Thermal power plant auxiliary control system dynamic diagnosis method and system based on time sequence state machine
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUANENG LINYI POWER GENERATION CO LTD
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-24
Smart Images
Figure CN122450102A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of diagnostic technology for auxiliary control systems in thermal power plants, and in particular to a dynamic diagnostic method and system for auxiliary control systems in thermal power plants based on a time-sequenced state machine. Background Technology
[0002] Thermal power plants, as the core of power energy security, rely heavily on their auxiliary control systems, such as coal conveying systems, ash removal systems, and desulfurization systems, whose efficient operation directly impacts the overall safety and economy of the unit. Due to the extremely complex processes, dense logical sequence control, and massive amounts of real-time monitoring points involved in these auxiliary control systems, traditional operation and maintenance models relying on manual inspections or simple post-event alarms are no longer sufficient to meet the demands of modern power plants for refined equipment management and early warning of faults. Therefore, developing a dynamic diagnostic solution capable of deeply mining production process data and sensing abnormal states in real time is of profound practical significance for improving the reliability of auxiliary system operation and preventing unplanned shutdowns.
[0003] Currently, monitoring and diagnostic solutions for auxiliary control systems in thermal power plants primarily rely on pre-defined logic judgments within the Distributed Control System (DCS). These existing diagnostic solutions typically rely on ideal parameters during the initial commissioning phase, verifying the correctness of control sequence actions by setting fixed timers or rigid time windows in the DCS logic. However, this fixed-threshold state machine monitoring mechanism based on rigid thresholds reveals serious adaptability issues when facing long-term industrial environments. This is because, after 3-5 years of operation, power generation equipment inevitably experiences mechanical characteristic degradation, wear of critical components, or operational drift. This leads to significant nonlinearity and uncertainty in the response time of actuators and the sequence of actions. Existing diagnostic solutions lack the flexibility to handle timing deviations, easily causing the system to enter a misjudgment state due to minor action lags, resulting in a persistently high false alarm rate. Furthermore, while logic timeout parameters can be manually adjusted through re-tuning, frequent modifications to the underlying DCS logic in actual production environments are not only cumbersome but also carry extremely high safety risks. The contradiction between this rigid logic and the degradation of equipment mechanical characteristics has become a key bottleneck restricting the intelligent development of auxiliary control systems in thermal power plants. Summary of the Invention
[0004] The present invention aims to solve at least one of the problems existing in the prior art, and provides a dynamic diagnosis method and system for auxiliary control systems of thermal power plants based on a time-sequenced state machine.
[0005] One aspect of the present invention provides a dynamic diagnostic method for auxiliary control systems of thermal power plants based on a time-series state machine, comprising: Step S1: Based on a preset sliding time window, perform buffering, slicing, and vectorization processing on the event stream of the thermal power plant auxiliary control system acquired in real time to obtain the set of observed events within the current window; Step S2: Based on the current state pointer, perform successor node search and concurrent logic block parsing on the pre-built state machine topology graph to obtain the target event set in the current state; Step S3: Perform set similarity matching and out-of-order measurement on the observed event set and the target event set to obtain the matching flag and out-of-order entropy value; Step S4: Based on the matching flag, out-of-order entropy value and preset tolerance threshold, perform elastic tolerance judgment and state transition decision on the current state transition conditions to obtain the diagnosis category and transition state pointer. Step S5: Using the migration status pointer as the associated index, perform trend accumulation and health profile updates for the diagnosis category and out-of-order entropy value to generate a health report.
[0006] Another aspect of the present invention provides a dynamic diagnostic system for auxiliary control systems of thermal power plants based on a time-series state machine, comprising: The windowed event observation module is used to perform buffering, slicing, and vectorization processing on the event stream of the auxiliary control system of the thermal power plant acquired in real time based on a preset sliding time window to obtain the set of observed events in the current window. The state diagram subsequent event parsing module is used to perform successor node search and concurrent logic block parsing on the pre-built state machine topology diagram based on the current state pointer to obtain the target event set in the current state; The event set matching and disordered quantization module is used to perform set similarity matching and disordered quantification on the observed event set and the target event set to obtain the matching flag and disordered entropy value. The elastic state transition decision module is used to make elastic tolerance judgments and state transition decisions on the current state transition conditions based on the matching flag, out-of-order entropy value and preset tolerance threshold, so as to obtain the diagnosis category and transition state pointer. The Health Trend Accumulation and Report Generation module is used to accumulate trends and update health profiles for diagnostic categories and out-of-order entropy values using migration status pointers as associated indexes, in order to generate health reports.
