Thermal power generating unit safety interlocking method and system based on protection logic

Abstract

The invention provides a thermal power generating unit safety interlocking method and system based on protection logic, and relates to the technical field of thermal power generation control, and the method comprises the steps: collecting operation parameters, analyzing a time sequence change rule, and constructing a protection logic incidence matrix; the association strength is converted into a distance relation, and a multi-level interlocking protection structure is constructed through adaptive clustering; tracking a parameter change transmission path, and selecting a control strategy based on risk assessment; and adjusting a control period through the adaptive factor, and determining an execution time sequence. According to the method, the accuracy, the real-time performance and the adaptability of safety interlocking control of the thermal power generating unit can be improved.

Description

Technical Field

[0001] This invention relates to the field of thermal power generation control technology, and in particular to a safety interlocking method and system for thermal power units based on protection logic. Background Technology

[0002] As a crucial component of the power system, the safe and stable operation of thermal power units has a significant impact on the reliability of the power grid. Thermal power units involve the coordinated operation of multiple subsystems, including fuel supply, boilers, turbines, and generators, resulting in high system complexity and close technological coupling between these subsystems. During operation, the parameters of each system influence each other; any anomaly in any link may trigger a chain reaction leading to system failure or even a major accident. Therefore, establishing an effective safety interlocking protection system is of paramount importance for the safe operation of thermal power units.

[0003] Traditional thermal power unit safety interlocking systems are mainly based on preset thresholds and simple logical relationships, triggering protection actions by monitoring whether key parameters exceed safety limits. As thermal power units develop towards larger capacity, higher parameters, and higher efficiency, system complexity has increased significantly, the coupling relationships between process parameters have become more intricate, and the interactions between equipment have become more pronounced, placing higher demands on safety interlocking protection systems.

[0004] However, existing safety interlocking technologies for thermal power units still have defects and shortcomings. Existing interlocking protection systems typically use fixed threshold judgment methods, which are difficult to adapt to the dynamic operating characteristics of thermal power units under different operating conditions. They cannot effectively identify the temporal change patterns and interactive influence characteristics between parameters, which may lead to protection failures or false protection under complex operating conditions. Traditional interlocking systems lack the ability to analyze the fault propagation path and scope of influence, and cannot achieve accurate tracking and prediction of fault propagation. When faced with cascading faults, it is difficult to determine the optimal interlocking execution strategy, which can easily cause unnecessary unit shutdowns or safety hazards. The control strategies of existing interlocking protection systems are relatively rigid, and the command execution sequence and timing lack dynamic optimization capabilities. They cannot adaptively adjust according to the real-time operating status and response characteristics of thermal power units, resulting in insufficient timeliness and accuracy of interlocking protection, affecting the safety level and economy of the unit. Summary of the Invention

[0005] This invention provides a safety interlocking method and system for thermal power units based on protection logic, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a safety interlocking method for thermal power units based on protection logic, comprising: Collect operating parameters of thermal power units; By analyzing the time-series variation of operating parameters of thermal power units through a coupled evaluation model, extracting the interactive influence characteristics of operating parameters, constructing a protection logic correlation matrix, and calculating the protection logic correlation strength value; The protection logic association strength value is converted into a distance relationship to construct an association distance matrix. Dynamic classification is performed through adaptive clustering radius. Iterative calculation is performed based on inter-class merging metric to generate a multi-level interlocking protection structure. By using the real-time deviation values ​​of thermal power unit operating parameters to trigger ripple effect analysis, the transmission path of operating parameter changes is traced in the multi-level interlocking protection structure, and related influencing nodes are identified. Key impact nodes are mapped to the risk propagation network. The strength of fuzzy rules is dynamically adjusted based on the propagation coupling coefficient, and a historical adjustment trajectory is established. Fuzzy rule reasoning is performed in combination with the historical adjustment trajectory to obtain the risk assessment value. Based on the risk assessment value, the interlocking control strategy is selected, and the order of instruction execution is determined according to the energy transfer chain of the thermal power unit process, forming an interlocking protection instruction sequence. By adjusting the control cycle parameters through the adaptive factor, fitting analysis is performed using the dynamic response characteristic curve to calculate the response time constraint and determine the execution sequence of the interlocking protection command. Record the response characteristic curves of thermal power units during the execution of interlocking protection commands.

[0007] In one optional embodiment, the protection logic association strength values ​​are converted into distance relationships to construct an association distance matrix. Dynamic classification is performed using an adaptive clustering radius, and iterative calculations are executed based on inter-class merging metrics to generate a multi-level interlocking protection structure, including: The protection logic association strength value is converted into the distance relationship between protected objects and preprocessed to calculate the association distance matrix of the protected objects. Based on the aforementioned correlation distance matrix, an adaptive clustering radius adjustment method is used to dynamically classify the protected objects, thereby obtaining the classification results of the protected objects. Based on the classification results of the protected objects, the inter-class closeness value and inter-class separation value are determined and combined to form a comprehensive evaluation value. Based on the comprehensive evaluation value, the merging metric between class pairs in the classification results of the protected objects is calculated. The class pair with the optimal merging metric is selected for iterative merging and the association distance matrix is ​​updated. When the preset maximum number of levels is reached or the similarity between all classes in the association distance matrix is ​​less than the preset similarity threshold, the iteration stops and the initial interlocking protection structure is obtained. Calculate the connectivity of each layer of protected objects and the transmission path length between each layer in the initial interlocking protection structure. Construct an important evaluation index for protected objects based on the connectivity and a hierarchical transmission efficiency index based on the transmission path length. Based on the important evaluation index for protected objects and the hierarchical transmission efficiency index, reallocate the layers of protected objects and reconstruct the association between the protected objects to generate an optimized multi-level interlocking protection structure.

[0008] In one optional embodiment, the adaptive clustering radius adjustment method includes: Based on the correlation distance matrix, the initial clustering radius is set, and the average intra-cluster distance and the minimum inter-cluster distance are calculated. The adaptive factor is determined based on the ratio of the average intra-class distance to the minimum inter-class distance; The clustering radius is dynamically adjusted according to a preset step size using the adaptive factor. The adaptive adjustment of the clustering radius is completed when the ratio of the average distance within a cluster to the minimum distance between clusters is within a preset range. During the adjustment process, the inter-class distance distribution characteristics after each adjustment are recorded to construct a historical adjustment trajectory. The direction of the next adjustment is predicted based on the historical adjustment trajectory, and the direction of the next adjustment is corrected by combining the current inter-class boundary density.

[0009] In one optional embodiment, the real-time deviation value of the operating parameters of the thermal power unit is used to trigger a ripple effect analysis. This involves tracing the transmission path of operating parameter changes within a multi-level interlocking protection structure and identifying associated influencing nodes, including: Based on the operating parameters of thermal power units, an operating parameter fluctuation range is established, and the real-time deviation value of the operating parameters is calculated. When the real-time deviation value exceeds the preset deviation threshold, a ripple effect analysis is triggered. Based on the real-time deviation value, the initial triggering parameter is located, and the interlocking protection action signal corresponding to the initial triggering parameter is obtained in the multi-level interlocking protection structure. Based on the interlocking protection action signal, the parameter change transmission path is tracked, and the response timing of each parameter on the parameter change transmission path is recorded. Obtain the propagation delay between adjacent parameters on the parameter change propagation path, calculate the influence attenuation ratio between adjacent parameters, and combine the influence attenuation ratio with the propagation delay to determine the influence weight of the propagation path; Based on the influence weight of the transmission path and the response timing, the influence intensity received by each node in the transmission path is calculated, and the nodes whose influence intensity exceeds a preset intensity threshold are identified as key influence nodes.

[0010] In one optional embodiment, the real-time deviation value of the operating parameters of the thermal power unit is used to trigger a ripple effect analysis. This involves tracing the transmission path of operating parameter changes within a multi-level interlocking protection structure and identifying associated influencing nodes, including: Map key impact nodes to the risk propagation network and construct propagation links between nodes; Based on the operating status parameters of the key influencing nodes, the response timing patterns in preset historical fault cases are extracted, and the propagation coupling coefficient between adjacent nodes on the propagation link is calculated based on the response timing patterns. Based on the propagation coupling coefficient, an initial fuzzy rule strength is established, and the rule strength difference between adjacent propagation links is calculated as the risk response ratio of the propagation link. The intensity of fuzzy rules is dynamically adjusted based on the correspondence between the rule intensity difference and the risk response ratio, and the risk propagation characteristics during the adjustment process are recorded to form a historical adjustment trajectory. The operational status parameters are input into the adjusted propagation path, and fuzzy rule reasoning is performed in conjunction with the historical adjustment trajectory to obtain the risk assessment value.

