locations, the current location is only set as reachable if it has at least one preceding reachable location, and the time advancement rules described above are satisfied. All locations are processed in a synchronous backward order along both time dimensions, and at each reachable location, the preceding location that can be entered and the corresponding advancement direction are recorded. All reachable locations and their preceding and following connections form a monotonic reachability domain.
[0127] Each reachable location is processed according to the time progression order in the monotonic reachability region, ensuring that all allowed preceding locations have been calculated before a location is included in the calculation. For reachable locations in the starting region, the residual reduction amount in the delay reduction field for that location is directly used as the initial bottleneck value. For the remaining reachable locations, each preceding location that can reach the current location is checked. For each preceding location, the bottleneck value already recorded at that preceding location is compared with the residual reduction amount of the current location, and the smaller of the two is used as the bottleneck value that can be maintained when reaching the current location via that preceding location. When there are multiple preceding locations, the bottleneck values corresponding to each preceding location are compared, and the largest one is used as the bottleneck value of the current location. Thus, the calculation result of the current location will not exceed the allowed residual reduction amount of any location along the way, and the connection with the largest minimum residual reduction amount along the way is preferentially retained. If multiple preceding positions produce the same bottleneck value, the cumulative residual reduction, the change in time difference between adjacent positions, and the number of consecutive advances in only one time dimension are compared sequentially. The preceding position with the larger cumulative residual reduction, smaller change in time difference, and fewer consecutive advances in one dimension is prioritized. If these conditions are still the same, the preceding position that advances simultaneously in both time dimensions is retained. For regions without a valid preceding position, no bottleneck value is set, and they are not allowed to participate in the calculation of subsequent positions. After processing all positions within the monotonic reachability domain, the bottleneck value, the retained preceding position, the cumulative residual reduction, and the change in time difference are recorded at the corresponding positions. These calculation results, interconnected in chronological order, form the path bottleneck field.
[0128] Within the terminal region of the path bottleneck field, bottleneck values at valid positions are compared sequentially. This terminal region includes positions in the row containing the last valid sampling time of factor data that meet the time difference requirement, and positions in the column containing the last valid sampling time of abnormal residual data that meet the time difference requirement. The largest bottleneck value is used as the terminal constraint value, and the position reaching this terminal constraint value is used as the starting position for reverse backtracking. When multiple starting positions exist, the position with the larger cumulative residual reduction, the smaller deviation from the preset allowable time difference center value, and the more complete coverage of the sampling range of factor data and abnormal residual data is selected. Starting from the selected terminal position, the process returns sequentially according to the previous position recorded in the path bottleneck field. Each return checks whether the bottleneck value of the previous position is not less than the terminal constraint value, whether the two positions before and after meet the time non-backtracking requirement, and whether the corresponding time difference is still within the preset allowable range. If any condition is not met, the process returns to the most recent position with a parallel previous position and uses the next matching previous position to continue processing. Upon reaching the starting region of the monotonic reachable domain, the positions traversed during the backtracking process are rearranged in reverse order. Each adjacent position is checked to ensure it uses only the three allowed time advancement methods, that the number of consecutive advancements in a single time dimension does not exceed the upper limit, and that the path does not contain any traversable positions. After all checks pass, the sampling times of the factor data and the abnormal residual data are matched one by one according to the rearrangement results. The time difference, residual reduction, and bottleneck value of each corresponding position are retained. These corresponding positions, which are sequentially continuous and do not undergo backtracking, form a time-delay aligned path.
[0129] In this embodiment, by limiting the time order of factor data and abnormal residual data to not backtrack, and using the path with the largest minimum reduction capability along the way as the backtracking basis, it is possible to avoid misalignment caused by local high reduction positions, so that the obtained time delay alignment path maintains time continuity and has stable overall reduction capability.