[0007] Compared with existing technologies, this invention provides a dynamic diagnostic method and system for auxiliary control systems of thermal power plants based on a time-series state machine. It slices and vectorizes the real-time event stream through a sliding time window to construct a set of currently observed events. It then analyzes the target event set under the current state using a pre-constructed state machine topology diagram. Furthermore, it quantifies the deviation of the event sequence by performing set matching and out-of-order measurement on observed and target events, and introduces a flexible tolerance mechanism. Considering network jitter and logical deviations, it achieves flexible judgment of state transitions, effectively distinguishing between normal fluctuations, minor network jitter, and severe logical misordering. Finally, based on the transition state pointer, it accumulates trends and updates the health profile of the diagnostic results, generating a health report that guides operation and maintenance. This effectively improves the adaptability to equipment state changes in complex industrial environments and significantly enhances the predictability of auxiliary control system operation and the scientific nature of operation and maintenance decisions. Attached Figure Description
[0008] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.
[0009] Figure 1 This is a flowchart of a dynamic diagnostic method for a thermal power plant auxiliary control system based on a time-sequenced state machine, according to an embodiment of the present invention. Figure 2 This is a data flow diagram illustrating the dynamic diagnostic method for auxiliary control systems of thermal power plants based on a time-series state machine, according to an embodiment of the present invention. Figure 3 This is a flowchart of step S2 in the dynamic diagnosis method for auxiliary control system of thermal power plant based on timing state machine according to an embodiment of the present invention; Figure 4 This is a block diagram of a dynamic diagnostic system for a thermal power plant auxiliary control system based on a time-sequenced state machine, according to an embodiment of the present invention. Detailed Implementation
[0010] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the embodiments of the present invention to facilitate a better understanding of the invention. However, the technical solutions claimed in the present invention can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.
[0011] As indicated in the specification and claims of this invention, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.
[0012] While this invention makes various references to certain modules in systems according to embodiments of the invention, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.
[0013] This invention uses flowcharts to illustrate the operations performed by the system according to embodiments of the invention. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously, as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.
[0014] One embodiment of the present invention provides a dynamic diagnostic method for auxiliary control systems of thermal power plants based on a time-sequential state machine. Figure 1 This is a flowchart of a dynamic diagnostic method for a thermal power plant auxiliary control system based on a time-series state machine, according to an embodiment of the present invention. Figure 2 This is a data flow diagram illustrating the dynamic diagnostic method for a thermal power plant auxiliary control system based on a time-series state machine, according to an embodiment of the present invention. (In conjunction with...) Figure 1 and Figure 2 According to an embodiment of the present invention, a dynamic diagnosis method for a thermal power plant auxiliary control system based on a time-series state machine includes: Step S1, based on a preset sliding time window, performing cache slicing and vectorization processing on the event stream of the thermal power plant auxiliary control system acquired in real time to obtain the observed event set within the current window; Step S2, based on the current state pointer, performing successor node search and concurrent logic block parsing on the pre-constructed state machine topology graph to obtain the target event set under the current state; Step S3, performing set similarity matching and out-of-order measurement on the observed event set and the target event set to obtain a matching flag and an out-of-order entropy value; Step S4, based on the matching flag, the out-of-order entropy value, and a preset tolerance threshold, performing elastic tolerance judgment and state transition decision on the current state transition conditions to obtain a diagnosis category and a transition state pointer; Step S5, using the transition state pointer as an associated index, performing trend accumulation and health profile update on the diagnosis category and the out-of-order entropy value to generate a health report.
[0015] Specifically, in step S1, based on a preset sliding time window, the event stream of the auxiliary control system of the thermal power plant, which is collected in real time, is cached, sliced, and vectorized to obtain the set of observed events within the current window. Here, the event stream includes auxiliary machine current values, valve feedback positions, pipeline pressure values, and switching signals. It should be understood that in the production environment of a thermal power plant, the event stream generated by the auxiliary machine control system has extremely strong temporal continuity and massive fragmentation characteristics. The raw data contains a large amount of invalid information due to network fluctuations or instantaneous sensor interference. If it is not preprocessed, it will directly face a messy and disordered full data stream, resulting in an overload of logic computation and easy noise interference. Therefore, in the technical solution of this invention, the dynamic and continuous raw event stream is transformed into a discrete and computable set of observations through a sliding time window, and invalid data is removed using a quality bit filtering mechanism, providing high-purity and structured data support for the subsequent logic matching of the state machine, ensuring the real-time performance and accuracy of the diagnostic results.
[0016] Step S1, in its specific implementation, includes: First, based on the quality bit verification rules, validity filtering and cache injection are performed on each event packet in the event stream. Valid events that pass the verification are pushed into a first-in-first-out (FIFO) circular buffer in the order of arrival to obtain a real-time event cache queue. It should be understood that each data packet in the original event stream, when entering the diagnostic processing flow, must be judged for validity according to predefined quality bit verification rules (e.g., checksum, data integrity flags, reasonableness of numerical range, etc.). Only valid events that pass all verification rules are allowed to enter the diagnostic processing flow. These valid events are pushed into a FIFO circular buffer in the order of their arrival. The circular buffer is a fixed-size circular queue data structure; when the buffer is full, newly arriving events will overwrite the oldest events. In this way, it is ensured that the cache queue always maintains events within the most recent period, avoiding unlimited memory growth and naturally achieving data timeliness management. Finally, a cleaned real-time event cache queue arranged in chronological order is obtained.