[0011] In an optional embodiment, the control cycle parameters are adjusted by controlling an adaptive factor, and the dynamic response characteristic curve is used for fitting analysis to calculate the response time constraint. The determination of the interlocking protection command execution sequence includes: The control adaptive factor is calculated by the ratio of the control error to the maximum permissible control error. The proportional gain coefficient and derivative gain coefficient of the control cycle are adjusted using the control adaptive factor to obtain the control cycle adjustment parameter. Based on the dynamic response data during the operation of thermal power units, the corresponding slope is determined, the fitting adaptive factor is calculated, and the fitting adaptive factor is combined with the Gompertz function to perform curve fitting to obtain the dynamic response characteristic curve. Calculate the response completion index corresponding to the dynamic response characteristic curve, determine the minimum response time based on the time when the response completion index reaches the preset completion threshold, and multiply the minimum response time by the coefficient of variation of the dynamic response characteristic curve to obtain the actual response time. A time optimization objective function containing a time deviation term and an adjacent time interval term is constructed. The actual response time is input into the state transition equation, and the time optimization objective function is solved using a dynamic programming algorithm to obtain the execution time interval of each control command. By combining the control cycle adjustment parameter with the execution time interval, the execution time of each control command is optimized to generate the interlocking protection command execution sequence.

[0012] In one optional embodiment, based on the dynamic response data during the operation of the thermal power unit, determining the corresponding slope and calculating the fitting adaptive factor includes: The dynamic response data is segmented according to the set time window interval, and the rate of change of adjacent data points is calculated to obtain the response change rate value within the time window. The response rate of change is decomposed into multiple scales using wavelet transform, and components of different frequency bands are extracted. Rate values ​​that exceed the preset fluctuation range are removed to obtain a corrected response rate of change sequence. Calculate the mean and standard deviation of the corrected response rate of change sequence within each time window to determine the fluctuation range of the response rate of change; The response acceleration segment and the response deceleration segment are divided according to the upper and lower limits of the fluctuation range, and the average rate of change of the response acceleration segment and the response deceleration segment are calculated as the dynamic response slope. The dynamic response slope is optimized using a recursive iterative method. When the difference between two adjacent dynamic response slopes is less than a preset convergence threshold, the optimization result of the current iteration is used as the fitting adaptive factor.

[0013] A second aspect of this invention provides a safety interlocking system for thermal power units based on protection logic, comprising: The first unit is used to collect operating parameters of thermal power units; The second unit is used to analyze the time-series variation of operating parameters of thermal power units through a coupled evaluation model, extract the interactive influence characteristics of operating parameters, construct the protection logic association matrix, and calculate the protection logic association strength value. The third unit is used to convert the protection logic association strength value into a distance relationship to construct an association distance matrix, perform dynamic classification through adaptive clustering radius, perform iterative calculations based on inter-class merging metric, and generate a multi-level interlocking protection structure. The fourth unit is used to trigger ripple effect analysis by utilizing the real-time deviation value of the operating parameters of thermal power units, to track the transmission path of changes in operating parameters in the multi-level interlocking protection structure, and to identify related influencing nodes. The fifth unit is used to map key influencing nodes to the risk propagation network, dynamically adjust the strength of fuzzy rules based on the propagation coupling coefficient and establish historical adjustment trajectories, and perform fuzzy rule reasoning in combination with the historical adjustment trajectories to obtain risk assessment values. The sixth unit is used to select interlocking control strategies based on risk assessment values, determine the order of instruction execution according to the energy transfer chain of the thermal power unit's process, and form an interlocking protection instruction sequence. The seventh unit is used to adjust the control cycle parameters by controlling the adaptive factor, perform fitting analysis using the dynamic response characteristic curve, calculate the response time constraint, and determine the execution sequence of the interlocking protection command. The eighth unit is used to record the response characteristic curves of thermal power units during the execution of interlocking protection commands.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] In this embodiment of the invention, by constructing a protection logic association matrix and dynamically classifying and generating a multi-level interlocking protection structure, the precision and systematization of safety protection for thermal power units are achieved. This avoids the malfunctions and missed protection problems caused by simple trigger-based protection in traditional methods, thus improving the safety and reliability of the system. By employing ripple effect analysis and risk propagation network assessment, the associated impact nodes can be accurately identified and the fuzzy rule strength can be dynamically adjusted, enabling accurate assessment of safety risks under complex operating conditions. Compared with traditional fixed threshold judgment methods, this significantly reduces the false alarm rate and improves the timeliness of risk warnings. By determining the instruction execution order based on the energy transfer chain and adjusting the control cycle parameters through adaptive factors, the intelligent coordinated execution of interlocking protection instructions is achieved. This solves the problem of secondary faults caused by instruction conflicts and unreasonable execution timing in traditional interlocking systems, effectively shortening the system response time. At the same time, recording response characteristic curves provides data support for system optimization. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the safety interlocking method for thermal power units based on protection logic, according to an embodiment of the present invention. Figure 2 Flowchart for optimizing the timing of interlocking protection commands. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the safety interlocking method for thermal power units based on protection logic, as described in an embodiment of the present invention. Figure 1 As shown, the method includes: Collect operating parameters of thermal power units; By analyzing the time-series variation of operating parameters of thermal power units through a coupled evaluation model, extracting the interactive influence characteristics of operating parameters, constructing a protection logic correlation matrix, and calculating the protection logic correlation strength value; The protection logic association strength value is converted into a distance relationship to construct an association distance matrix. Dynamic classification is performed through adaptive clustering radius. Iterative calculation is performed based on inter-class merging metric to generate a multi-level interlocking protection structure. By using the real-time deviation values ​​of thermal power unit operating parameters to trigger ripple effect analysis, the transmission path of operating parameter changes is traced in the multi-level interlocking protection structure, and related influencing nodes are identified. Key impact nodes are mapped to the risk propagation network. The strength of fuzzy rules is dynamically adjusted based on the propagation coupling coefficient, and a historical adjustment trajectory is established. Fuzzy rule reasoning is performed in combination with the historical adjustment trajectory to obtain the risk assessment value. Based on the risk assessment value, the interlocking control strategy is selected, and the order of instruction execution is determined according to the energy transfer chain of the thermal power unit process, forming an interlocking protection instruction sequence. By adjusting the control cycle parameters through the adaptive factor, fitting analysis is performed using the dynamic response characteristic curve to calculate the response time constraint and determine the execution sequence of the interlocking protection command. Record the response characteristic curves of thermal power units during the execution of interlocking protection commands.

[0021] In one optional implementation, the protection logic association strength values ​​are converted into distance relationships to construct an association distance matrix. Dynamic classification is performed using an adaptive clustering radius, and iterative calculations are executed based on inter-class merging metrics to generate a multi-level interlocking protection structure, including: The protection logic association strength value is converted into the distance relationship between protected objects and preprocessed to calculate the association distance matrix of the protected objects. Based on the aforementioned correlation distance matrix, an adaptive clustering radius adjustment method is used to dynamically classify the protected objects, thereby obtaining the classification results of the protected objects. Based on the classification results of the protected objects, the inter-class closeness value and inter-class separation value are determined and combined to form a comprehensive evaluation value. Based on the comprehensive evaluation value, the merging metric between class pairs in the classification results of the protected objects is calculated. The class pair with the optimal merging metric is selected for iterative merging and the association distance matrix is ​​updated. When the preset maximum number of levels is reached or the similarity between all classes in the association distance matrix is ​​less than the preset similarity threshold, the iteration stops and the initial interlocking protection structure is obtained. Calculate the connectivity of each layer of protected objects and the transmission path length between each layer in the initial interlocking protection structure. Construct an important evaluation index for protected objects based on the connectivity and a hierarchical transmission efficiency index based on the transmission path length. Based on the important evaluation index for protected objects and the hierarchical transmission efficiency index, reallocate the layers of protected objects and reconstruct the association between the protected objects to generate an optimized multi-level interlocking protection structure.

[0022] In one specific implementation, the protection logic association strength value is transformed into a distance relationship between protected objects and preprocessed to calculate the association distance matrix of the protected objects. Specifically, logical association data of each protected object in the thermal power unit is collected, including protection triggering conditions, protected objects, and their degree of influence. For example, the association strength between turbine overspeed protection and generator trip protection is 0.85, indicating that there is an 85% probability that the turbine overspeed protection will trigger the generator trip protection. The association strength between boiler low water level protection and feedwater pump trip protection is 0.92, indicating that there is a 92% probability that the boiler low water level protection will trigger the feedwater pump trip protection. The reciprocal of the association strength value is calculated as the distance between protected objects; for example, the distance between turbine overspeed protection and generator trip protection is 1 / 0.85 ≈ 1.18, and the distance between boiler low water level protection and feedwater pump trip protection is 1 / 0.92 ≈ 1.09. Considering the influence of data noise and outliers, the distance matrix is ​​normalized, limiting the distance values ​​to between 0 and 1, thus obtaining the preprocessed association distance matrix.

