A data analysis-based intelligent fault diagnosis method for photovoltaic fuses

Abstract

The application discloses a photovoltaic fuse intelligent fault diagnosis method based on data analysis, which comprises the following steps: obtaining the operation data of a target photovoltaic fuse and recording a time stamp; performing time alignment and association on the operation data to form a photovoltaic fuse operation data sequence; extracting state characteristics and generating a photovoltaic fuse state characteristic sequence; performing data analysis and constructing a semi-ordered degradation state space with degradation order constraints; calculating a degradation intensity parameter and determining a state transition constraint rule; modeling the transition process to construct a semi-ordered Markov degradation model; deducing the operation state of the photovoltaic fuse to generate a degradation state evolution path; based on the degradation state evolution path, judging the evolution consistency and evolution trend of the degradation state of the photovoltaic fuse, and outputting the fault diagnosis result of the photovoltaic fuse. The application adopts semi-ordered Markov degradation modeling, realizes the degradation path diagnosis of the fuse, and has the advantages of noise resistance and early warning.

Description

Technical Field

[0001] This invention relates to the field of intelligent fault diagnosis technology for photovoltaic fuses, and in particular to an intelligent fault diagnosis method for photovoltaic fuses based on data analysis. Background Technology

[0002] As a crucial component of distributed renewable energy, photovoltaic (PV) power generation systems typically utilize fuses for overcurrent protection on their DC side to prevent damage to modules and lines from fault currents. Existing PV fuse operation monitoring and fault diagnosis technologies primarily rely on threshold judgments based on single or limited operating parameters such as current, voltage, or temperature, or on classifying and identifying the characteristics of the operating state at a specific moment to determine whether the fuse has malfunctioned. While some existing technologies incorporate data analysis or state recognition models, they are mostly based on instantaneous state determinations and lack a systematic characterization of the degradation behavior of fuses during long-term operation.

[0003] However, in actual operation, the performance degradation of photovoltaic fuses is usually gradual and irreversible, with the operating state exhibiting a continuous evolution over time. Existing technologies generally lack structured constraints on the sequence of degradation states, making it difficult to avoid misjudgments caused by noise fluctuations or short-term operating condition changes. Furthermore, existing diagnostic methods based on Markov or statistical models often employ fixed transition assumptions or simple health index classifications, failing to effectively combine the degree of cumulative degradation with state transition rules, resulting in insufficient timeliness and reliability in identifying potential failure risks. Summary of the Invention

[0004] One objective of this invention is to propose an intelligent fault diagnosis method for photovoltaic fuses based on data analysis. This invention uses semi-sequential Markov degradation modeling to realize fuse degradation path diagnosis and has the advantage of noise resistance and early warning.

[0005] A data analysis-based intelligent fault diagnosis method for photovoltaic fuses according to an embodiment of the present invention includes the following steps:

[0006] Acquire the operating data of the target photovoltaic fuse in the photovoltaic power generation system and record the corresponding timestamps;

[0007] The operating data is time-aligned and associated based on a unified time index to form a photovoltaic fuse operating data sequence.

[0008] State features are extracted from photovoltaic fuse operation data sequences, and a photovoltaic fuse state feature sequence is generated.

[0009] Data analysis of the operating status of photovoltaic fuses is performed based on the state characteristic sequence of photovoltaic fuses, and multiple degradation state intervals are divided. A semi-order degradation state space with degradation sequence constraints is constructed according to the state constraint relationship.

[0010] In the semi-order degraded state space, the degradation intensity parameter is calculated and associated with the boundary conditions of the degradation state interval to determine the state transition constraint rules of the photovoltaic fuse operation state in the semi-order degraded state space.

[0011] Based on the state transition constraint rules, the transition process of the photovoltaic fuse's operating state in the semi-order degenerate state space is modeled, and a semi-order Markov degenerate model is constructed.

[0012] Based on the semi-order Markov degradation model, the operating state of photovoltaic fuses is deduced, and the degradation state evolution path is generated.

[0013] Based on the degradation state evolution path, the consistency and trend of the degradation state evolution of photovoltaic fuses are determined, and the fault diagnosis results of photovoltaic fuses are output.

[0014] Optionally, the generation of the photovoltaic fuse operation data sequence includes:

[0015] Based on the timestamps corresponding to current data, temperature data, and operating condition data, the operating data undergoes unified time base processing to determine a unified time index covering the preset monitoring time window;

[0016] Based on a unified time index, timestamp alignment is performed on current data, temperature data, and operating condition data.

[0017] Truncation is performed on current data, temperature data, and operating condition data whose timestamps are earlier than the start position of the unified time index or later than the end position of the unified time index.

[0018] For current data, temperature data, and operating condition data that have missing index positions under a unified time index, perform completion processing;

[0019] The current data, temperature data, and operating condition data corresponding to the same time index are associated to form associated data corresponding to a unified time index.

[0020] The associated data are arranged according to the chronological order of a unified time index, forming a sequence of photovoltaic fuse operation data arranged in chronological order.

[0021] Optionally, the generation of the photovoltaic fuse state characteristic sequence includes:

[0022] Read the corresponding current data, temperature data and operating condition data in the photovoltaic fuse operation data sequence in chronological order to form an operation data set that corresponds one-to-one with the chronological order.

[0023] Based on the running data set, state features that correspond one-to-one with the time sequence are generated.

[0024] Within a preset time window, based on the state features corresponding to adjacent time sequences, the difference between the state features corresponding to the previous time sequence and the state features corresponding to the next time sequence is calculated to form the state feature change results corresponding to the time sequence.

[0025] Based on the state characteristics and the results of state characteristic changes, the state characteristics are combined in chronological order, and the combined state characteristics are arranged in chronological order to generate a photovoltaic fuse state characteristic sequence.

[0026] Optionally, constructing a semi-order degenerate state space with degenerate order constraints includes:

[0027] Using the continuous time index in the photovoltaic fuse state feature sequence as the analysis order, statistical processing is performed on the state features in the photovoltaic fuse state feature sequence to generate state feature distribution results;

[0028] Based on the state feature distribution results, the distribution change results between adjacent state feature distribution results are calculated, and the distribution change results are arranged in the order of continuous time index to generate a distribution change sequence.