[0017] Next, based on the current system clock and the sliding time window, the real-time event buffer queue is subjected to backtracking interval truncation. All events whose source timestamps fall within a closed interval defined by a window length traced back from the current time are selected to obtain a candidate event sequence. During this process, a preset sliding time window length is maintained. At any diagnostic time, a closed interval is defined on the timeline based on the current system clock. Subsequently, the processor iterates through each event in the aforementioned real-time event buffer queue, but the selection criterion is not the time the event entered the buffer queue, but rather the event's source timestamp (i.e., the actual time the event was generated at the sensor or device). Only events whose source timestamps fall within the defined closed interval are selected, ultimately resulting in a candidate event sequence aligned along the time dimension.
[0018] Then, the event sequence is processed by extracting and encapsulating the identifier and timestamp tuples of each event to obtain the observed event set. In this process, firstly, the events selected from the candidate event sequence typically contain rich field information, such as device ID, measured value, and status code. Secondly, to perform efficient event matching and logical judgment, key features are extracted from these events. Specifically, for each event in the candidate event sequence, its event identifier (used to uniquely identify the event type, such as "#Pump A start command", "#Valve B opening feedback") and source timestamp are extracted, forming a tuple. Subsequently, all event tuples are encapsulated into a set data structure, which is the final observed event set.
[0019] Specifically, in step S2, based on the current state pointer, successor node search and concurrent logic block parsing are performed on the pre-constructed state machine topology diagram to obtain the target event set under the current state. It should be understood that in the operation analysis of auxiliary control systems in thermal power plants, simply obtaining real-time observation data is insufficient; the pre-constructed state machine topology diagram is essentially a mapping of the physical equipment control logic in digital space. Since auxiliary equipment control logic often involves complex processes such as branching, parallelism, and merging, the technical solution of this invention uses the state pointer to accurately locate the logical coordinates of the current auxiliary control system operation and automatically derives the set of events that should theoretically occur under the current environment based on the topology diagram. This step provides a diagnostic reference, enabling the diagnostic algorithm to shift from passive correlation analysis to active logic verification, thereby solving the lack of flexibility in traditional diagnostic schemes caused by the inability to understand the process stages.
[0020] Figure 3 This is a flowchart of step S2 in the dynamic diagnosis method for auxiliary control systems of thermal power plants based on a time-series state machine according to an embodiment of the present invention. Figure 3As shown, step S2 includes: S21, based on the state pointer, performing outgoing edge traversal and condition pre-screening in the state machine topology graph to identify successor candidate nodes on all active paths of the current node to obtain the next-hop node list; S22, based on the attribute type label of each node in the state machine topology graph, performing atomic decomposition of complex logic blocks on each node in the next-hop node list to obtain the atomic event list; S23, performing deduplication and standardization processing on the atomic event list, and binding the corresponding expected logic value to each event identifier according to the expected state defined in the state machine topology graph to obtain the target event set.
[0021] Specifically, in step S21, based on the state pointer, outgoing edge traversal and condition pre-screening are performed in the state machine topology graph to identify successor candidate nodes on all active paths of the current node, thus obtaining a list of next-hop nodes. The current state pointer points to a specific node in the pre-constructed state machine topology graph, which represents the theoretical state of the auxiliary control system. Starting from this node, the processor traverses all outgoing edges (i.e., all possible state transition paths) originating from it. Each outgoing edge is associated with a condition that triggers the transition. At this stage, the processor pre-screens these conditions based on a preset context or simplified logic (e.g., excluding globally disabled paths), identifying all transition paths that are active in the current context. The destination nodes pointed to by these paths are collected to form a list of next-hop nodes, which contains all the next legal theoretical states that the auxiliary control system may transition to from the current state.
[0022] Specifically, in step S22, based on the attribute type labels of each node in the state machine topology diagram, the complex logic block of each node in the next-hop node list is atomically decomposed to obtain a list of atomic events. Since nodes in the state machine topology diagram may have different attribute type labels, for example, one node may represent a simple "waiting for event A," while another node may represent a complex concurrent logic block, such as "waiting for events B and C to occur simultaneously" (logical AND) or "waiting for either event D or event E to occur" (logical OR). During this process, the attribute label of each node in the next-hop node list is checked. For nodes labeled as complex logic blocks (such as AND gates, OR gates, or sequential sequences), they are decomposed into several indivisible basic units, i.e., atomic events, according to their internally defined logical relationships. For example, an "AND (Event_B, Event_C)" logic block will be decomposed into atomic events "Event_B" and "Event_C"; a "Sequence (Step1:Event_D, Step2: Event_E)" logic block will be decomposed into atomic events "Event_D" and "Event_E." After performing this operation on all complex nodes, you get a list of atomic events containing all the basic events.