[0023] Based on the association distance matrix, an adaptive clustering radius adjustment method is used to dynamically classify the protected objects. In the initial stage, each protected object is treated as an independent class, and the initial clustering radius is set to the minimum value of the non-zero elements in the association distance matrix, such as 0.1. For the 50 protected objects in a thermal power unit, there are initially 50 classes. As the clustering process progresses, the clustering radius is dynamically adjusted, with a growth rate set to 15% of the original radius. When the distance between two protected objects is less than the current clustering radius, they are grouped into the same class. For example, when the clustering radius is 0.12, boiler low water level protection and feedwater pump trip protection (normalized distance 0.11) are grouped into the same class, while turbine overspeed protection and generator trip protection (normalized distance 0.14) remain independent classes. As the clustering radius increases to 0.15, turbine overspeed protection and generator trip protection are also grouped into the same class. In this way, 10 protected object categories are ultimately obtained, each containing protected objects with close logical relationships.

[0024] Based on the classification results of the protected objects, the inter-class closeness value and inter-class separation value are determined and combined to form a comprehensive evaluation value. For each pair of categories, the average distance between their internal members is calculated as the inter-class closeness value, and the distance between the nearest members of the two categories is calculated as the inter-class separation value. For example, the average internal distance of category 1 (including boiler-related protection) is 0.18, the average internal distance of category 2 (including steam turbine-related protection) is 0.22, and the distance between the nearest members of category 1 and category 2 is 0.35. The comprehensive evaluation value is calculated by weighting, with weights of 0.6 and 0.4, i.e., comprehensive evaluation value = 0.6 × (inter-class closeness value) + 0.4 × (1 - inter-class separation value). For categories 1 and 2, the comprehensive evaluation value is 0.6 × ((0.18 + 0.22) / 2) + 0.4 × (1 - 0.35) = 0.26. The merging metric between all class pairs is calculated, and the class pair with the optimal merging metric (smallest value) is selected for iterative merging, and the association distance matrix is ​​updated. For example, the merging metric for Category 3 (which includes condenser-related protection) and Category 4 (which includes cooling water system-related protection) is 0.19, which is the smallest among all class pairs. Therefore, Categories 3 and 4 are merged into a new category. The iteration continues until a preset maximum number of levels (e.g., 4 levels) is reached or the similarity between all classes is less than a preset similarity threshold (e.g., 0.1), at which point the initial interlocking protection structure is obtained.

[0025] Calculate the connectivity of each layer of protected objects and the transmission path length between layers in the initial interlocking protection structure. Connectivity represents the degree of association between a protected object and other protected objects. It is calculated by dividing the sum of the association strengths of the protected object with other protected objects at the same layer by the maximum possible association strength value. For example, the sum of the association strengths of the turbine overspeed protection with 8 protected objects at the same layer is 5.2, and the maximum possible value is 8, so its connectivity is 5.2 / 8 = 0.65. Transmission path length represents the efficiency of influence transmission between protected objects at different levels. It is calculated by the shortest path length between two layers of protected objects. For example, the shortest path length between the boiler layer and the turbine layer is 2, indicating that it takes 2 transmission steps for the protected object at the boiler layer to trigger and affect the protected object at the turbine layer.

[0026] An importance evaluation index for protected objects is constructed based on connectivity, and a hierarchical transmission efficiency index is constructed based on transmission path length. The importance evaluation index considers both connectivity and the triggering frequency of the protected object, and is calculated by multiplying connectivity by the normalized triggering frequency. For example, the connectivity of turbine overspeed protection is 0.65, and the normalized triggering frequency is 0.8, so its importance evaluation index is 0.65 × 0.8 = 0.52. The hierarchical transmission efficiency index considers the inter-layer transmission path length and transmission reliability, and is calculated by dividing 1 by the path length and then multiplying by the transmission reliability. For example, the path length between the boiler layer and the turbine layer is 2, and the transmission reliability is 0.9, so the transmission efficiency index is (1 / 2) × 0.9 = 0.45.

[0027] The hierarchy of protected objects is redistributed and the relationships between them are reconstructed based on the importance evaluation index and the hierarchical transfer efficiency index. Protected objects with an importance index higher than 0.5 are upgraded to the next higher level, while those with an importance index lower than 0.2 are downgraded to the next lower level. For example, the importance index of feedwater pump trip protection is 0.55, so it is upgraded from level 2 to level 1; while the importance index of condenser low vacuum protection is 0.18, so it is downgraded from level 2 to level 3. For inter-level relationships with a hierarchical transfer efficiency index lower than 0.3, direct connections are added to improve transfer efficiency. For example, the transfer efficiency index between the boiler level and the turbine level is 0.25, and the transfer efficiency is improved by adding direct connections between the two key protected objects. In this way, an optimized multi-level interlocking protection structure is finally generated, including four levels. The first level contains the 10 most critical protected objects, the second level contains 15 protected objects, the third level contains 18 protected objects, and the fourth level contains 7 protected objects. An efficient transmission relationship is established between the levels to ensure that the thermal power unit can trigger the corresponding protection measures in order of priority under abnormal conditions, thereby improving the response speed and reliability of the safety interlocking system.

[0028] The multi-level interlocking protection structure constructed by the above method not only considers the logical relationship between the protected objects, but also fully considers the importance of the protected objects and the transmission efficiency between levels. This enables the thermal power unit to quickly and reliably trigger corresponding protection measures when facing various abnormal situations, avoid equipment damage and safety accidents, and improve the overall safety and reliability of the thermal power unit.

[0029] In one optional implementation, the adaptive clustering radius adjustment method includes: Based on the correlation distance matrix, the initial clustering radius is set, and the average intra-cluster distance and the minimum inter-cluster distance are calculated. The adaptive factor is determined based on the ratio of the average intra-class distance to the minimum inter-class distance; The clustering radius is dynamically adjusted according to a preset step size using the adaptive factor. The adaptive adjustment of the clustering radius is completed when the ratio of the average distance within a cluster to the minimum distance between clusters is within a preset range. During the adjustment process, the inter-class distance distribution characteristics after each adjustment are recorded to construct a historical adjustment trajectory. The direction of the next adjustment is predicted based on the historical adjustment trajectory, and the direction of the next adjustment is corrected by combining the current inter-class boundary density.

[0030] In one specific implementation, the initial clustering radius is set based on the association distance matrix. Specifically, the system calculates the distance values ​​between all pairs of sample points, forming an N×N association distance matrix, where N is the number of sample points. The initial clustering radius can be set as the average of all distance values ​​in the association distance matrix multiplied by an initial coefficient. For example, the initial coefficient can be set to 0.5, meaning the initial clustering radius is half the average distance between all sample points. Taking a dataset containing 100 sample points as an example, assuming the calculated average distance between all pairs of points is 10 units, the initial clustering radius can be set to 5 units.

[0031] After setting the initial cluster radius, perform preliminary clustering and calculate the average intra-cluster distance and the minimum inter-cluster distance. The average intra-cluster distance is the average distance between all pairs of sample points within the same cluster. The minimum inter-cluster distance is the distance between the two nearest sample points in different clusters. For example, if the preliminary clustering results in three clusters, assuming the average intra-cluster distances of the three clusters are 2.5, 3.0, and 2.8 units respectively, the overall average intra-cluster distance is 2.77 units; and the minimum inter-cluster distance is 6.2 units (occurring between cluster 1 and cluster 2).

[0032] The adaptive factor is determined based on the ratio of the average intra-class distance to the minimum inter-class distance. The ratio is calculated by dividing the average intra-class distance by the minimum inter-class distance. In the example above, this ratio is approximately 2.77 / 6.2 ≈ 0.45. The adaptive factor can be obtained through linear or non-linear mapping of this ratio. In one embodiment, when the ratio is less than 0.3, the adaptive factor is set to 1.2, indicating a need to increase the cluster radius; when the ratio is greater than 0.7, the adaptive factor is set to 0.8, indicating a need to decrease the cluster radius; when the ratio is between 0.3 and 0.7, the adaptive factor can be calculated through linear interpolation, for example, adaptive factor = 1.2 - 0.4 × (ratio - 0.3) / 0.4. In the example above, the ratio is 0.45, and the calculated adaptive factor is 1.2 - 0.4 × (0.45 - 0.3) / 0.4 = 1.2 - 0.15 = 1.05.