[0029] The continuous time index positions that meet the preset change conditions in the distribution change sequence are determined, and the continuous time index positions are used as interval division positions to divide the photovoltaic fuse state characteristic sequence and generate multiple degradation state intervals.

[0030] Perform interval statistical processing on the state features within each degenerate state interval to generate interval state features that correspond one-to-one with each degenerate state interval;

[0031] According to the direction of change of each degenerate state interval in time sequence, the degeneracy order among multiple degenerate state intervals is determined, and state constraint relationships are constructed among multiple degenerate state intervals based on the degeneracy order, thus constructing a semi-order degenerate state space with degeneracy order constraints.

[0032] Optionally, the determination of the state transition constraint rules includes:

[0033] In the semi-sequence degradation state space, the current data, temperature data and operating condition data corresponding to each time position are read sequentially based on the time sequence. The change amplitude of the current data and the change rate of the temperature data are calculated. At the same time, the values ​​of the operating condition data are obtained, and the degradation intensity parameter calculation results corresponding to the time sequence are generated.

[0034] Based on the calculation results of the degradation intensity parameters, cumulative calculations are performed on the degradation intensity parameters in chronological order within a preset time window to generate degradation intensity parameters;

[0035] Based on multiple degenerate state intervals in the semi-order degenerate state space, the corresponding interval boundary conditions are determined for each degenerate state interval, and a correspondence is established between each degenerate state interval and its corresponding interval boundary conditions.

[0036] The degradation intensity parameter is associated with the interval boundary conditions corresponding to the current degradation state interval. Based on the positional relationship between the degradation intensity parameter and the interval boundary conditions, the state transition constraint rules of the photovoltaic fuse operating state in the semi-sequence degradation state space are determined.

[0037] Optionally, the construction of the semi-order Markov degenerate model includes:

[0038] In the semi-order degradation state space, each degradation state interval is taken as the state set of the photovoltaic fuse operation state, and the corresponding subsequent degradation state interval is determined for each degradation state interval according to the degradation order.

[0039] Based on the state set and the subsequent degradation state interval, and in accordance with the state transition constraint rules, the permissible changes in the operating state of the photovoltaic fuse in the semi-sequence degradation state space are limited, forming a set of state changes limited by the state transition constraint rules.

[0040] Based on the set of state changes, the maintenance behavior within each degenerate state interval and the transition behavior to the subsequent degenerate state interval are statistically analyzed, generating the number of maintenance occurrences and the number of transition occurrences corresponding one-to-one with each degenerate state interval;

[0041] Based on the number of times maintenance occurs and the number of times transition occurs, the maintenance probability value corresponding to each degradation state interval is determined, and the transition probability value corresponding to each degradation state interval is also determined.

[0042] Based on the set of state changes, according to the order of changes in the operating state of the photovoltaic fuse in the semi-order degradation state space, the continuous holding time of the operating state of the photovoltaic fuse in each degradation state interval is statistically analyzed and generated to generate a distribution of holding duration values ​​that correspond one-to-one with each degradation state interval.

[0043] Based on the values ​​of the hold probability, the transition probability, and the hold duration distribution, a semi-order Markov degradation model is constructed to model the hold process of the photovoltaic fuse's operating state and the transition process to the subsequent degradation state interval in the semi-order degradation state space.

[0044] Optionally, the generation of the degradation state evolution path includes:

[0045] Based on the semi-order Markov degradation model, the starting continuous time index in the continuous time index is determined, and the starting degradation state interval of the photovoltaic fuse operation state corresponding to the starting continuous time index is determined as the starting state of the degradation state evolution path.

[0046] Using the initial continuous time index as the current continuous time index and the initial degradation state interval as the current degradation state interval, the photovoltaic fuse operating state under the current continuous time index is deduced based on the state description corresponding to the current degradation state interval in the semi-order Markov degradation model, and the corresponding next degradation state interval is obtained.

[0047] Associating the next degenerate state interval with its corresponding subsequent continuous time index forms a state pair consisting of a continuous time index and a degenerate state interval. The state pairs are then recorded in the order of the continuous time indexes to generate the degenerate state evolution path.

[0048] The state pairs are verified based on the semi-order Markov degeneration model. When there is a reverse degeneration transition between the next degeneration state interval and the current degeneration state interval that does not conform to the degeneration direction constraint, the next degeneration state interval is corrected to the current degeneration state interval, and the degeneration state evolution path is updated with the corrected state pairs.

[0049] The revised next degenerate state interval is used as the new current degenerate state interval, and the continuous time index is used as the new current continuous time index. The process of updating the degenerate state evolution path is repeated until the preset termination continuous time index is reached, thus generating the degenerate state evolution path.

[0050] Optionally, the generation of the fault diagnosis results includes:

[0051] Based on the evolution path of the degenerate state, a preset time range corresponding to the evolution path of the degenerate state is determined, and a sequence of degenerate states located within the preset time range and arranged in chronological order is extracted from the evolution path of the degenerate state as the evolution path of the degenerate state for judgment.

[0052] Based on the evolution path of the degradation state used for judgment, the degradation states corresponding to adjacent time positions are compared in chronological order. The number of times the degradation state remains unchanged within a preset time range and the number of times the degradation state changes along the degradation direction are counted. The total number of degradation state changes is calculated based on the number of times the degradation state remains unchanged and the number of times the degradation state changes along the degradation direction.

[0053] Based on the number of hold times, the number of transfer times, and the total number of degradation state changes, the evolution consistency of the degradation state of photovoltaic fuses within a preset time range is determined according to the ratio between the number of hold times and the total number of degradation state changes, as well as the ratio between the number of transfer times and the total number of degradation state changes, and an evolution consistency determination result is generated.

[0054] Based on the degradation state evolution path used for judgment, the changes in the degradation state along the degradation direction within a preset time range are accumulated in chronological order to obtain the cumulative change result of the degradation state within the preset time range. Based on the ratio between the cumulative change result and the time length of the preset time range, the evolution trend of the photovoltaic fuse degradation state within the preset time range is judged, and the evolution trend judgment result is generated.