[0023] Specifically, in step S23, the atomic event list is deduplicated and standardized, and each event identifier is bound to a corresponding expected logical value according to the expected state defined in the state machine topology diagram to obtain the target event set. Since different successor nodes may share some of the same atomic events (e.g., multiple migration paths rely on the same valve feedback signal), duplicate event identifiers may exist in the atomic event list. Therefore, a deduplication operation is further performed on the atomic event list to ensure that each unique event identifier appears only once in the final set. Simultaneously, standardization is performed, such as unifying the naming format of event identifiers. More importantly, the state machine topology diagram not only defines which events are expected to occur but also defines the logical values that these events should possess when triggering a migration (e.g., a switch signal should be "1" (high level) to indicate on, and "0" to indicate off; analog values should be within a specific range). According to the definitions in the state machine topology diagram, a corresponding expected logical value is bound to each deduplicated event identifier. Finally, this series of tuples (event identifier, expected logical value) are encapsulated into a set, namely the target event set, which fully and unambiguously defines the theoretical event snapshots that must be satisfied for a legal transition from the current state.
[0024] Specifically, in step S3, set similarity matching and out-of-order measurement are performed on the observed event set and the target event set to obtain a matching flag and an out-of-order entropy value. It should be understood that set matching alone (whether an event occurred) is insufficient to comprehensively assess the health of the auxiliary control system, because the arrival order of events in industrial networks may be disordered due to network latency, jitter, and other factors. This disorder itself is a symptom of potential faults (such as communication anomalies or controller logic errors). Therefore, in the technical solution of this invention, set similarity matching is used to determine whether the expected event has occurred completely, obtaining a Boolean matching flag; simultaneously, out-of-order measurement is used to quantify the degree of temporal disorder in the event occurrence, obtaining a continuous numerical out-of-order entropy value. These two results provide a qualitative (whether missing) and quantitative (degree of temporal disorder) basis for subsequent steps in making flexible state transition decisions.
[0025] Step S3, in its specific implementation, includes: First, by traversing each element in the target event set and searching for the corresponding event with a matching identifier in the observed event set, performing a set intersection operation and cardinality comparison, a subset of matching events is obtained, consisting of successfully matched events. If the number of elements in the matching event subset is equal to the number of elements in the target event set, the matching flag is set to true; otherwise, the matching flag is set to false. During this process, each tuple (Event_ID, Expected_Value) in the target event set is traversed. For each target event, it searches for an observed event with the same Event_ID in the observed event set (where the elements are (Event_ID, Observed_Timestamp)). This implicitly involves a logical value matching check: the target event is considered successfully matched only if the observed event identifier matches the target identifier, and its carried logical value (this logical value information is already contained in the data packet in the observed event; although the observed event set stores timestamps, the logical value in the original data packet can be associated with the identifier during lookup) matches the Expected_Value of the target event (for example, for a switch signal expected to be "1", the observed value must also be "1"). All successfully matched events (meaning their identifiers match) constitute a subset of matched events. Then, the number of elements in this subset of matched events (i.e., the number of successfully matched events) is compared with the total number of elements in the target event set. If they are equal, it means all expected events have been observed according to the expected logical value, so the matching flag is set to "True"; otherwise, it means some expected events are missing or the logical value is inconsistent, so the matching flag is set to "False". This step completes the completeness check based on set inclusion relationships and logical value consistency.
[0026] Next, based on the predefined logical index order of each element in the target event set, a typical order mapping and time series reconstruction are performed on the matching event subset to obtain a typical order time series. Since the events in the matching event subset only contain identifiers, their actual timestamps need to be obtained to measure the timing. In this process, a logical index (such as 1, 2, 3...) is assigned to each event according to the predefined logical order in the target event set (e.g., in state machine design, event A should occur before event B). Then, based on the identifiers of each event in the matching event subset, the corresponding observation timestamps are retrieved from the observation event set. These timestamps are arranged into a new sequence according to the logical index order (rather than the actual order in which the timestamps occur), namely the typical order time series, which reflects the proper arrangement of timestamps if the events occurred exactly in the theoretical order.