[0033] The cluster radius is dynamically adjusted using an adaptive factor with a preset step size. The adjustment method is: New cluster radius = Current cluster radius × Adaptive factor. The preset step size can be understood as the magnitude of each adjustment, determined by the size of the adaptive factor. In the example above, assuming the current cluster radius is 5 units and the adaptive factor is 1.05, the new cluster radius is 5 × 1.05 = 5.25 units.

[0034] Re-clustering is performed using the new cluster radius, and the ratio of the average intra-cluster distance to the minimum inter-cluster distance is recalculated. When this ratio falls within a preset range (e.g., 0.4 to 0.6), the adaptive adjustment of the cluster radius is considered complete. If the ratio is still outside the preset range, the cluster radius will continue to be adjusted. In the example above, after adjustment, the average intra-cluster distance might be 3.1 units, the minimum inter-cluster distance 6.0 units, and the ratio 3.1 / 6.0 ≈ 0.52, which falls within the preset range of 0.4 to 0.6, thus completing the cluster radius adjustment.

[0035] During the adjustment process, the distribution characteristics of inter-class distances after each adjustment are recorded to construct a historical adjustment trajectory. The distribution characteristics of inter-class distances can include the minimum inter-class distance, the variance of inter-class distances, and the shape of the inter-class distance distribution. The historical adjustment trajectory refers to the data sequence that records each adjustment of the cluster radius and the corresponding changes in the inter-class distance characteristics. For example, it can be recorded that in five consecutive adjustments, the cluster radius is [5, 5.25, 5.51, 5.79, 6.08] units, the corresponding minimum inter-class distance is [6.2, 6.0, 5.9, 5.75, 5.6] units, and the average intra-class distance is [2.77, 3.1, 3.25, 3.38, 3.51] units.

[0036] Predicting the direction of the next adjustment based on historical adjustment trajectories. The prediction method can employ trend analysis, observing the changing trend of the ratio of the average intra-cluster distance to the minimum inter-cluster distance. If this ratio continuously increases, it indicates that the clusters are becoming increasingly compact, and the next adjustment may require a reduction in the cluster radius; conversely, if the ratio continuously decreases, it indicates that the clusters are becoming increasingly dispersed, and the next adjustment may require an increase in the cluster radius. In the historical trajectory of the example above, the calculated ratio sequence is [0.45, 0.52, 0.55, 0.59, 0.63], showing a continuous upward trend, predicting that the next adjustment should reduce the cluster radius.

[0037] The adjustment direction is then adjusted based on the current inter-cluster boundary density. Inter-cluster boundary density refers to the density of sample points near the boundaries of two clusters. A high boundary density indicates many critical points between the two clusters, resulting in blurred cluster boundaries. In this case, the cluster radius should be reduced to enhance the discriminative power between clusters. Conversely, a low boundary density indicates clear cluster boundaries, allowing for a more comprehensive inclusion of relevant points by increasing the cluster radius. In practical applications, the sample point density in each cluster boundary region (e.g., within 20% of the current cluster radius from the boundary) can be calculated and compared with the density in the cluster center region. If the boundary region density exceeds 80% of the center region density, the boundary density is considered high, and the adjustment direction should be adjusted to reduce the cluster radius. If the boundary region density is less than 30% of the center region density, the boundaries are considered clear, and the adjustment direction should be adjusted to increase the cluster radius.

[0038] The above adaptive clustering radius adjustment method can intelligently adjust clustering parameters according to data characteristics, effectively improving clustering quality and making it suitable for various complex data scenarios.

[0039] In one optional implementation, the real-time deviation value of the operating parameters of the thermal power unit is used to trigger a ripple effect analysis. This involves tracing the transmission path of operating parameter changes within a multi-level interlocking protection structure and identifying associated influencing nodes, including: Based on the operating parameters of thermal power units, an operating parameter fluctuation range is established, and the real-time deviation value of the operating parameters is calculated. When the real-time deviation value exceeds the preset deviation threshold, a ripple effect analysis is triggered. Based on the real-time deviation value, the initial triggering parameter is located, and the interlocking protection action signal corresponding to the initial triggering parameter is obtained in the multi-level interlocking protection structure. Based on the interlocking protection action signal, the parameter change transmission path is tracked, and the response timing of each parameter on the parameter change transmission path is recorded. Obtain the propagation delay between adjacent parameters on the parameter change propagation path, calculate the influence attenuation ratio between adjacent parameters, and combine the influence attenuation ratio with the propagation delay to determine the influence weight of the propagation path; Based on the influence weight of the transmission path and the response timing, the influence intensity received by each node in the transmission path is calculated, and the nodes whose influence intensity exceeds a preset intensity threshold are identified as key influence nodes.

[0040] In one specific implementation, based on the operating parameters of the thermal power unit, a fluctuation range for the operating parameters is established, and the real-time deviation value of the operating parameters is calculated. When the real-time deviation value exceeds a preset deviation threshold, a ripple effect analysis is triggered. Specifically, key operating parameters of the thermal power unit are collected in real time, including boiler steam pressure, turbine speed, generator power, and condenser vacuum. Through historical data analysis, normal fluctuation ranges for each parameter are established. For example, the normal fluctuation range for boiler main steam pressure is 8.5-9.5 MPa, the normal fluctuation range for turbine speed is 2990-3010 rpm, and the normal fluctuation range for generator power is 295-305 MW. The deviation between the current parameter value and the parameter set value is calculated in real time to obtain the real-time deviation value. When the real-time deviation value exceeds a preset deviation threshold, a ripple effect analysis is triggered. For example, the preset deviation threshold for boiler main steam pressure is ±0.8 MPa. When the measured pressure is 10.4 MPa, the real-time deviation value is 1.4 MPa, exceeding the preset deviation threshold, thus triggering a ripple effect analysis.

[0041] The initial trigger parameter is located based on the real-time deviation value. Within the multi-level interlocking protection structure, the corresponding interlocking protection action signal is obtained. The parameter change propagation path is traced based on the interlocking protection action signal, and the response timing of each parameter along the propagation path is recorded. In the example above, the boiler main steam pressure is located as the initial trigger parameter. Within the multi-level interlocking protection structure, interlocking protection action signals related to the boiler main steam pressure are searched, such as the boiler high-pressure protection signal. This signal triggers a series of interlocking protection actions, including boiler emergency shutdown, turbine tripping, and generator disconnection. By analyzing the propagation relationship of the interlocking protection action signals, the propagation path of parameter changes is traced. For example, boiler main steam pressure increases → boiler high-pressure protection activates → main steam valve closes → turbine steam intake decreases → turbine speed decreases → generator power decreases. Simultaneously, the response timing of each parameter change is recorded. For example, the main steam pressure of the boiler reaches 10.4 MPa at 0 seconds, the boiler high-pressure protection activates at 0.5 seconds, the main steam valve begins to close at 1.2 seconds, the turbine speed begins to decrease at 3.5 seconds, and the generator power begins to decrease at 4.8 seconds.

[0042] The propagation delay between adjacent parameters along the parameter change propagation path is obtained. The influence attenuation ratio between adjacent parameters is calculated, and the influence attenuation ratio is combined with the propagation delay to determine the influence weight of the propagation path. The propagation delay refers to the response time difference between adjacent parameters. For example, the propagation delay from the boiler high-pressure protection action to the closure of the main steam valve is 0.7 seconds (1.2-0.5), and the propagation delay from the closure of the main steam valve to the decrease in turbine speed is 2.3 seconds (3.5-1.2). The influence attenuation ratio is the ratio of the change amplitude of the latter parameter to the change amplitude of the former parameter. For example, if the boiler main steam pressure increases by 1.4 MPa, causing the main steam valve to close by 30%, the influence attenuation ratio is 0.21 (0.3 / 1.4); if the main steam valve closes by 30%, causing the turbine speed to decrease by 15 rpm, the influence attenuation ratio is 0.5 (15 / 30). The influence weight of the propagation path is calculated by dividing the influence attenuation ratio by the propagation delay. For example, the influence weight of the transmission path from the boiler high-pressure protection action to the closure of the main steam valve is 0.3 (0.21 / 0.7), and the influence weight of the transmission path from the closure of the main steam valve to the decrease in turbine speed is 0.217 (0.5 / 2.3).

[0043] Based on the influence weights of the transmission path and the response timing, the influence intensity received by each node in the transmission path is calculated, and nodes whose influence intensity exceeds a preset threshold are identified as critical influence nodes. The calculation of influence intensity considers the deviation of the initial triggering parameters, the cumulative influence weights on the transmission path, and the response timing. The influence intensity of the initial node is equal to the ratio of the real-time deviation value to the deviation threshold; for example, the influence intensity of the boiler main steam pressure is 1.75 (1.4 / 0.8). The influence intensity of subsequent nodes is calculated by multiplying the influence intensity of the previous node by the influence weight of the transmission path, and considering the time decay factor. For example, the influence intensity of the main steam valve node is 1.75 × 0.3 × 0.9 = 0.473 (time decay factor is 0.9), and the influence intensity of the turbine speed node is 0.473 × 0.217 × 0.85 = 0.087 (time decay factor is 0.85). Assuming the preset intensity threshold is 0.4, the boiler main steam pressure node (1.75) and the main steam valve node (0.473) are identified as critical influencing nodes, while the turbine speed node (0.087) is not a critical influencing node.