[0130] In an exemplary embodiment, a maximum-minimum bottleneck recursion is performed on the monotonically reachable domain to obtain the path bottleneck field, including steps 902 to 908. Wherein:
[0131] Step 902: Based on the residual reduction amount of the initial boundary of the delay reduction field, assign an initial bottleneck value to the monotonic reachable domain to obtain the initial bottleneck boundary.
[0132] Step 904: Based on the time order of the factor data and the time order of the abnormal residual data, limit the predecessor range of each position in the monotonic reachable domain to obtain each predecessor position.
[0133] Step 906: Calculate the smaller value of the residual reduction amount at the current position in the delay reduction field and the path bottleneck value at each predecessor position to obtain each candidate bottleneck value.
[0134] Step 908: According to the rule of the candidate bottleneck value being the largest and the adjacent time delay change being the smallest, the bottleneck value is recursively calculated for each position in the monotonic reachable domain to obtain the path bottleneck field.
[0135] The initial boundary is a set of continuous positions in the monotonic reachable domain that correspond to the first valid sampling time of the factor data or the first valid sampling time of the abnormal residual data and can be used as the starting point of the path.
[0136] The initial bottleneck assignment is a process of setting the residual reduction amount at the starting position as the corresponding path bottleneck value, and then retaining the smaller value between the previous path bottleneck value and the current position residual reduction amount along the starting boundary.
[0137] The initial bottleneck boundary is the boundary composed of all the initial boundary positions that record the path bottleneck value and adjacent connection relationships after the initial bottleneck assignment is completed.
[0138] Here, the location is a two-dimensional corresponding unit in the delay reduction field, which is jointly determined by the sampling time of a factor data and the sampling time of an abnormal residual data.
[0139] Among them, the predecessor range limitation is the process of restricting the earlier adjacent positions that can be directly connected to the current position to an allowable range, according to the requirement that neither of the two time sequences should be backdated.
[0140] Among them, the precursor position is a position located within the monotonic reachability domain, whose time sequence is no later than the current position, and which can directly reach the current position in the allowed direction of advancement.
[0141] Among them, residual reduction is the amount of residual energy reduction in the current abnormal residual data before and after the factor data participates in the anomaly interpretation under the corresponding time delay.
[0142] The path bottleneck value is the maximum value of the lowest residual reduction among the feasible paths from the starting boundary to the current position.
[0143] The smaller value calculation involves comparing the residual reduction at the current position with the path bottleneck value at a predecessor position and using the smaller of the two values.
[0144] The candidate bottleneck value is the minimum residual reduction of the path calculated from the smaller value when the current position is reached via each predecessor position.
[0145] Among them, the minimum change in adjacent time delay means that when the candidate bottleneck values are the same, the connection with the smallest absolute amount of time delay difference between the current position and the predecessor position is preferred.
[0146] The bottleneck value recursion is a process of selecting the largest candidate bottleneck value as the current path bottleneck value in chronological order and using this value for calculations at subsequent positions.
[0147] Specifically, the row order of the delay reduction field is set according to the sampling time of the factor data, and the column order is set according to the sampling time of the abnormal residual data. The position that earliest simultaneously satisfies the time order constraint, delay range constraint, and effective state requirement within the monotonic reachability domain is taken as the starting position. For this starting position, the residual reduction amount recorded in the delay reduction field at the corresponding position is directly used as the path bottleneck value. For other boundary positions that are continuously connected to this starting position along the first row or first column, the path bottleneck value already determined at the previous boundary position is compared with the residual reduction amount of the current boundary position, and the smaller value is assigned to the current boundary position to ensure that the path bottleneck value of the boundary position does not exceed the lowest residual reduction amount among the previously passed positions. If a boundary position does not belong to the monotonic reachability domain, the corresponding residual reduction amount is invalid, or there is no allowed connection between it and the previous boundary position, the assignment along that direction is stopped to avoid establishing boundary connections across unreachable areas. After completing the assignment of all effective boundary positions, each boundary position, the corresponding residual reduction amount, the path bottleneck value, the delay, and the adjacent connection relationship together form the initial bottleneck boundary.