[0055] Based on the results of evolution consistency determination and evolution trend determination, the degradation status of photovoltaic fuses within a preset time range is comprehensively determined, and the fault diagnosis results of photovoltaic fuses are generated and output.

[0056] The beneficial effects of this invention are:

[0057] This invention introduces a degradation modeling approach based on operational data analysis, elevating the assessment of photovoltaic fuse status changes during long-term operation from traditional "instantaneous judgment" to "evolutionary process judgment." This approach more realistically and stably reflects the actual health status of the fuse. By performing correlation analysis on current, temperature, and operating condition data under a unified time index, a time-ordered sequence of operational data and state characteristic sequences is constructed. This ensures the continuity and traceability of fuse operating status changes, fundamentally avoiding misjudgments caused by abnormal fluctuations at a single moment in existing technologies. This provides a reliable data foundation for subsequent degradation analysis and fault diagnosis.

[0058] Furthermore, this invention analyzes the distribution and variation of state characteristics under continuous-time indexing to construct a semi-order degenerate state space with degradation sequence constraints. This explicitly limits the fuse's operating state to a structured state system that only allows it to remain unchanged or evolve along the degradation direction, eliminating unreasonable reverse state jumps at the model level. Based on the correlation between degradation intensity parameters and boundary conditions of degradation state intervals, this invention implements state transition constraint rules driven by the degree of degradation accumulation. This ensures that state evolution no longer depends on fixed transition probabilities or simple threshold judgments, but rather remains consistent with the actual degradation process of the fuse, thereby improving the engineering rationality and stability of the diagnostic results.

[0059] Based on this, by constructing a semi-order Markov degradation model and generating degradation state evolution paths that do not include reverse degradation transitions, this invention can characterize the degradation behavior of fuses at the path level and comprehensively determine the consistency and trend of degradation state evolution within a preset time range, thereby achieving early identification of potential fault risks. Compared with existing technologies, this invention not only improves the accuracy and noise immunity of photovoltaic fuse fault diagnosis, but also enhances the ability to express gradual and irreversible degradation processes, resulting in higher diagnostic reliability and stronger engineering applicability. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is an overall flowchart of a data analysis-based intelligent fault diagnosis method for photovoltaic fuses proposed in this invention.

[0062] Figure 2 This is a schematic diagram of a semi-order degenerate state space with degradation sequence constraints constructed based on the distribution change of state characteristics in a data analysis-based intelligent fault diagnosis method for photovoltaic fuses proposed in this invention.

[0063] Figure 3 This is a schematic diagram illustrating the generation of degradation state evolution paths and fault diagnosis based on a semi-sequence Markov degradation model in the intelligent fault diagnosis method for photovoltaic fuses based on data analysis proposed in this invention. Detailed Implementation

[0064] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0065] refer to Figures 1-3 A data analysis-based intelligent fault diagnosis method for photovoltaic fuses includes the following steps:

[0066] The system acquires the operating data of the target photovoltaic fuse in the photovoltaic power generation system within a preset monitoring time window. The operating data includes current data, temperature data, and operating condition data, and records the corresponding timestamps for the current data, temperature data, and operating condition data.

[0067] Time alignment processing is performed on the operating data, and current data, temperature data and operating condition data are associated based on a unified time index to form a photovoltaic fuse operating data sequence arranged in chronological order;

[0068] Based on the photovoltaic fuse operation data sequence, state features characterizing the changes in the operating state of the photovoltaic fuse are extracted, and the state features are arranged in chronological order to generate a photovoltaic fuse state feature sequence.

[0069] The operating status of photovoltaic fuses is analyzed based on the state characteristic sequence of photovoltaic fuses. Multiple degradation state intervals are divided according to the distribution and change relationship of state characteristics under continuous time index. Based on the state constraint relationship between each degradation state interval, which only allows changes to occur along the degradation direction, a semi-order degradation state space with degradation sequence constraint is constructed.

[0070] In the semi-sequence degradation state space, the degradation intensity parameter reflecting the cumulative degree of degradation of the photovoltaic fuse is calculated based on the photovoltaic fuse operation data sequence, and the degradation intensity parameter is associated with the interval boundary conditions of the degradation state interval to determine the state transition constraint rules of the photovoltaic fuse operation state in the semi-sequence degradation state space.

[0071] Based on the state transition constraint rules, the process of maintaining or transitioning the operating state of a photovoltaic fuse in the semi-sequence degradation state space to the subsequent degradation state interval is modeled, and a semi-sequence Markov degradation model is constructed.

[0072] Based on the semi-order Markov degradation model, the operating state of photovoltaic fuses under continuous time index is deduced, and the degradation state evolution path is generated in time order without reverse degradation transition.

[0073] Based on the degradation state evolution path, the consistency and trend of the degradation state of the photovoltaic fuse within a preset time range are determined, and the fault diagnosis results of the photovoltaic fuse are output.

[0074] In this embodiment, the generation of the photovoltaic fuse operation data sequence includes:

[0075] Based on the timestamps corresponding to current data, temperature data, and operating condition data, the operating data undergoes unified time base processing to determine a unified time index covering the preset monitoring time window;

[0076] Based on a unified time index, timestamp alignment is performed on current data, temperature data, and operating condition data to establish a correspondence between them under the unified time index.

[0077] Based on the unified time index, current data, temperature data, and operating condition data with timestamps earlier than the start position of the unified time index or later than the end position of the unified time index are truncated to make the time range of current data, temperature data, and operating condition data consistent with the unified time index.

[0078] Based on the unified time index, the current data, temperature data, and operating condition data are filled in when there are missing index positions under the unified time index, so that the current data, temperature data, and operating condition data are arranged continuously under the unified time index.

[0079] Based on a unified time index, current data, temperature data, and operating condition data corresponding to the same time index are associated to form associated data according to the unified time index;

[0080] The associated data are arranged according to the chronological order of a unified time index, forming a sequence of photovoltaic fuse operation data arranged in chronological order.