[0044] Using the above method, when abnormal operating parameters of thermal power units occur, the initial triggering parameters can be quickly located, the transmission path of parameter changes can be analyzed, and key influencing nodes can be identified, providing accurate decision-making basis for the safety interlocking system. This method considers the transmission delay, impact attenuation, and time decay factors of parameter changes, enabling a more accurate assessment of the impact of parameter anomalies on thermal power units and achieving more intelligent and reliable safety interlocking control.

[0045] Taking the abnormal increase in main steam pressure as an example, when the main steam pressure rises from 9.0 MPa to 10.4 MPa, the system detects a real-time deviation of 1.4 MPa, exceeding the preset deviation threshold of 0.8 MPa, triggering a ripple effect analysis. The initial triggering parameter is located as the boiler main steam pressure, and the corresponding interlock protection action signal "boiler high-pressure protection" is obtained. The parameter change transmission path is traced: boiler main steam pressure (10.4 MPa, 0 seconds) → boiler high-pressure protection (action, 0.5 seconds) → main steam valve (closed 30%, 1.2 seconds) → turbine speed (decreased by 15 rpm, 3.5 seconds) → generator power (decreased by 15 MW, 4.8 seconds). The transmission delay between adjacent parameters is calculated as follows: 0.5 seconds, 0.7 seconds, 2.3 seconds, 1.3 seconds; the influence attenuation ratio is calculated as follows: 1.0, 0.21, 0.5, 1.0; the influence weight of the transmission path is calculated as follows: 2.0, 0.3, 0.217, 0.769. Considering the time decay factor, the influence intensity of each node is calculated: boiler main steam pressure (1.75), boiler high-pressure protection (3.5), main steam valve (0.473), turbine speed (0.087), and generator power (0.057). Compared with the preset intensity threshold of 0.4, boiler main steam pressure, boiler high-pressure protection, and main steam valve are identified as critical influencing nodes. Based on the identification results of critical influencing nodes, the safety interlock system can prioritize handling the abnormal states of these nodes and take corresponding protective measures, such as reducing boiler load, adjusting feedwater flow, and monitoring the opening of the main steam valve, to avoid serious failures in the thermal power unit.

[0046] In this embodiment, by real-time monitoring and impact analysis of the operating parameters of thermal power units, the safety interlocking system is made intelligent and precise, which improves the safety and reliability of thermal power units, reduces unnecessary downtime events, and has important practical value.

[0047] In one optional implementation, key influencing nodes are mapped to a risk propagation network, the strength of fuzzy rules is dynamically adjusted based on the propagation coupling coefficient, and historical adjustment trajectories are established. Fuzzy rule inference is then performed using these historical adjustment trajectories to obtain a risk assessment value, including: Map key impact nodes to the risk propagation network and construct propagation links between nodes; Based on the operating status parameters of the key influencing nodes, the response timing patterns in preset historical fault cases are extracted, and the propagation coupling coefficient between adjacent nodes on the propagation link is calculated based on the response timing patterns. Based on the propagation coupling coefficient, an initial fuzzy rule strength is established, and the rule strength difference between adjacent propagation links is calculated as the risk response ratio of the propagation link. The intensity of fuzzy rules is dynamically adjusted based on the correspondence between the rule intensity difference and the risk response ratio, and the risk propagation characteristics during the adjustment process are recorded to form a historical adjustment trajectory. The operational status parameters are input into the adjusted propagation path, and fuzzy rule reasoning is performed in conjunction with the historical adjustment trajectory to obtain the risk assessment value.

[0048] In one specific implementation, key influencing nodes in the thermal power unit system are identified and mapped to construct a risk propagation network. Key influencing nodes include core equipment units such as the turbine itself, boiler feedwater system, condensate system, and generator cooling system. By analyzing the physical connections and control logic dependencies between nodes, propagation links between nodes are established. For example, there is a steam parameter transmission link between the boiler feedwater system and the turbine itself, and a water circulation link between the condensate system and the boiler feedwater system. The establishment of propagation links is based on the equipment topology and process flow diagram to ensure coverage of the main risk propagation paths.

[0049] Collect operational status parameters of key influencing nodes, including monitoring data on temperature, pressure, flow rate, and vibration. Extract the response time-series patterns of node parameters during various faults from a pre-set historical fault case database for thermal power units. For example, analyze the time-series relationship of abnormal turbine vibration caused by boiler feedwater pump failure, extracting the time-series pattern that after the feedwater pressure drops to 80% of the rated value, the turbine vibration value increases by 15% within 120 seconds. Through statistical analysis of multiple similar fault cases, obtain characteristic quantities such as the time delay and amplitude ratio of parameter changes between nodes. Based on these response time-series patterns, calculate the propagation coupling coefficient between adjacent nodes in the propagation chain. The propagation coupling coefficient is represented by a value between 0 and 1; the larger the value, the higher the coupling degree and the greater the probability of risk propagation. For example, the propagation coupling coefficient between the boiler feedwater system and the turbine body may be 0.85, indicating that there is an 85% probability that a feedwater system fault will propagate to the turbine system.

[0050] Based on the calculated propagation coupling coefficient, an initial fuzzy rule strength is established. The fuzzy rule is expressed in the form "If the state of node A is X, then the state of node B is Y," and the initial rule strength is assigned according to the propagation coupling coefficient. For example, for a link with a propagation coupling coefficient of 0.85, its initial fuzzy rule strength is set to 0.85. The rule strength difference between adjacent propagation links is calculated, which is the absolute value of the difference between the fuzzy rule strengths of two adjacent propagation links. Simultaneously, the risk response ratio of the propagation link is calculated, which is the ratio of the change in risk parameters of the downstream node to the change in risk parameters of the upstream node. For example, when the feedwater temperature rises by 10℃, the turbine vibration increases by 0.5 mm / s, and the risk response ratio is 0.05 mm / s / ℃.

[0051] The strength of fuzzy rules is dynamically adjusted based on the correlation between rule strength difference and risk response ratio. When the rule strength difference is less than 0.2 and the risk response ratio is greater than 0.8, the rule strength of the two links is adjusted to their average value; when the rule strength difference is greater than 0.5 and the risk response ratio is less than 0.3, the original rule strength remains unchanged; for other cases, the rule strength is adjusted by weighting according to the magnitude of the risk response ratio. During the adjustment process, the rule strength change value and the corresponding risk assessment accuracy change for each iteration are recorded to form a historical adjustment trajectory. The historical adjustment trajectory is stored as a data structure, containing information such as the adjustment time point, the rule strength before adjustment, the rule strength after adjustment, and the corresponding risk assessment accuracy.

[0052] The real-time collected operating status parameters are input into the adjusted propagation link network. For the boiler feedwater system's status parameters of 325℃ temperature, 16.5MPa pressure, and 410t / h flow rate, inference calculations are performed using adjusted fuzzy rules. The fuzzy rule inference process is executed by combining the rule adjustment direction and magnitude recorded in the historical adjustment trajectory. The fuzzy rule inference uses the max-min synthesis method, first fuzzifying the input parameters, then inferring based on rule strength, and finally defuzzifying to obtain the risk assessment value. The risk assessment value is represented by a value from 0 to 100, with higher values ​​indicating higher risk levels. For example, when the boiler feedwater system parameters are abnormal, the fuzzy rule inference yields a risk assessment value of 78 for the turbine system, indicating a high safety risk.

[0053] In this embodiment, by combining risk propagation network modeling and fuzzy rule reasoning, the limitations of traditional safety interlocking systems that rely solely on threshold triggering are overcome. This approach enables the perception of risk propagation relationships between devices, early detection of potential safety hazards, and provides effective protection for the safe operation of thermal power units.