[0027] Furthermore, inverse pair counting and normalized entropy value calculation are performed on the typical order time series to obtain the disorder entropy value, which specifically includes: performing inverse pair counting and normalized entropy value calculation on the typical order time series using the following formula: ; where, is the disorder entropy value, is the total number of events, is the typical order time series, is the th timestamp in the typical order time series, is the jth timestamp in the typical order time series; is the indicator function, which takes the value of 1 when the condition inside the parentheses holds, and 0 otherwise. Specifically, first, calculate the total number of inverse pairs in the typical order time series. The definition of an inverse pair is: in the typical order time series, for any two events, if the actual timestamp of the event that should be earlier in the theoretical order (represented by the smaller index i) is greater than the actual timestamp of the event that should be later in the theoretical order (represented by the larger index j), then these two events form an inverse pair. To count all inverse pairs, traverse all possible, ordered (i < j) event pairs in the typical order time series. For each pair of events, use an indicator function to judge: if the condition "the timestamp of the previous event is greater than the timestamp of the later event" holds, the count value is 1, otherwise it is 0. Summing up the count values of all event pairs gives the total number of inverse pairs. Then, perform normalized entropy value calculation to eliminate the influence of the total number of events N on the result and make the chaos index comparable. In the specific calculation, multiply the total number of inverse pairs by 2 and then divide by the total number of all possible event pairs, where the total number of all possible event pairs is equal to N multiplied by (N - 1), and N is the total number of events. The finally calculated ratio is the disorder entropy value, and the range of this value is normalized between 0 and 1. When the disorder entropy value is 0, it means the sequence is completely in the forward order and there are no inverse pairs; when the disorder entropy value is 1, it means the sequence is completely in the reverse order, that is, the order of every possible pair of events is reversed; when the disorder entropy value is between 0 and 1, it indicates different degrees of time series chaos. The whole process converts the absolute number of inverse pairs into a standard and relative chaos scale.
[0028] Specifically, in step S4, based on the matching flag, out-of-order entropy value, and preset tolerance threshold, a flexible tolerance judgment and state transition decision are made for the current state transition conditions to obtain the diagnostic category and transition state pointer. It should be understood that in the complex industrial electromagnetic environment of thermal power plants and under the background of long-term equipment operation, the feedback of auxiliary actuators often has uncertain time delays or timing reversals. If rigid logic judgments are used in traditional distributed control systems, the system may frequently trigger false alarms or even cause cascading trips due to normal hysteresis caused by minor network signal jitter or mechanical wear. Therefore, in the technical solution of this invention, a flexible tolerance judgment mechanism is introduced. By integrating logical integrity (matching flag) and timing offset (out-of-order entropy value) in multiple dimensions, it can intelligently identify deviations belonging to reliable process actions, deviations belonging to tolerable random noise, and logical faults that must be intercepted. This flexible processing method not only realizes the operating habits of aging equipment through programming, but also fundamentally balances logical rigor and physical ambiguity while ensuring process safety, significantly reducing the false alarm rate and improving the production adaptive capability of diagnosis.
[0029] Step S4, in its specific implementation, includes: First, based on the entry time of the current state and a preset timeout threshold, a joint verification of the integrity and timeout status of the matching flag is performed to obtain a pre-verified status code. During this process, the start time of entering the current state (pointed to by the status pointer) is recorded, and a preset timeout threshold is set. If the matching flag is false and there is no timeout (i.e., the preset timeout threshold has not been exceeded), it is determined to be in a waiting state. If the matching flag is false and there is a timeout (i.e., the preset timeout threshold has been exceeded), it is determined to be in a missing component state. If the matching flag is true, it is determined to be in a ready state. Here, the timeout is determined by whether the difference between the current time and the state entry time is greater than the timeout threshold. If the difference is greater than the timeout threshold, it is "timeout"; otherwise, it is "no timeout". Through this step, a pre-verified status code is obtained, which initially categorizes the situation into three basic states: "waiting," "missing component," or "ready."
[0030] Next, based on the pre-verification status code and the tolerance threshold, the out-of-order entropy value is segmented logically to obtain the diagnostic category. In this process, when the pre-verification status code is "Ready" and the out-of-order entropy value does not exceed the noise floor, the diagnostic category is determined to be normal. When the pre-verification status code is "Ready" and the out-of-order entropy value is greater than the noise floor but does not exceed the tolerance threshold, the diagnostic category is determined to be network micro-jitter. When the pre-verification status code is "Ready" and the out-of-order entropy value exceeds the tolerance threshold, the diagnostic category is determined to be severe logical misordering. Here, the noise floor represents the minimum out-of-order measure that may occur under completely normal and fault-free conditions due to inherent factors such as slight clock asynchrony and sampling errors. The tolerance threshold defines the maximum degree of timing disorder that the system can accept while ensuring functionality and safety. Furthermore, for cases where the pre-verification status code is not "Ready," the diagnostic category is directly mapped from the pre-verification status code: "Waiting State" maps to the diagnostic category "Waiting," and "Component Missing" maps to the diagnostic category "Component Missing." At this point, the final diagnostic category is obtained, which belongs to the set {normal, network minor jitter, severe logical disorder, component missing, waiting}.
[0031] Furthermore, based on the state transition table, a transition mapping is performed between the diagnostic category and the current state pointer to generate a transition state pointer. During this process, a state transition table is maintained, defining the target state to transition from any current state under various diagnostic categories. Specifically, when the diagnostic category is normal or network jitter, a forward transition is triggered to obtain a normal successor state. When the diagnostic category is component missing or severe logical misordering, exception handling is triggered to obtain a safe rollback state. When the diagnostic category is waiting, the current state pointer remains unchanged. During execution, the state transition table is queried using the current state pointer and diagnostic category as a join key. This table can directly store pointers to the target state or store a function call entry point for "obtaining a normal successor state" or "obtaining a safe rollback state." "Obtaining a normal successor state" typically points to the next standard state preset for that state in the state machine topology diagram, used for process advancement. "Obtaining a safe rollback state" may point to a preset safe state, the previous state, or a dedicated fault handling state. The query result is then updated with the new transition state pointer.