[0148] Positions within the monotonic reachability domain are arranged from earliest to latest according to the sampling times of the factor data and the abnormal residual data, with the fundamental condition that neither time dimension can be backtracked. For any current position that is not part of the initial bottleneck boundary, allowed positions as predecessors include: positions where the factor data is at the previous valid sampling time and the abnormal residual data is at the current sampling time; positions where the factor data is at the current sampling time and the abnormal residual data is at the previous valid sampling time; and positions where both the factor data and the abnormal residual data are at their respective previous valid sampling times. Each of these positions is checked to see if it is within the monotonic reachability domain, if it already has a path bottleneck value, and if the sampling interval and time delay changes between it and the current position meet the preset range. If any condition is not met, the corresponding position is not included in the predecessor range of the current position. For factor data and abnormal residual data with uneven sampling intervals, the temporally adjacent valid sampling times are used as the standard, and skipping still valid intermediate sampling times to directly establish a connection is not allowed. Each position is processed in ascending order of the sum of the two time numbers, so that each predecessor position corresponding to the current position is calculated before the current position. Finally, zero, one, or more predecessor positions are determined for each non-boundary position in the monotonic reachable domain.
[0149] For the current position in the monotonically reachable domain, the residual reduction amount recorded at that position in the delay reduction field is called, and each residual reduction amount is compared with the path bottleneck value already determined for each predecessor position. If the current residual reduction amount is less than the path bottleneck value of a predecessor position, the current residual reduction amount is used as a candidate bottleneck value for reaching the current position via that predecessor position; if the current residual reduction amount is greater than the path bottleneck value of that predecessor position, the path bottleneck value of that predecessor position is used as a candidate bottleneck value; if the two are equal, the same value is directly used. During the calculation, the residual reduction amount of the current position is not summed or averaged with the path bottleneck value of the predecessor position, so that each candidate bottleneck value always corresponds to the lowest residual reduction level in the entire connection from the corresponding predecessor position to the current position. When there are multiple predecessor positions, the one-to-one correspondence between each candidate bottleneck value and its corresponding predecessor position is maintained; if the predecessor position does not have a valid path bottleneck value configured or the residual reduction amount of the current position is invalid, no corresponding candidate bottleneck value is generated, thus obtaining the candidate bottleneck values corresponding to the current position.
[0150] The path bottleneck value for each location in the monotonic reachability domain is processed sequentially according to the time order of factor data and abnormal residual data. The candidate bottleneck value with the largest value is selected as the path bottleneck value for the current location. When multiple predecessor locations produce the same maximum candidate bottleneck value, the absolute difference between the current location's delay and the delays of each predecessor location is calculated, and the predecessor location with the smallest absolute difference is selected first to ensure a smooth delay change between adjacent locations. When both the maximum candidate bottleneck value and the adjacent delay changes are the same, the predecessor location whose time is simultaneously advanced from the factor data time and the abnormal residual data time is selected first. If they are still the same, all predecessor locations that meet the conditions are retained. After the path bottleneck value for the current location is determined, the corresponding predecessor location, the current location's residual reduction, the current location's delay, and the adjacent delay changes are recorded synchronously. This path bottleneck value is then used for calculating later locations in the time order. For locations without candidate bottleneck values, no path bottleneck value is configured, and the location is not allowed to connect to subsequent locations. After all locations in the monotonic reachability domain have been recursively processed, the path bottleneck field is formed by the path bottleneck values of each location and their corresponding connection relationships.
[0151] In this embodiment, by retaining the lowest residual reduction amount in the path position by position, and prioritizing the use of adjacent connections with smaller delay changes when candidate values are the same, it is possible to prevent large local reduction amounts from masking weak positions, and at the same time reduce the interference of delay abrupt changes on path bottleneck judgment.
[0152] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise expressly stated herein, there is no strict order restriction on the execution of these steps, and they may be executed in other orders.