[0054] In one optional implementation, the control cycle parameters are adjusted by controlling an adaptive factor, and the dynamic response characteristic curve is used for fitting analysis to calculate the response time constraint. The timing of the interlocking protection command execution is then determined, including: The control adaptive factor is calculated by the ratio of the control error to the maximum permissible control error. The proportional gain coefficient and derivative gain coefficient of the control cycle are adjusted using the control adaptive factor to obtain the control cycle adjustment parameter. Based on the dynamic response data during the operation of thermal power units, the corresponding slope is determined, the fitting adaptive factor is calculated, and the fitting adaptive factor is combined with the Gompertz function to perform curve fitting to obtain the dynamic response characteristic curve. Calculate the response completion index corresponding to the dynamic response characteristic curve, determine the minimum response time based on the time when the response completion index reaches the preset completion threshold, and multiply the minimum response time by the coefficient of variation of the dynamic response characteristic curve to obtain the actual response time. A time optimization objective function containing a time deviation term and an adjacent time interval term is constructed. The actual response time is input into the state transition equation, and the time optimization objective function is solved using a dynamic programming algorithm to obtain the execution time interval of each control command. By combining the control cycle adjustment parameter with the execution time interval, the execution time of each control command is optimized to generate the interlocking protection command execution sequence.

[0055] In one specific implementation, to achieve adaptive adjustment of control error, a control adaptive factor is calculated by the ratio of the control error to the maximum permissible control error. During the operation of the thermal power unit, the current control error value is collected in real time, representing the difference between the actual operating parameters and the set parameters. Simultaneously, the maximum permissible control error value is determined according to the thermal power unit's operating specifications; this value is typically preset based on the unit type, capacity, and operating environment. The error ratio is obtained by dividing the current control error value by the maximum permissible control error value. When the error ratio is less than 0.3, the control adaptive factor is set to 1.2; when the error ratio is between 0.3 and 0.7, the control adaptive factor is set to 1.0; and when the error ratio is greater than 0.7, the control adaptive factor is set to 0.8. For example, for a 600MW thermal power unit, when a turbine speed error of 30rpm is detected, while the maximum permissible error is 50rpm, the error ratio is 0.6, and the corresponding control adaptive factor is 1.0.

[0056] The calculated adaptive control factor is used to adjust the proportional gain coefficient and derivative gain coefficient of the control cycle. The original value of the proportional gain coefficient is multiplied by the adaptive control factor to obtain the adjusted proportional gain coefficient; the original value of the derivative gain coefficient is divided by the adaptive control factor to obtain the adjusted derivative gain coefficient. For example, when the original proportional gain coefficient is 2.5, the derivative gain coefficient is 0.5, and the adaptive control factor is 0.8, the adjusted proportional gain coefficient is 2.0, and the derivative gain coefficient is 0.625. These adjusted parameters constitute the control cycle adjustment parameters, which are used for subsequent control command execution.

[0057] The corresponding slope is determined based on the dynamic response data during the operation of thermal power units. Time-series data of key parameters such as boiler pressure, turbine speed, and generator power are collected during unit startup, load changes, or emergency conditions. These data are arranged chronologically, and the rate of change of parameters between adjacent time points, i.e., the slope, is calculated. For a 600MW unit during load ramp-up, it may be observed that during the load increase from 300MW to 400MW, the average load growth rate is 5MW / min for the first 5 minutes, 12MW / min for the middle 10 minutes, and 3MW / min for the last 5 minutes.

[0058] The fitting adaptive factor is calculated based on the calculated slope. The slopes for each time period are normalized by dividing each slope value by the maximum slope value. For the example above, the normalized slopes are 0.417, 1.0, and 0.25, respectively. The fitting adaptive factor is obtained by weighted averaging of the normalized slopes, with the weights determined according to the importance of each time period. For example, for the load increase process, the weight of the intermediate stage is set to 0.5, and the weights of the initial and final stages are each set to 0.25. Then the fitting adaptive factor is 0.417×0.25 + 1.0×0.5 + 0.25×0.25 = 0.667.

[0059] Curve fitting is performed by combining an adaptive fitting factor with the Gompertz function to obtain the dynamic response characteristic curve. The Gompertz function is an S-shaped curve, suitable for describing parameter changes during the response process of thermal power units. By adjusting the parameters of the Gompertz function, the fitting error between the function and the actual data points is minimized. The adaptive fitting factor is used to adjust the growth rate parameter of the function. For example, for a adaptive fitting factor of 0.667, the growth rate parameter of the Gompertz function can be adjusted to 0.667 times the original value to obtain a curve that better reflects the actual response characteristics.

[0060] Calculate the response completion rate index corresponding to the dynamic response characteristic curve. The response completion rate index represents the degree to which the system has reached the target state, with a value ranging from 0 to 1. The response completion rate at any given time can be obtained from the dynamic response characteristic curve. A preset completion rate threshold is set, typically 0.95, indicating that the response is considered essentially complete when the response completion rate reaches 95%. Based on the dynamic response characteristic curve, identify the moment when the response completion rate reaches the preset completion rate threshold, and determine this as the minimum response time. For example, for a load increase from 300MW to 400MW, if the response completion rate reaches 0.95 at the 18th minute, the minimum response time is 18 minutes.

[0061] Calculate the coefficient of variation of the dynamic response characteristic curve. This coefficient represents the degree of fluctuation of the curve and is calculated by dividing the standard deviation of the curve by the mean. For example, for the above response process, if the calculated coefficient of variation is 1.2, then the actual response time is the product of the minimum response time and the coefficient of variation, i.e., 18 × 1.2 = 21.6 minutes.

[0062] A time optimization objective function is constructed, comprising a time deviation term and an adjacent time interval term. The time deviation term represents the deviation between the actual execution time and the theoretical optimal time, while the adjacent time interval term represents the time interval between adjacent control commands. A smaller time deviation results in more precise control; an appropriate time interval avoids excessively frequent or sparse control commands. For the safety interlocking system of thermal power units, the weight of the time deviation term is typically set to 0.7, and the weight of the adjacent time interval term is set to 0.3, to balance control precision and system stability.

[0063] The actual response time is input into the state transition equation, and the time optimization objective function is solved using a dynamic programming algorithm. The state transition equation describes the change of the system state over time, with the actual response time serving as the key parameter for state transition. The dynamic programming algorithm solves the optimal subproblems step by step, ultimately obtaining the global optimal solution. For example, for an interlocking sequence containing 5 control commands, the optimal execution time intervals calculated by the dynamic programming algorithm are 2.5 seconds, 3.0 seconds, 4.2 seconds, and 3.8 seconds, respectively.

[0064] By combining the control cycle adjustment parameter with the execution time interval, the execution timing of each control command can be optimized. The control cycle adjustment parameter affects the execution effect of the control command, while the execution time interval determines the timing of the control command issuance. Combining the two can achieve the optimal control effect. For example, with an adjusted proportional gain coefficient of 2.0 and a derivative gain coefficient of 0.625, combined with the above execution time interval, the execution times of the five control commands can be determined to be 0 seconds, 2.5 seconds, 5.5 seconds, 9.7 seconds, and 13.5 seconds, respectively.

[0065] The interlocking protection command execution sequence is generated based on the optimized execution time. This sequence includes the content, execution time, and execution conditions of each control command. For example, for turbine overspeed protection, the possible execution sequence is: detect the speed at 0 seconds and determine if it exceeds the limit; close the main steam valve at 2.5 seconds; close the regulating valve at 5.5 seconds; close the extraction valve at 9.7 seconds; and trigger generator protection at 13.5 seconds. This execution sequence ensures timely response of safety interlocks while avoiding system oscillations caused by overly aggressive operations.

[0066] like Figure 2 The diagram shown illustrates the timing optimization flowchart for interlocking protection commands.

[0067] In one optional implementation, based on the dynamic response data during the operation of the thermal power unit, the corresponding slope is determined, and the fitting adaptive factor is calculated, including: The dynamic response data is segmented according to the set time window interval, and the rate of change of adjacent data points is calculated to obtain the response change rate value within the time window. The response rate of change is decomposed into multiple scales using wavelet transform, and components of different frequency bands are extracted. Rate values ​​that exceed the preset fluctuation range are removed to obtain a corrected response rate of change sequence. Calculate the mean and standard deviation of the corrected response rate of change sequence within each time window to determine the fluctuation range of the response rate of change; The response acceleration segment and the response deceleration segment are divided according to the upper and lower limits of the fluctuation range, and the average rate of change of the response acceleration segment and the response deceleration segment are calculated as the dynamic response slope. The dynamic response slope is optimized using a recursive iterative method. When the difference between two adjacent dynamic response slopes is less than a preset convergence threshold, the optimization result of the current iteration is used as the fitting adaptive factor.

[0068] In one specific implementation, dynamic response data is segmented according to a set time window interval. Parameter data collected during the operation of thermal power units, such as boiler steam pressure, turbine speed, and generator power, are divided according to fixed time windows. The size of the time window is determined based on the unit capacity and response characteristics; for a 600MW unit, it can be set to 30 seconds. For example, during an emergency load reduction process in a thermal power unit, if the generator power decreases from 500MW to 300MW, and a total of 200 data points are collected over a time span of 100 minutes, the data can be divided into 200 segments using 30-second time windows. For the data within each time window, the rate of change between adjacent data points is calculated, i.e., the difference in parameter values ​​divided by the sampling time interval. For example, if the power changes from 498MW to 495MW with a sampling interval of 15 seconds, the rate of change is -0.2MW / second. In this way, a sequence of response change rate values ​​within the time window is obtained.