[0081] In this embodiment, the generation of the photovoltaic fuse state characteristic sequence includes:

[0082] Using the time sequence corresponding to the photovoltaic fuse operation data sequence as the processing order, the current data, temperature data and operating condition data corresponding to the photovoltaic fuse operation data sequence are read in the order of time to form an operation data set that corresponds one-to-one with the time sequence.

[0083] Based on the set of operating data, in the corresponding time sequence, the instantaneous value of the current, the average value of the current within the preset time window, and the current change amplitude in adjacent time sequences are calculated for the current data. The instantaneous value of the temperature, the average value of the temperature within the preset time window, and the rate of temperature change in adjacent time sequences are calculated for the temperature data. The operating condition data are then converted into operating condition state values ​​corresponding to the time sequence, generating state features that correspond one-to-one with the time sequence.

[0084] Within a preset time window, based on the state features corresponding to adjacent time sequences, the difference between the state features corresponding to the previous time sequence and the state features corresponding to the next time sequence is calculated to form the state feature change results corresponding to the time sequence.

[0085] Based on the state characteristics and the results of state characteristic changes, the state characteristics are combined in chronological order, and the combined state characteristics are arranged in chronological order to generate a photovoltaic fuse state characteristic sequence.

[0086] In this embodiment, constructing a semi-order degenerate state space with degenerate order constraints includes:

[0087] Using the continuous time index in the photovoltaic fuse state feature sequence as the analysis order, the state features in the photovoltaic fuse state feature sequence are statistically processed within the preset analysis time window. The statistical values ​​of the corresponding state features are calculated for each continuous time index, and the distribution pattern of the state features within the continuous time range is characterized to support the subsequent division of the degradation state interval based on the distribution change, and generate the state feature distribution results that correspond one-to-one with each continuous time index.

[0088] Based on the state feature distribution results corresponding to adjacent continuous time indices, the distribution change results between adjacent state feature distribution results are calculated. The calculation of the distribution change results between adjacent state feature distribution results includes quantifying the structural differences in the statistical distribution of state features within adjacent time windows, which is used to characterize the degree of change in degradation evolution over time, and arranging the distribution change results according to the order of the continuous time indices to generate a distribution change sequence.

[0089] Based on the distribution change sequence, the continuous time index position that meets the preset change condition in the distribution change sequence is determined, and the continuous time index position is used as the interval division position to divide the photovoltaic fuse state characteristic sequence and generate multiple degradation state intervals.

[0090] The preset change conditions include: within a preset analysis time window, the distribution change results of the corresponding continuous time index in the distribution change sequence satisfy a monotonically increasing or monotonically non-decreasing change trend at multiple consecutive time positions, and satisfy the condition that the number of consecutive satisfying is not less than a preset number of times; or within a preset analysis time window, the proportion of time positions where the distribution change results exceed the preset quantile statistical benchmark is not less than a preset proportion; or the distribution change sequence shows an inflection point position where the change changes from a stable change to a continuous increasing change within the preset analysis time window. The continuous time index position is the time index position that satisfies at least one of the above change conditions.

[0091] Based on multiple degenerate state intervals, interval statistical processing is performed on the state features within each degenerate state interval to generate interval state features that correspond one-to-one with each degenerate state interval.

[0092] Based on the characteristics of the interval state, the degradation order among multiple degradation state intervals is determined according to the direction of change of each degradation state interval in time sequence. Based on the degradation order, a state constraint relationship is constructed among multiple degradation state intervals. The state constraint relationship restricts any degradation state interval to only remain in the current degradation state interval or change to the subsequent degradation state interval.

[0093] Based on multiple degenerate state intervals and state constraint relationships, a semi-order degenerate state space with degenerate order constraints is constructed.

[0094] In this embodiment, the determination of the state transition constraint rules includes:

[0095] In the semi-sequence degradation state space, based on the time sequence of the photovoltaic fuse operation data sequence, the current data, temperature data and operating condition data corresponding to each time position are read sequentially. Based on the numerical change relationship between adjacent time positions, the change amplitude of the current data and the change rate of the temperature data are calculated. At the same time, the operating condition data values ​​of the corresponding time positions are obtained, and the degradation intensity parameter calculation results corresponding one-to-one with the time sequence of the photovoltaic fuse operation data sequence are generated.

[0096] The calculation of the magnitude of change of current data and the rate of change of temperature data specifically includes: at adjacent time positions, taking the difference between the values ​​of current data before and after the time to characterize the magnitude of change of current over time; and taking the change of temperature data before and after the time and normalizing it in combination with the time interval to characterize the rate of change of temperature over time.

[0097] Based on the calculation results of the degradation intensity parameter, the cumulative calculation of the degradation intensity parameter is performed in chronological order within a preset time window to generate degradation intensity parameters that reflect the cumulative degree of degradation of the photovoltaic fuse.

[0098] Based on multiple degenerate state intervals in the semi-order degenerate state space, the corresponding interval boundary conditions are determined for each degenerate state interval, and a correspondence is established between each degenerate state interval and its corresponding interval boundary conditions.

[0099] The interval boundary conditions include the lower boundary conditions and the upper boundary conditions of the corresponding degenerate state interval. The lower boundary conditions and the upper boundary conditions are determined by the interval start position and the interval end position of the degenerate state interval in the semi-order degenerate state space.

[0100] The degradation intensity parameter is associated with the boundary conditions of the current degradation state interval. Based on the positional relationship between the degradation intensity parameter and the boundary conditions, the state transition constraint rules of the photovoltaic fuse operating state in the semi-sequence degradation state space are determined. When the degradation intensity parameter falls within the range of the boundary conditions of the current degradation state interval, the photovoltaic fuse operating state is determined to remain in the current degradation state interval. When the degradation intensity parameter reaches or exceeds the upper boundary condition of the current degradation state interval and satisfies the stable holding condition under the preset continuous time index, the photovoltaic fuse operating state is determined to transition from the current degradation state interval to the subsequent degradation state interval.