[0153] Based on the same inventive concept, this application also provides a hydropower anomaly analysis device for implementing the above-mentioned hydropower anomaly analysis method. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations of the one or more hydropower anomaly analysis device embodiments provided below can be found in the limitations of the hydropower anomaly analysis method described above, and will not be repeated here.
[0154] In one exemplary embodiment, such as Figure 3 As shown, the device includes: an anomaly data analysis module 302, an influencing factor identification module 304, an anomaly joint analysis module 306, and a hydropower cycle analysis module 308. Each module in the aforementioned hydropower anomaly analysis device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.
[0155] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. This computer device includes a processor, memory, input / output interfaces (I / O), and communication interfaces.
[0156] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0157] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.
[0158] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0159] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.
[0160] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0161] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.
[0162] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0163] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for analyzing abnormal hydropower situations, characterized in that, The method includes: A transfer delay residual analysis based on the normal water energy input-electricity output conversion relationship is performed on the abnormal water energy data of the hydropower system to obtain the current abnormal residual data; Based on the current abnormal residual data, an abnormal residual reduction analysis is performed on the preset set of hydropower influencing factors of the hydropower system to obtain the target hydropower influencing factors. A time-delay aligned orthogonal joint analysis is performed on the factor data corresponding to the target hydropower influencing factors and the current abnormal residual data to obtain joint anomaly analysis data. The joint anomaly analysis data is used as the current anomaly residual data. The process of performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current anomaly residual data is repeated until each of the target hydropower influencing factors forms a complete influence path from the hydropower input end to the power output end in the preset hydropower coupling topology, thus obtaining the target anomaly analysis data.
2. The method according to claim 1, characterized in that, The time-delay aligned orthogonal joint analysis of the factor data corresponding to the target hydropower influencing factors and the current abnormal residual data yields joint anomaly analysis data, including: Based on the transmission path of the target hydropower influencing factors to the power output end in the preset hydropower coupling topology, the factor data is calculated stepwise gain transfer to obtain equivalent power disturbance data. Based on the cumulative transmission time of the transmission path, the equivalent power disturbance data and the current abnormal residual data are aligned with the energy centroid time delay to obtain time delay aligned disturbance data; Based on the inertial duration of the water flow and the rotational inertial duration of the hydropower system, a dual-inertial orthogonal joint analysis is performed on the time delay alignment disturbance data and the current abnormal residual data to obtain the joint anomaly analysis data.
3. The method according to claim 2, characterized in that, The method involves performing a dual-inertial orthogonal joint analysis on the time delay alignment disturbance data and the current abnormal residual data based on the water flow inertia duration and the unit rotation inertia duration of the hydropower system, to obtain the joint anomaly analysis data, including: Perform a dual-inertial state transition on the time-delay aligned perturbation data and the current abnormal residual data to obtain the perturbation dual-inertial state trajectory and the residual dual-inertial state trajectory; The inner product of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory is calculated to obtain the contribution coefficient of the target factor. Based on the target factor contribution coefficient, the factor contribution corresponding to the time delay alignment disturbance data is subtracted from the current abnormal residual data to obtain the joint anomaly analysis data.
4. The method according to claim 3, characterized in that, The step of performing a dual-inertial state transition on the time-delay aligned perturbation data and the current abnormal residual data to obtain the perturbation dual-inertial state trajectory and the residual dual-inertial state trajectory includes: Based on the inertial duration of the water flow, the time-delay aligned disturbance data and the current abnormal residual data are respectively subjected to first-order inertial recursion to obtain the disturbed water flow state sequence and the residual water flow state sequence. Based on the rotational inertia duration of the unit, the first-order inertial recursion is performed on the disturbed water flow state sequence and the residual water flow state sequence to obtain the disturbed rotational state sequence and the residual rotational state sequence. Hourly directed difference calculations are performed on the disturbed flow state sequence, the disturbed rotation state sequence, the residual flow state sequence, and the residual rotation state sequence to obtain the disturbed inertial conversion difference sequence and the residual inertial conversion difference sequence. Based on the inertial duration of the water flow and the rotational inertial duration of the unit, the disturbance inertial conversion difference sequence and the residual inertial conversion difference sequence are respectively subjected to dual-duration delay coordinate embedding to obtain the disturbance dual-inertial state trajectory and the residual dual-inertial state trajectory.