[0032] Specifically, in step S5, using the migration status pointer as the associated index, trend accumulation and health profile updates are performed on the diagnostic category and out-of-order entropy value to generate a health report. It should be understood that in the industrial operation scenario of auxiliary control systems in thermal power plants, a single logical diagnosis can only reflect the instantaneous equipment status and cannot depict the gradual performance degradation or potential fault hazards that occur over time. Since the wear and tear of mechanical components or the communication load of network nodes in auxiliary equipment changes dynamically during long-term operation, the technical solution of this invention, by performing trend accumulation on discrete diagnostic events over time, transforms instantaneous logical judgments into a profile analysis of system operation quality to capture the regular evolution of out-of-order entropy values. Thus, before a fault actually occurs, system degradation can be perceived through the health report, achieving a historic leap from traditional passive alarms to proactive operation and maintenance, providing predictive data support for power plant operation and maintenance personnel.
[0033] Step S5, in its specific implementation, includes: First, based on the enumerated values of the diagnostic categories, implementing a tiered alarm strategy and persisting logs for the current diagnostic event to obtain structured log entries. During this process, based on the diagnostic category of the current diagnostic cycle (e.g., normal, minor network jitter, severe logical misordering, etc.), a pre-configured tiered alarm strategy is triggered. For example, "severe logical misordering" and "component missing" may trigger immediate high-level alarms, while "minor network jitter" may only log or trigger a low-level alert. Regardless of whether an external alarm is triggered, all key information from this diagnosis (including timestamps, migration status pointers, diagnostic categories, out-of-order entropy values, matching event identifiers, etc.) will be formatted into a standardized, machine-readable structured log entry and written to a database or log file for persistent storage. This ensures that every diagnostic event can be traced and analyzed.
[0034] Next, based on the exponential moving average algorithm and the migration state pointer as the primary key index, the out-of-order entropy value is weighted and fused with the historical out-of-order mean value corresponding to that state retrieved from the database to obtain network profile data. This step aims to calculate and dynamically update the long-term trend of the out-of-order entropy value for each system state (uniquely identified by the migration state pointer). In this process, using the migration state pointer as the primary key index, the historical out-of-order mean value calculated and stored for that state is retrieved from the database's historical records. Then, the exponential moving average algorithm is used to weight and fuse the out-of-order entropy value calculated in this diagnosis with the historical mean value to calculate a new out-of-order mean value, i.e., the updated out-of-order mean value. That is, by giving more weight to recent data, the out-of-order mean value can more sensitively reflect the latest changing trends. Finally, the calculated new out-of-order mean value is the updated network profile data reflecting the latest time-series health trend of that state, and it is immediately written back to the database using the migration state pointer as the primary key, overwriting the old historical mean value, completing the iterative update of the profile.
[0035] Furthermore, a health score calculation and maintenance suggestion generation are performed on the network profile data and structured log entries. The deviation between the updated out-of-order mean and the baseline value is mapped to a health score ranging from zero to one hundred. When the health score falls below a preset threshold, communication network inspection suggestions are automatically generated to produce a health report. In this process, firstly, a baseline value is set for each state, namely the baseline value of the out-of-order entropy. This typically represents the expected value of the out-of-order entropy measured when the auxiliary control system is brand new or fully healthy (potentially close to the noise floor). Secondly, the degree to which the current profile data deviates from the baseline value is quantified and mapped linearly or non-linearly to a score range of 0 to 100. For example, the calculation of a health score and maintenance suggestion generation for the network profile data and structured log entries, mapping the deviation between the updated out-of-order mean and the baseline value to a health score ranging from zero to one hundred, includes: determining the health score using the following formula: ; in, This indicates taking the maximum value. Rate your health. This is the new mean, i.e., the updated disordered mean. The tolerance threshold, This is the baseline value for the out-of-order entropy. This means that when... Equal to or better than At that time, the health score was close to 100 points; when Approaching or exceeding the tolerance threshold The health score approaches 0. A health score is calculated for each important status. Simultaneously, structured log entries from recent periods (e.g., the past 24 hours) are scanned to analyze the frequency and distribution of various diagnostic events (especially abnormal events). Finally, the report generator integrates all information: listing the health scores for each status, highlighting statuses with health scores below a preset threshold (e.g., 60 points), and displaying statistical charts of key events. Furthermore, when the health score falls below the preset threshold, a communication network check suggestion is automatically generated, for example: "Status 'S3: Pump Unit Operation' health score dropped to 45, 12 'network micro-jitter' events were detected in the past hour, it is recommended to check the network connection and switch port of this pump unit controller." The document integrating quantitative scores, trend analysis, and specific suggestions is the final health report.