[0101] In this embodiment, constructing the semi-order Markov degeneracy model includes:

[0102] In the semi-sequence degradation state space, each degradation state interval is taken as the state set of the photovoltaic fuse operation state, and according to the degradation order determined in the semi-sequence degradation state space, the corresponding subsequent degradation state interval is determined for each degradation state interval.

[0103] Based on the state set and subsequent degradation state intervals, and in accordance with the state transition constraint rules, the permissible changes in the operating state of the photovoltaic fuse in the semi-sequence degradation state space are restricted, so that the operating state of the photovoltaic fuse is only allowed to remain in the current degradation state interval or to transfer from the current degradation state interval to the subsequent degradation state interval in any degradation state interval, thus forming a set of state changes restricted by the state transition constraint rules.

[0104] Based on the set of state changes, according to the order of changes in the operating state of the photovoltaic fuse in the semi-sequence degradation state space, the holding behavior in each degradation state interval and the transfer behavior to the subsequent degradation state interval are statistically analyzed, and the number of holding occurrences and the number of transfer occurrences corresponding to each degradation state interval are generated.

[0105] Based on the number of times maintenance occurs and the number of times transition occurs, the maintenance probability value corresponding to each degradation state interval is determined according to the ratio of the number of times maintenance occurs to the sum of the number of times maintenance occurs and the number of times transition occurs, and the transition probability value corresponding to each degradation state interval is also determined.

[0106] Based on the set of state changes, according to the order of changes in the operating state of the photovoltaic fuse in the semi-sequence degradation state space, the continuous holding time of the operating state of the photovoltaic fuse in each degradation state interval is statistically analyzed, and the continuous holding time is statistically analyzed according to the preset time division, generating a distribution value of holding duration corresponding to each degradation state interval, which is used to characterize the typical holding time range of the operating state of the photovoltaic fuse in the corresponding degradation state interval.

[0107] Based on the values ​​of the hold probability, the transition probability, and the hold duration distribution, a semi-order Markov degradation model is constructed to model the hold process of the photovoltaic fuse's operating state and the transition process to the subsequent degradation state interval in the semi-order degradation state space.

[0108] In this embodiment, the generation of the degradation state evolution path includes:

[0109] Based on the semi-order Markov degradation model, the starting continuous time index in the continuous time index is determined, and the starting degradation state interval of the photovoltaic fuse operation state corresponding to the starting continuous time index is determined as the starting state of the degradation state evolution path.

[0110] Using the initial continuous time index as the current continuous time index and the initial degradation state interval as the current degradation state interval, the photovoltaic fuse operating state under the current continuous time index is deduced based on the state description corresponding to the current degradation state interval in the semi-order Markov degradation model, and the corresponding next degradation state interval is obtained.

[0111] The state deduction is based on the retention probability value, transition probability value, and retention duration distribution value corresponding to the current degraded state interval. When the retention time length under the current continuous time index does not reach the typical retention time range corresponding to the retention duration distribution value, it is preferentially determined to remain in the current degraded state interval. When the retention time length reaches or exceeds the typical retention time range, it is determined whether to transition to the subsequent degraded state interval based on the transition probability value.

[0112] Associating the next degenerate state interval with its corresponding subsequent continuous time index forms a state pair consisting of a continuous time index and a degenerate state interval. The state pairs are then recorded in the order of the continuous time indexes to generate the degenerate state evolution path.

[0113] Based on the constraint on the direction of change of the degenerate state interval in the semi-order Markov degeneracy model, the state pairs are checked. When there is a reverse degeneracy transition between the next degenerate state interval and the current degenerate state interval that does not conform to the constraint of the degeneracy direction, the next degenerate state interval is corrected to the current degenerate state interval, and the degenerate state evolution path is updated with the corrected state pair.

[0114] The revised next degenerate state interval is used as the new current degenerate state interval, and the continuous time index is used as the new current continuous time index. The process of updating the degenerate state evolution path is repeated until the preset termination continuous time index is reached, thereby generating a degenerate state evolution path arranged in the order of continuous time index and without containing reverse degenerate transitions.

[0115] In this embodiment, the generation of the fault diagnosis result includes:

[0116] Based on the evolution path of the degenerate state, a preset time range corresponding to the evolution path of the degenerate state is determined, and a sequence of degenerate states located within the preset time range and arranged in chronological order is extracted from the evolution path of the degenerate state as the evolution path of the degenerate state for judgment.

[0117] Based on the evolution path of the degradation state used for judgment, the degradation states corresponding to adjacent time positions are compared in chronological order. The number of times the degradation state remains unchanged within a preset time range and the number of times the degradation state changes along the degradation direction are counted. The total number of degradation state changes is calculated based on the number of times the degradation state remains unchanged and the number of times the degradation state changes along the degradation direction.

[0118] Based on the number of hold times, the number of transfer times, and the total number of degradation state changes, the evolution consistency of the degradation state of photovoltaic fuses within a preset time range is determined according to the ratio between the number of hold times and the total number of degradation state changes, as well as the ratio between the number of transfer times and the total number of degradation state changes, and an evolution consistency determination result is generated.

[0119] The generation of the evolution consistency determination result specifically includes: within a preset time range, based on the evolution path of the degenerate state used for determination, counting the number of times the degenerate state remains unchanged, the number of times it changes along the degenerate direction, and the total number of times the degenerate state changes, and generating the evolution consistency determination result based on the ratio of the number of times the degenerate state remains unchanged to the total number of times the degenerate state changes and the ratio of the number of times the degenerate state changes to the total number of times the degenerate state changes.

[0120] Based on the degradation state evolution path used for judgment, the changes in the degradation state along the degradation direction within a preset time range are accumulated in chronological order to obtain the cumulative change result of the degradation state within the preset time range. Based on the ratio between the cumulative change result and the time length of the preset time range, the evolution trend of the photovoltaic fuse degradation state within the preset time range is judged, and the evolution trend judgment result is generated.

[0121] Based on the results of evolution consistency determination and evolution trend determination, the degradation status of photovoltaic fuses within a preset time range is comprehensively determined, and the fault diagnosis results of photovoltaic fuses are generated and output.