5. The method according to claim 3, characterized in that, The calculation of the inner product of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory to obtain the target factor contribution coefficient includes: The local contribution sequence is obtained by calculating the local inner product of the state displacements of the disturbed dual-inertial state trajectory and the residual dual-inertial state trajectory in each of the same sampling intervals; Based on the water flow inertia duration, the local contribution sequence is recursively calculated in the forward inertia to obtain the water flow side contribution sequence. Based on the unit's rotational inertia duration, the local contribution sequence is recursively extrapolated in reverse to obtain the unit-side contribution sequence. The target factor contribution coefficient is obtained by performing a directed maximum-minimum bottleneck convergence calculation under the same-sign continuous interval constraint on the water flow side contribution sequence and the unit side contribution sequence.
6. The method according to claim 1, characterized in that, The step involves performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current abnormal residual data to obtain the target hydropower influencing factors, including: The factor data is subjected to hourly displacement-based residual reduction calculation on the current abnormal residual data to obtain the time delay reduction field; The time delay reduction field is solved by monotonically maximum-minimum path calculation to obtain the time delay alignment path; Based on the time delay alignment path, the factor data and the current abnormal residual data are subjected to path-constrained orthogonal projection to obtain the joint anomaly analysis data.
7. The method according to claim 6, characterized in that, The step of performing a monotonically maximum-minimum path solution on the time delay reduction field to obtain the time delay alignment path includes: With the constraint that neither the time order of the factor data nor the time order of the abnormal residual data can be backtracked, the time delay reduction field is forward connected to obtain a monotonically reachable domain. The path bottleneck field is obtained by performing a maximum-minimum bottleneck recursion on the monotonically reachable domain. The time-delay aligned path is obtained by performing reverse backtracking under the constraint of the maximum bottleneck value at the end of the path bottleneck field.
8. The method according to claim 7, characterized in that, The step of performing a maximum-minimum bottleneck recursion on the monotonically reachable domain to obtain the path bottleneck field includes: Based on the residual reduction amount of the initial boundary of the time delay reduction field, the initial bottleneck value is assigned to the monotonic reachable domain to obtain the initial bottleneck boundary. Based on the time order of the factor data and the time order of the abnormal residual data, the predecessor range of each position in the monotonic reachability domain is defined to obtain each predecessor position. The residual reduction amount at the current position in the delay reduction field and the path bottleneck value at each of the predecessor positions are calculated by taking the smaller value to obtain each candidate bottleneck value; According to the rule that the candidate bottleneck value is the largest and the adjacent time delay change is the smallest, the bottleneck value is recursively calculated for each position in the monotonic reachable domain to obtain the path bottleneck field.
9. A device for analyzing abnormal hydropower conditions, characterized in that, The device includes: The abnormal data analysis module is used to perform transmission delay residual analysis on abnormal water energy data of the hydropower system based on the normal water energy input-electricity output conversion relationship, and obtain the current abnormal residual data. The influencing factor identification module is used to perform abnormal residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current abnormal residual data, so as to obtain the target hydropower influencing factors. The anomaly joint analysis module is used to perform time-delay aligned orthogonal joint analysis on the factor data corresponding to the target hydropower influencing factors and the current anomaly residual data to obtain joint anomaly analysis data. The hydropower cycle analysis module is used to take the joint anomaly analysis data as the current anomaly residual data and return to execute the step of performing anomaly residual reduction analysis on the preset set of hydropower influencing factors of the hydropower system based on the current anomaly residual data to obtain the target hydropower influencing factors, until each of the target hydropower influencing factors forms a complete influence path from the hydropower input end to the power output end in the preset hydropower coupling topology, and the target anomaly analysis data is obtained.
10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.