[0036] In summary, the dynamic diagnosis method for auxiliary control systems of thermal power plants based on time-series state machines according to embodiments of the present invention is explained. It slices and vectorizes the real-time event stream through a sliding time window to construct a set of currently observed events. It then analyzes the target event set under the current state using a pre-constructed state machine topology diagram. Furthermore, it quantifies the deviation of the event sequence by performing set matching and out-of-order measurement on observed and target events, and introduces a flexible tolerance mechanism to achieve flexible judgment of state transitions while considering network jitter and logical deviations. This effectively distinguishes between normal fluctuations, minor network jitter, and severe logical misordering. Finally, based on the transition state pointer, it accumulates trends and updates the health profile of the diagnostic results, generating a health report that can guide operation and maintenance. This effectively improves the adaptability to equipment state changes in complex industrial environments and significantly enhances the predictability of auxiliary control system operation and the scientific nature of operation and maintenance decisions.
[0037] An embodiment of the present invention also provides a dynamic diagnostic system for auxiliary control systems of thermal power plants based on a time-series state machine.
[0038] Figure 4 This is a block diagram of a dynamic diagnostic system for a thermal power plant auxiliary control system based on a time-sequential state machine, according to an embodiment of the present invention. Figure 4 As shown, the dynamic diagnostic system 300 for a thermal power plant auxiliary control system based on a time-series state machine according to an embodiment of the present invention includes: a windowed event observation module 310, used to perform buffering, slicing, and vectorization processing on the event stream of the thermal power plant auxiliary control system acquired in real time based on a preset sliding time window to obtain the observed event set within the current window; a state graph subsequent event parsing module 320, used to perform successor node search and concurrent logic block parsing on a pre-constructed state machine topology graph based on the current state pointer to obtain the target event set under the current state; an event set matching and disorder quantification module 330, used to perform set similarity matching and disorder measurement on the observed event set and the target event set to obtain a matching flag and disorder entropy value; a flexible state transition decision module 340, used to perform flexible tolerance judgment and state transition decision on the current state transition conditions based on the matching flag, disorder entropy value, and preset tolerance threshold to obtain a diagnostic category and a transition state pointer; and a health trend accumulation and report generation module 350, used to perform trend accumulation and health profile update on the diagnostic category and disorder entropy value using the transition state pointer as an associated index to generate a health report.
[0039] The specific implementation method of the dynamic diagnosis system for auxiliary control system of thermal power plant based on time-sequential state machine provided in the embodiments of the present invention can be found in the description of the dynamic diagnosis method for auxiliary control system of thermal power plant based on time-sequential state machine provided in the embodiments of the present invention, and will not be repeated here.
[0040] The dynamic diagnostic system 300 for auxiliary control systems of thermal power plants based on a time-sequential state machine, according to embodiments of the present invention, can be implemented in various wireless terminals, such as servers with a dynamic diagnostic algorithm for auxiliary control systems of thermal power plants based on a time-sequential state machine. In one possible implementation, the dynamic diagnostic system 300 for auxiliary control systems of thermal power plants based on a time-sequential state machine, according to embodiments of the present invention, can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the dynamic diagnostic system 300 for auxiliary control systems of thermal power plants based on a time-sequential state machine can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the dynamic diagnostic system 300 for auxiliary control systems of thermal power plants based on a time-sequential state machine can also be one of many hardware modules of the wireless terminal.
[0041] Alternatively, in another example, the dynamic diagnostic system 300 for the auxiliary control system of a thermal power plant based on a sequential state machine and the wireless terminal can also be separate devices, and the dynamic diagnostic system 300 for the auxiliary control system of a thermal power plant based on a sequential state machine can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.
[0042] Those skilled in the art will understand that the above embodiments are specific implementations of the present invention, and in practical applications, various changes can be made in form and detail without departing from the spirit and scope of the present invention.
Claims
1. A dynamic diagnostic method for auxiliary control systems of thermal power plants based on sequential state machines, characterized in that, include: Step S1: Based on a preset sliding time window, perform buffering, slicing, and vectorization processing on the event stream of the thermal power plant auxiliary control system acquired in real time to obtain the set of observed events within the current window; Step S2: Based on the current state pointer, perform successor node search and concurrent logic block parsing on the pre-built state machine topology graph to obtain the target event set in the current state; Step S3: Perform set similarity matching and out-of-order measurement on the observed event set and the target event set to obtain the matching flag and out-of-order entropy value; Step S4: Based on the matching flag, out-of-order entropy value and preset tolerance threshold, perform elastic tolerance judgment and state transition decision on the current state transition conditions to obtain the diagnosis category and transition state pointer. Step S5: Using the migration status pointer as the associated index, perform trend accumulation and health profile updates for the diagnosis category and out-of-order entropy value to generate a health report.
2. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequenced state machines according to claim 1, characterized in that, The event flow includes auxiliary machine current values, valve feedback positions, pipeline pressure values, and switching signals.
3. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 1, characterized in that, Step S1 includes: Based on the quality bit verification rule, the validity filtering and cache injection of each event packet in the event stream are performed. The valid events that pass the verification are pushed into the first-in-first-out circular buffer in the order of arrival to obtain the real-time event cache queue. Based on the current system clock and the sliding time window, the real-time event cache queue is subjected to backtracking interval truncation. All events whose source timestamps fall within the closed interval defined by a window length traced back from the current time are selected to obtain the candidate event sequence. The candidate event sequence is processed by extracting the identifier and timestamp tuples of each event and encapsulating them into a set to obtain the observed event set.
4. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 1, characterized in that, Step S2 includes: Based on the state pointer, outgoing edge traversal and condition pre-filtering are performed in the state machine topology graph to identify the successor candidate nodes on all active paths of the current node in order to obtain the list of next-hop nodes. Based on the attribute type label of each node in the state machine topology diagram, the complex logic block of each node in the next hop node list is atomically decomposed to obtain the atomic event list. The atomic event list is deduplicated and standardized, and the expected logical values are bound to each event identifier according to the expected state defined in the state machine topology diagram to obtain the target event set.
5. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 1, characterized in that, Step S3 includes: By traversing each element in the target event set and searching for the corresponding event with the matching flag in the observation event set, performing set intersection operation and cardinality comparison, a subset of matching events consisting of successfully matched events is obtained. If the number of elements in the subset of matching events is equal to the number of elements in the target event set, the matching flag is set to true; otherwise, the matching flag is set to false. Based on the predefined logical index order of each element in the target event set, a typical order mapping and time series reconstruction are performed on the matching event subset to obtain a typical order time series. The out-of-order entropy value is obtained by counting inversion pairs and calculating normalized entropy values for typical sequential time series.
6. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 5, characterized in that, To obtain the out-of-order entropy value, the following formula is used to calculate the inversion pair count and normalized entropy value of a typical time series: ； in, This is the out-of-order entropy value. The total number of events, It is a typical time series. The first in a typical time series A timestamp, This represents the j-th timestamp in a typical ordered time series. This is an indicator function; it takes the value 1 when the condition within the parentheses is true, and 0 otherwise.
7. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 1, characterized in that, Step S4 includes: Based on the entry time of the current state and the preset timeout threshold, the matching flag is jointly verified for integrity and timeout status to obtain the pre-verified status code. Based on the pre-checked status code and tolerance threshold, the out-of-order entropy value is segmented logically to obtain the diagnostic category. Based on the state transition table, a transition mapping lookup is performed between the diagnostic category and the current state pointer to generate a transition state pointer.
8. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-sequential state machines according to claim 1, characterized in that, Step S5 includes: Based on the enumeration value of the diagnostic category, a hierarchical alarm strategy is implemented and logs are persisted for the current diagnostic event to obtain structured log entries. Based on the exponential moving average algorithm and the migration state pointer as the primary key index, the out-of-order entropy value and the historical out-of-order mean value corresponding to the state retrieved from the database are weighted and fused to update the network profile data. The system calculates health scores and generates maintenance recommendations based on network profile data and structured log entries. It maps the deviation between the updated out-of-order mean and the baseline value to a health score ranging from zero to one hundred. When the health score falls below a preset threshold, it automatically generates communication network inspection recommendations to produce a health report.
9. The dynamic diagnostic method for auxiliary control systems of thermal power plants based on time-series state machines according to claim 8, characterized in that, Health scores are calculated and maintenance recommendations are generated based on network profile data and structured log entries. The deviation between the updated out-of-order mean and the baseline value is mapped to a health score ranging from zero to one hundred, including: determining the health score using the following formula: ； in, Rate your health. This indicates taking the maximum value. This is the updated random mean. This serves as the baseline value for the out-of-order entropy. This is the tolerance threshold.
10. A dynamic diagnostic system for auxiliary control systems of thermal power plants based on a time-sequential state machine, characterized in that, include: The windowed event observation module is used to perform buffering, slicing, and vectorization processing on the event stream of the auxiliary control system of the thermal power plant acquired in real time based on a preset sliding time window to obtain the set of observed events in the current window. The state diagram subsequent event parsing module is used to perform successor node search and concurrent logic block parsing on the pre-built state machine topology diagram based on the current state pointer to obtain the target event set in the current state; The event set matching and disordered quantization module is used to perform set similarity matching and disordered quantification on the observed event set and the target event set to obtain the matching flag and disordered entropy value. The elastic state transition decision module is used to make elastic tolerance judgments and state transition decisions on the current state transition conditions based on the matching flag, out-of-order entropy value and preset tolerance threshold, so as to obtain the diagnosis category and transition state pointer. The Health Trend Accumulation and Report Generation module is used to accumulate trends and update health profiles for diagnostic categories and out-of-order entropy values using migration status pointers as associated indexes, in order to generate health reports.