[0122] Example 1:

[0123] To verify the feasibility of this invention in practice, it was applied to the DC-side operation monitoring scenario of a large-scale ground-mounted photovoltaic (PV) power plant. This PV power plant is located in an area with relatively stable solar radiation conditions and significant diurnal temperature variations. The plant has a large number of DC combiner circuits and a long operating history, resulting in PV fuses operating under conditions of frequent current fluctuations, significant ambient temperature changes, and continuous switching of operating modes. During actual operation, maintenance personnel found that some fuses often underwent a long period of performance degradation before failure. However, existing monitoring methods mainly rely on instantaneous threshold alarms for current or temperature, making it difficult to identify fuses in the early stages of degradation in a timely manner, leading to false alarms or missed alarms, increasing maintenance risks and costs. This application scenario is precisely the core problem that this invention aims to solve: how to continuously characterize the degradation evolution process of PV fuses based on operational data and provide reliable diagnostic conclusions before failure occurs.

[0124] In this scenario, the operating status of the target photovoltaic fuse in the photovoltaic power generation system is first continuously monitored. Current data, temperature data, and operating condition data reflecting load changes, grid connection status, and environmental conditions are collected during long-term operation. All operating data are timestamped upon acquisition to ensure accurate reflection of the time correspondence between various data types during subsequent analysis. Subsequently, the collected operating data undergoes unified time-base processing. Through time alignment and data association, current data, temperature data, and operating condition data from different sampling frequencies and sources are integrated into a chronologically arranged sequence of photovoltaic fuse operating data, allowing the operating status at each moment to be described by a complete dataset.

[0125] Based on this, features of the fuse operating status are extracted according to the time sequence of the operating data sequence. Through comprehensive analysis of current changes, temperature trends, and operating conditions, state features reflecting changes in fuse operating status are generated. These state features are then arranged in chronological order to form a photovoltaic fuse state feature sequence. This state feature sequence not only retains operating information at a single moment but also includes the changing characteristics between adjacent time periods, providing a continuous basic description for subsequent degradation analysis.

[0126] Subsequently, based on the state characteristic sequence of the photovoltaic fuse, statistical analysis of the state characteristics is performed over a continuous time range, focusing on the distribution and changes of state characteristics over time. By comparing the distribution results of state characteristics in adjacent time periods, the locations where structural changes in the operating state occur are identified, and based on this, the entire operating process is divided into multiple degradation state intervals. Each degradation state interval corresponds to a relatively stable operating state level of the fuse within a certain stage. Furthermore, according to the evolution direction of these degradation state intervals in time sequence, the constraint relationship between each degradation state interval is clarified, allowing only maintenance or changes along the degradation direction, thereby constructing a semi-order degradation state space with degradation sequence constraints. In this way, the originally continuous and complex operating state change process is transformed into a state space representation with a clear structure and degradation logic constraints.

[0127] After constructing the semi-sequence degradation state space, the cumulative degradation of photovoltaic fuses during long-term operation is quantitatively described based on the photovoltaic fuse operating data sequence. By analyzing the current variation amplitude and temperature variation rate at adjacent time points, and combining the operating conditions, the degradation intensity, reflecting the degree of fuse degradation, is calculated and accumulated over time to obtain a degradation intensity parameter that reflects the gradual worsening trend of degradation. Subsequently, the degradation intensity parameter is correlated with the boundary conditions corresponding to each degradation state interval. Based on the variation of degradation intensity within the interval, the state transition constraint rules of the fuse operating state in the semi-sequence degradation state space are determined, thus closely linking the state transition process with the actual degree of degradation.

[0128] Building upon the above, a semi-order Markov degradation model is constructed by utilizing the established semi-order degradative state space and state transition constraints to model the holding behavior of the fuse's operating state and its transition behavior to subsequent degradative state intervals. This model, while describing the state transition probability, incorporates the holding duration characteristics of the operating state within each degradation state interval, enabling it to reflect the differences in the dwell time of the fuse at different degradation stages. Through this model, the operating states of the fuse under continuous-time indexing are deduced, progressively generating degradation state evolution paths that do not include reverse degradation transitions, thus comprehensively depicting the entire process of the fuse's evolution from its initial operating state towards failure.

[0129] During the fault diagnosis phase, based on the generated degradation state evolution path, a time range covering a certain operating cycle is selected as the analysis object. The changes in the degradation state within this time range are comprehensively assessed. On one hand, by statistically analyzing cases where the degradation state remains unchanged and undergoes degradation transitions within this time range, the consistency of the degradation state evolution is analyzed to determine whether the fuse is in a stable operating phase or a continuous degradation phase. On the other hand, by analyzing the cumulative changes in the degradation state along the degradation direction, the degradation evolution trend is judged. Finally, the results of the degradation evolution consistency and degradation evolution trend judgments are combined to output the fault diagnosis conclusion for the photovoltaic fuse.

[0130] As demonstrated by its application in the aforementioned real-world scenarios, this invention can continuously and structurally characterize the degradation process of photovoltaic fuses without relying on a single threshold alarm. Compared to traditional methods that only focus on instantaneous states, this invention is more conducive to identifying potential degradation risks in advance, reducing the impact of operating noise and fluctuations in operating conditions on diagnostic results, thereby improving the reliability and practicality of photovoltaic power generation system operation monitoring and fault diagnosis.

[0131] Table 1. Experimental Comparison Results of Intelligent Fault Diagnosis Methods for Photovoltaic Fuses

[0132] Performance indicators Traditional threshold diagnostic methods Statistical model diagnostic methods The method of this invention (based on semi-order Markov degeneracy modeling) Fault identification accuracy (percentage) 84.6 89.8 93.2 False alarm rate (percentage) 10.7 6.3 2.4 Missed report rate (percentage) 9.8 5.6 4.1 Advance warning time (hours) 6.5 14.2 27.6 Consistency of operational status assessment (percentage) 82.3 88.9 92.1 Diagnostic reliability (percentage) under complex operating conditions 78.5 85.7 93.4

[0133] As shown in the table, in terms of fault identification accuracy, the traditional threshold diagnosis method achieves 84.6%, the statistical model diagnosis method improves to 89.8%, while the method of this invention further improves to 93.2%. Compared with the traditional threshold method, the accuracy is improved by 8.6 percentage points, and compared with the statistical model method, the accuracy is still improved by 3.4 percentage points. This result indicates that the method of this invention has a higher level of overall accuracy in fault identification and can reduce operational decision-making biases caused by misjudgments.