[0069] Wavelet transform is used to perform multi-scale decomposition of the response rate of change. Wavelet transform can decompose a signal into components of different frequency bands, making it suitable for processing non-stationary signals. The obtained rate of change sequence is decomposed into three levels using the Daubechies wavelet basis function, yielding high-frequency detail components and low-frequency approximate components. High-frequency components typically contain noise and fluctuation information, while low-frequency components reflect the main trend of the signal. For example, after wavelet decomposition of the power rate of change, an abnormal fluctuation value of -0.6 MW / s may be found during the rapid load decline phase, far exceeding the normal range. By setting a preset fluctuation range, such as the mean of the rate of change plus or minus two standard deviations, rate values ​​exceeding this range are identified as outliers and removed. Assuming the normal power rate of change range is -0.3 MW / s to -0.1 MW / s, the data point with a value of -0.6 MW / s will be removed, resulting in a corrected response rate of change sequence.

[0070] Calculate the mean and standard deviation of the corrected response rate of change sequence within each time window. For the rate sequence after outlier removal, calculate the mean and standard deviation of the rate within each window, according to the original time window division. For example, in the first 30-second window, the corrected power change rate has a mean of -0.15 MW / s and a standard deviation of 0.03 MW / s; in the second window, the mean is -0.18 MW / s and the standard deviation is 0.04 MW / s. Based on the mean and standard deviation, determine the fluctuation range of the response rate of change, with the upper limit being the mean plus the standard deviation and the lower limit being the mean minus the standard deviation. For the first window, the fluctuation range is -0.18 MW / s to -0.12 MW / s; for the second window, it is -0.22 MW / s to -0.14 MW / s.

[0071] The response is divided into acceleration and deceleration phases based on the upper and lower limits of the fluctuation range. The acceleration or deceleration state of the response process is determined by comparing the fluctuation ranges of adjacent time windows. When the absolute value of the rate of change in the later window is greater than that in the earlier window, it is considered an acceleration phase; otherwise, it is a deceleration phase. For the load reduction process, if the power change rate in the second window is -0.18 MW / s, which is greater than that in the first window (-0.15 MW / s), then the process from the first window to the second window is an acceleration phase. The average rate of change within all windows of both the acceleration and deceleration phases is taken as the dynamic response slope for each stage. Assuming that three acceleration phases and two deceleration phases are identified throughout the entire load reduction process, with an average rate of change of -0.22 MW / s for the acceleration phases and -0.09 MW / s for the deceleration phases, these two values ​​are the dynamic response slopes for their respective stages.

[0072] A recursive iterative method is used to optimize the dynamic response slope calculation. Initially, the calculated slopes for the acceleration and deceleration segments are used as the starting point for iteration. In each iteration, the acceleration and deceleration states of each time window are reassessed based on the current slope values, and the slope calculations for the acceleration and deceleration segments are updated. For example, in the first iteration, using an acceleration segment slope of -0.22 MW / s and a deceleration segment slope of -0.09 MW / s, the state of each window is reassessed. It might be found that a window originally classified as an acceleration segment has an actual rate of change of -0.16 MW / s, which is less than the average rate of the acceleration segment and closer to the characteristics of the deceleration segment; therefore, it is reclassified as a deceleration segment. After reclassification, the new average rates of change for the acceleration and deceleration segments are calculated, such as -0.24 MW / s for the acceleration segment and -0.11 MW / s for the deceleration segment. The difference between the results of two adjacent iterations is compared, and the absolute value of the slope difference for the acceleration segment is calculated to be 0.02, and for the deceleration segment, it is 0.02. A preset convergence threshold of 0.01 is set. Since the current difference is greater than the threshold, iteration continues. After the fourth iteration, the slope of the acceleration segment is -0.255 MW / s, and the slope of the deceleration segment is -0.105 MW / s. Compared with the result of the third iteration, the differences are 0.005 and 0.005 respectively, both less than the preset convergence threshold of 0.01, and the iteration terminates. The slopes of the acceleration and deceleration segments obtained from the final iteration are combined to calculate the fitting adaptive factor. For example, if the acceleration segment is weighted at 0.7 and the deceleration segment at 0.3, the fitting adaptive factor is -0.255 × 0.7 - 0.105 × 0.3 = -0.2085.

[0073] In existing technologies, thermal power unit safety interlocking systems typically employ fixed threshold triggering and preset timing execution, which is insufficient to adapt to the dynamic characteristics of the unit under different operating conditions. Traditional methods rely heavily on empirical parameter setting, resulting in low calculation accuracy and weak handling of data fluctuations and outliers. This leads to suboptimal timing of interlocking protection actions, sometimes triggering prematurely and causing unnecessary shutdowns, and sometimes reacting too slowly to prevent the escalation of accidents. This embodiment introduces wavelet transform to perform multi-scale analysis of the response change rate, effectively identifying and eliminating abnormal data, thus improving the accuracy of dynamic response characteristic extraction. By distinguishing between the acceleration and deceleration phases of the response and calculating the slope separately, and then combining this with a recursive iterative method to dynamically optimize the calculation results, the fitting adaptive factor better matches the actual operating characteristics of the unit. This significantly improves the reliability and stability of the thermal power unit safety interlocking system, minimizing unplanned shutdowns while ensuring safety and improving the unit's economic efficiency.

[0074] The safety interlocking system for thermal power units based on protection logic according to embodiments of the present invention includes: The first unit is used to collect operating parameters of thermal power units; The second unit is used to analyze the time-series variation of operating parameters of thermal power units through a coupled evaluation model, extract the interactive influence characteristics of operating parameters, construct the protection logic association matrix, and calculate the protection logic association strength value. The third unit is used to convert the protection logic association strength value into a distance relationship to construct an association distance matrix, perform dynamic classification through adaptive clustering radius, perform iterative calculations based on inter-class merging metric, and generate a multi-level interlocking protection structure. The fourth unit is used to trigger ripple effect analysis by utilizing the real-time deviation value of the operating parameters of thermal power units, to track the transmission path of changes in operating parameters in the multi-level interlocking protection structure, and to identify related influencing nodes. The fifth unit is used to map key influencing nodes to the risk propagation network, dynamically adjust the strength of fuzzy rules based on the propagation coupling coefficient and establish historical adjustment trajectories, and perform fuzzy rule reasoning in combination with the historical adjustment trajectories to obtain risk assessment values. The sixth unit is used to select interlocking control strategies based on risk assessment values, determine the order of instruction execution according to the energy transfer chain of the thermal power unit's process, and form an interlocking protection instruction sequence. The seventh unit is used to adjust the control cycle parameters by controlling the adaptive factor, perform fitting analysis using the dynamic response characteristic curve, calculate the response time constraint, and determine the execution sequence of the interlocking protection command. The eighth unit is used to record the response characteristic curves of thermal power units during the execution of interlocking protection commands.

[0075] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0076] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0077] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for safety interlocking of thermal power generating units based on protection logic, characterized in that, The method comprises the following steps: Collecting operation parameters of a thermal power unit; Analyzing the time sequence variation law of the operation parameters of the thermal power unit through a coupling evaluation model, extracting interaction influence characteristics of the operation parameters, constructing a protection logic correlation matrix, and calculating a protection logic correlation strength value; Converting the protection logic correlation strength value into a distance relationship to construct a correlation distance matrix, dynamically classifying through an adaptive clustering radius, performing iterative operation based on an inter-class merging metric, and generating a multi-level interlock protection structure; Triggering a ripple effect analysis by using real-time deviation values of the operation parameters of the thermal power unit, tracking the change transmission path of the operation parameters in the multi-level interlock protection structure, and identifying an associated influence node; Mapping the key influence node to a risk propagation network, dynamically adjusting the strength of a fuzzy rule based on a propagation coupling coefficient and establishing a historical adjustment trajectory, and combining the historical adjustment trajectory to perform fuzzy rule reasoning to obtain a risk assessment value; Selecting an interlock control strategy according to the risk assessment value, determining an instruction execution order according to an energy transmission chain of a process of the thermal power unit, and forming an interlock protection instruction sequence; Adjusting control cycle parameters through a control adaptive factor, fitting analysis by using a dynamic response characteristic curve, calculating a response time constraint, and determining an interlock protection instruction execution time sequence; Recording the response characteristic curve of the thermal power unit in the interlock protection instruction execution process.

2. The method of claim 1, wherein, Converting the protection logic correlation strength value into a distance relationship to construct a correlation distance matrix, dynamically classifying through an adaptive clustering radius, performing iterative operation based on an inter-class merging metric, and generating a multi-level interlock protection structure comprises: Converting the protection logic correlation strength value into a distance relationship between protection objects and preprocessing, and calculating a correlation distance matrix of the protection objects; Based on the correlation distance matrix, the protection objects are dynamically classified by using an adaptive clustering radius adjustment method to obtain a protection object classification result; Based on the protection object classification result, the class tightness value and the class separation value are determined, the comprehensive evaluation value is combined, the merging metric between classes in the protection object classification result is calculated based on the comprehensive evaluation value, the class pair with the optimal merging metric is selected for iterative merging, the correlation distance matrix is updated, and the iteration is stopped when a preset maximum layer number is reached or the similarity between all classes in the correlation distance matrix is less than a preset similarity threshold, and an initial interlock protection structure is obtained; The connectivity of each layer of protection objects in the initial interlock protection structure and the transmission path length between layers are calculated, a protection object importance evaluation index is constructed based on the connectivity, a hierarchical transmission efficiency index is constructed based on the transmission path length, the protection object hierarchy is redistributed according to the protection object importance evaluation index and the hierarchical transmission efficiency index, and the correlation relationship between the protection objects is reconstructed, and an optimized multi-level interlock protection structure is generated.

3. The method of claim 2, wherein, The adaptive clustering radius adjustment method comprises: Based on the correlation distance matrix, an initial clustering radius is set, and the intra-class average distance and the inter-class minimum distance are calculated; An adaptive factor is determined according to the ratio of the intra-class average distance to the inter-class minimum distance; Adaptive factor is used to dynamically adjust clustering radius according to preset step size, and adaptive adjustment of clustering radius is completed when the ratio of the intra-class average distance to the inter-class minimum distance is located in a preset interval; During the adjustment process, the inter-class distance distribution characteristics after each adjustment are recorded to construct a historical adjustment trajectory, the next adjustment direction is predicted based on the historical adjustment trajectory, and the next adjustment direction is corrected in combination with the current inter-class boundary density.

4. The method of claim 1, wherein, Real-time deviation value of thermal power unit operating parameters is used to trigger wave and influence analysis, and the change transmission path of operating parameters is tracked in a multi-level interlocking protection structure to identify the associated impact nodes, including: Based on the operating parameters of the thermal power unit, the operating parameter fluctuation interval is established, the real-time deviation value of the operating parameter is calculated, and the wave and influence analysis is triggered when the real-time deviation value exceeds the preset deviation threshold; According to the real-time deviation value, the initial trigger parameter is located, the interlocking protection action signal corresponding to the initial trigger parameter is obtained in the multi-level interlocking protection structure, the parameter change transmission path is tracked based on the interlocking protection action signal, and the response time sequence of each parameter on the parameter change transmission path is recorded; The transmission delay between adjacent parameters on the parameter change transmission path is obtained, the influence attenuation ratio between adjacent parameters is calculated, and the influence attenuation ratio is combined with the transmission delay to determine the transmission path influence weight; According to the transmission path influence weight and the response time sequence, the influence intensity received by each node in the transmission path is calculated, and the node whose influence intensity exceeds the preset intensity threshold is determined as the key impact node.

5. The method of claim 1, wherein, Real-time deviation value of thermal power unit operating parameters is used to trigger wave and influence analysis, and the change transmission path of operating parameters is tracked in a multi-level interlocking protection structure to identify the associated impact nodes, including: Map the key impact nodes to the risk propagation network, and construct the propagation link between nodes; Based on the operating state parameters of the key impact nodes, the response time sequence rule in the preset historical fault case is extracted, and the propagation coupling coefficient between adjacent nodes on the propagation link is calculated based on the response time sequence rule; Based on the propagation coupling coefficient, the initial fuzzy rule strength is established, the rule strength difference degree between adjacent propagation links and the risk response ratio of the propagation link are calculated; According to the corresponding relationship between the rule strength difference degree and the risk response ratio, the fuzzy rule strength is dynamically adjusted, and the risk propagation characteristics in the adjustment process are recorded to form a historical adjustment trajectory; The operating state parameters are input into the adjusted propagation link, and the fuzzy rule reasoning is performed in combination with the historical adjustment trajectory to obtain the risk assessment value.

6. The method of claim 1, wherein, By controlling the adaptive factor adjustment control period parameter, using dynamic response characteristic curve for fitting analysis, calculating the response time constraint, and determining the interlocking protection instruction execution time sequence, including: The control adaptive factor is calculated by the ratio of the control error to the maximum allowed control error, the proportional gain coefficient and the differential gain coefficient of the control period are adjusted by using the control adaptive factor, and the control period adjustment parameter is obtained; Based on dynamic response data in the operation process of the thermal power generating unit, a corresponding slope is determined, a fitting adaptive factor is calculated, a Gompertz function is combined with the fitting adaptive factor for curve fitting, and a dynamic response characteristic curve is obtained; A response completion degree index corresponding to the dynamic response characteristic curve is calculated, a minimum response time is determined according to a time when the response completion degree index reaches a preset completion degree threshold, and an actual response time is obtained by multiplying the minimum response time and a variation coefficient of the dynamic response characteristic curve; A time optimization objective function including a time deviation term and an adjacent time interval term is constructed, the actual response time is input into a state transition equation, a dynamic programming algorithm is used to solve the time optimization objective function, and an execution time interval of each control instruction is obtained; The control cycle adjustment parameter and the execution time interval are combined, an execution time of each control instruction is optimized, and an interlocking protection instruction execution sequence is generated.

7. The method of claim 6, wherein, Based on dynamic response data in the operation process of the thermal power generating unit, a corresponding slope is determined, a fitting adaptive factor is calculated, and a Gompertz function is combined with the fitting adaptive factor for curve fitting, and a dynamic response characteristic curve is obtained, including: The dynamic response data is segmented according to a set time window interval, the change rate of adjacent data points is calculated, and a response change rate value in the time window is obtained; A multi-scale decomposition is performed on the response change rate value by using wavelet transform, components in different frequency bands are extracted, rate values exceeding a preset fluctuation range are removed, and a modified response change rate sequence is obtained; The mean and standard deviation of the modified response change rate sequence in each time window are calculated, and a fluctuation interval of the response change rate is determined; The upper limit and lower limit of the fluctuation interval are used to divide a response acceleration segment and a response deceleration segment, and the average change rate of the response acceleration segment and the response deceleration segment is calculated as the dynamic response slope; A recursive iteration method is used to optimize and calculate the dynamic response slope, and when the difference between the dynamic response slopes calculated in two adjacent times is less than a preset convergence threshold, the optimization result of the current iteration is taken as the fitting adaptive factor.

8. A protection logic based safety interlocking system for thermal power generating units for implementing the method according to any one of the preceding claims 1 to 7, characterized in that, It includes: A first unit for collecting thermal power generating unit operation parameters; A second unit for analyzing the time sequence variation law of the thermal power generating unit operation parameters by coupling an evaluation model, extracting the interactive influence characteristics of the operation parameters, constructing a protection logic correlation matrix, and calculating a protection logic correlation strength value; A third unit for converting the protection logic correlation strength value into a distance relationship to construct a correlation distance matrix, dynamically classifying by using an adaptive clustering radius, performing iterative operation based on an inter-class merging measure, and generating a multi-level interlocking protection structure; A fourth unit for triggering a ripple effect analysis using real-time deviation values of the thermal power generating unit operation parameters, tracking the operation parameter change transmission path in the multi-level interlocking protection structure, and identifying the correlation influence nodes; A fifth unit for mapping the key influence nodes to a risk propagation network, dynamically adjusting the fuzzy rule strength based on a propagation coupling coefficient and establishing a historical adjustment trajectory, and combining the historical adjustment trajectory to perform fuzzy rule reasoning to obtain a risk assessment value; The sixth unit is configured to select an interlocking control strategy according to the risk assessment value, determine an execution sequence of instructions according to an energy transmission chain of a process of the thermal power generating unit, and form an interlocking protection instruction sequence. The seventh unit is configured to adjust a control period parameter by using a control adaptive factor, perform fitting analysis by using a dynamic response characteristic curve, calculate a response time constraint, and determine an execution timing of the interlocking protection instruction. The eighth unit is configured to record a response characteristic curve of the thermal power generating unit in an execution process of the interlocking protection instruction.

9. An electronic device, comprising: The thermal power generating unit comprises: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon computer program instructions, wherein, The computer program instructions, when executed by the processor, implement the method in any one of claims 1 to 7.

Images