[0134] Regarding false alarm rates, the traditional threshold diagnostic method has a false alarm rate of 10.7%, the statistical model diagnostic method reduces it to 6.3%, while the method of this invention further reduces it to 2.4%. Compared to the traditional threshold method, the false alarm rate is reduced by 8.3 percentage points; compared to the statistical model method, the false alarm rate is reduced by 3.9 percentage points. This indicates that the method of this invention has a stronger ability to suppress non-realistic degradation states under complex operating conditions and can effectively reduce unnecessary alarm triggering.

[0135] Regarding the false negative rate, the traditional threshold diagnostic method has a false negative rate of 9.8%, the statistical model diagnostic method has a false negative rate of 5.6%, and the method of this invention has a false negative rate of 4.1%. Compared with the traditional threshold method, the false negative rate is reduced by 5.7 percentage points; compared with the statistical model method, the false negative rate is reduced by 1.5 percentage points. These data indicate that the method of this invention is more effective in identifying potential faults and reduces the risk caused by failure to identify degradation in a timely manner.

[0136] In terms of early warning time, the traditional threshold diagnostic method provides an early warning time of 6.5 hours, the statistical model diagnostic method improves to 14.2 hours, while the method of this invention achieves 27.6 hours. Compared with the traditional threshold method, the early warning time is extended by 21.1 hours; compared with the statistical model method, the early warning time is extended by 13.4 hours. This invention's method can identify fuse degradation trends at an earlier stage, providing maintenance personnel with more time to respond.

[0137] Regarding the consistency of operational status determination, the traditional threshold diagnostic method achieves 82.3%, the statistical model diagnostic method achieves 88.9%, and the method of this invention reaches 92.1%. Compared with the traditional threshold method, the consistency is improved by 9.8 percentage points; compared with the statistical model method, the consistency is improved by 3.2 percentage points. This reflects that the status determination results given by the method of this invention are more stable across different time periods, reducing the occurrence of repeated status changes.

[0138] In the diagnostic reliability index under complex operating conditions, the traditional threshold diagnostic method achieved 78.5%, the statistical model diagnostic method achieved 85.7%, and the method of this invention achieved 93.4%. Compared with the traditional threshold method, the reliability is improved by 14.9 percentage points; compared with the statistical model method, the reliability is improved by 7.7 percentage points. This data shows that the method of this invention can maintain high diagnostic reliability even under conditions of significant variation in operating conditions.

[0139] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A data analysis-based intelligent fault diagnosis method for photovoltaic fuses, characterized in that, Includes the following steps: Acquire the operating data of the target photovoltaic fuse in the photovoltaic power generation system and record the corresponding timestamps; The operating data is time-aligned and associated based on a unified time index to form a photovoltaic fuse operating data sequence. State features are extracted from photovoltaic fuse operation data sequences, and a photovoltaic fuse state feature sequence is generated. Data analysis of the operating status of photovoltaic fuses is performed based on the state characteristic sequence of photovoltaic fuses, and multiple degradation state intervals are divided. A semi-order degradation state space with degradation sequence constraints is constructed according to the state constraint relationship. In the semi-order degraded state space, the degradation intensity parameter is calculated and associated with the boundary conditions of the degradation state interval to determine the state transition constraint rules of the photovoltaic fuse operation state in the semi-order degraded state space. Based on the state transition constraint rules, the transition process of the photovoltaic fuse's operating state in the semi-order degenerate state space is modeled, and a semi-order Markov degenerate model is constructed. Based on the semi-order Markov degradation model, the operating state of photovoltaic fuses is deduced, and the degradation state evolution path is generated. Based on the degradation state evolution path, the consistency and trend of the degradation state evolution of photovoltaic fuses are determined, and the fault diagnosis results of photovoltaic fuses are output.

2. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The generation of the photovoltaic fuse operation data sequence includes: Based on the timestamps corresponding to current data, temperature data, and operating condition data, the operating data undergoes unified time base processing to determine a unified time index covering the preset monitoring time window; Based on a unified time index, timestamp alignment is performed on current data, temperature data, and operating condition data. Truncation is performed on current data, temperature data, and operating condition data whose timestamps are earlier than the start position of the unified time index or later than the end position of the unified time index. For current data, temperature data, and operating condition data that have missing index positions under a unified time index, perform completion processing; The current data, temperature data, and operating condition data corresponding to the same time index are associated to form associated data corresponding to a unified time index. The associated data are arranged according to the chronological order of a unified time index, forming a sequence of photovoltaic fuse operation data arranged in chronological order.

3. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The generation of the photovoltaic fuse state characteristic sequence includes: Read the corresponding current data, temperature data and operating condition data in the photovoltaic fuse operation data sequence in chronological order to form an operation data set that corresponds one-to-one with the chronological order. Based on the running data set, state features that correspond one-to-one with the time sequence are generated. Within a preset time window, based on the state features corresponding to adjacent time sequences, the difference between the state features corresponding to the previous time sequence and the state features corresponding to the next time sequence is calculated to form the state feature change results corresponding to the time sequence. Based on the state characteristics and the results of state characteristic changes, the state characteristics are combined in chronological order, and the combined state characteristics are arranged in chronological order to generate a photovoltaic fuse state characteristic sequence.

4. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The construction of the semi-order degenerate state space with degenerate order constraints includes: Using the continuous time index in the photovoltaic fuse state feature sequence as the analysis order, statistical processing is performed on the state features in the photovoltaic fuse state feature sequence to generate state feature distribution results; Based on the state feature distribution results, the distribution change results between adjacent state feature distribution results are calculated, and the distribution change results are arranged in the order of continuous time index to generate a distribution change sequence. The continuous time index positions that meet the preset change conditions in the distribution change sequence are determined, and the continuous time index positions are used as interval division positions to divide the photovoltaic fuse state characteristic sequence and generate multiple degradation state intervals. Perform interval statistical processing on the state features within each degenerate state interval to generate interval state features that correspond one-to-one with each degenerate state interval; According to the direction of change of each degenerate state interval in time sequence, the degeneracy order among multiple degenerate state intervals is determined, and state constraint relationships are constructed among multiple degenerate state intervals based on the degeneracy order, thus constructing a semi-order degenerate state space with degeneracy order constraints.

5. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The determination of the state transition constraint rules includes: In the semi-sequence degradation state space, the current data, temperature data and operating condition data corresponding to each time position are read sequentially based on the time sequence. The change amplitude of the current data and the change rate of the temperature data are calculated. At the same time, the values ​​of the operating condition data are obtained, and the degradation intensity parameter calculation results corresponding to the time sequence are generated. Based on the calculation results of the degradation intensity parameters, cumulative calculations are performed on the degradation intensity parameters in chronological order within a preset time window to generate degradation intensity parameters; Based on multiple degenerate state intervals in the semi-order degenerate state space, the corresponding interval boundary conditions are determined for each degenerate state interval, and a correspondence is established between each degenerate state interval and its corresponding interval boundary conditions. The degradation intensity parameter is associated with the interval boundary conditions corresponding to the current degradation state interval. Based on the positional relationship between the degradation intensity parameter and the interval boundary conditions, the state transition constraint rules of the photovoltaic fuse operating state in the semi-sequence degradation state space are determined.

6. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The construction of the semi-order Markov degenerate model includes: In the semi-order degradation state space, each degradation state interval is taken as the state set of the photovoltaic fuse operation state, and the corresponding subsequent degradation state interval is determined for each degradation state interval according to the degradation order. Based on the state set and the subsequent degradation state interval, and in accordance with the state transition constraint rules, the permissible changes in the operating state of the photovoltaic fuse in the semi-sequence degradation state space are limited, forming a set of state changes limited by the state transition constraint rules. Based on the set of state changes, the maintenance behavior within each degenerate state interval and the transition behavior to the subsequent degenerate state interval are statistically analyzed, generating the number of maintenance occurrences and the number of transition occurrences corresponding one-to-one with each degenerate state interval; Based on the number of times maintenance occurs and the number of times transition occurs, the maintenance probability value corresponding to each degradation state interval is determined, and the transition probability value corresponding to each degradation state interval is also determined. Based on the set of state changes, according to the order of changes in the operating state of the photovoltaic fuse in the semi-order degradation state space, the continuous holding time of the operating state of the photovoltaic fuse in each degradation state interval is statistically analyzed and generated to generate a distribution of holding duration values ​​that correspond one-to-one with each degradation state interval. Based on the values ​​of the hold probability, the transition probability, and the hold duration distribution, a semi-order Markov degradation model is constructed to model the hold process of the photovoltaic fuse's operating state and the transition process to the subsequent degradation state interval in the semi-order degradation state space.

7. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The generation of the degradation state evolution path includes: Based on the semi-order Markov degradation model, the starting continuous time index in the continuous time index is determined, and the starting degradation state interval of the photovoltaic fuse operation state corresponding to the starting continuous time index is determined as the starting state of the degradation state evolution path. Using the initial continuous time index as the current continuous time index and the initial degradation state interval as the current degradation state interval, the photovoltaic fuse operating state under the current continuous time index is deduced based on the state description corresponding to the current degradation state interval in the semi-order Markov degradation model, and the corresponding next degradation state interval is obtained. Associating the next degenerate state interval with its corresponding subsequent continuous time index forms a state pair consisting of a continuous time index and a degenerate state interval. The state pairs are then recorded in the order of the continuous time indexes to generate the degenerate state evolution path. The state pairs are verified based on the semi-order Markov degeneration model. When there is a reverse degeneration transition between the next degeneration state interval and the current degeneration state interval that does not conform to the degeneration direction constraint, the next degeneration state interval is corrected to the current degeneration state interval, and the degeneration state evolution path is updated with the corrected state pairs. The revised next degenerate state interval is used as the new current degenerate state interval, and the continuous time index is used as the new current continuous time index. The process of updating the degenerate state evolution path is repeated until the preset termination continuous time index is reached, thus generating the degenerate state evolution path.

8. The intelligent fault diagnosis method for photovoltaic fuses based on data analysis according to claim 1, characterized in that, The generation of the fault diagnosis results includes: Based on the evolution path of the degenerate state, a preset time range corresponding to the evolution path of the degenerate state is determined, and a sequence of degenerate states located within the preset time range and arranged in chronological order is extracted from the evolution path of the degenerate state as the evolution path of the degenerate state for judgment. Based on the evolution path of the degradation state used for judgment, the degradation states corresponding to adjacent time positions are compared in chronological order. The number of times the degradation state remains unchanged within a preset time range and the number of times the degradation state changes along the degradation direction are counted. The total number of degradation state changes is calculated based on the number of times the degradation state remains unchanged and the number of times the degradation state changes along the degradation direction. Based on the number of hold times, the number of transfer times, and the total number of degradation state changes, the evolution consistency of the degradation state of photovoltaic fuses within a preset time range is determined according to the ratio between the number of hold times and the total number of degradation state changes, as well as the ratio between the number of transfer times and the total number of degradation state changes, and an evolution consistency determination result is generated. Based on the degradation state evolution path used for judgment, the changes in the degradation state along the degradation direction within a preset time range are accumulated in chronological order to obtain the cumulative change result of the degradation state within the preset time range. Based on the ratio between the cumulative change result and the time length of the preset time range, the evolution trend of the photovoltaic fuse degradation state within the preset time range is judged, and the evolution trend judgment result is generated. Based on the results of evolution consistency determination and evolution trend determination, the degradation status of photovoltaic fuses within a preset time range is comprehensively determined, and the fault diagnosis results of photovoltaic fuses are generated and output.

Images