Power plant centralized control operation sequence generation method and device, electronic equipment and storage medium

Abstract

The application discloses a power plant centralized control operation sequence generation method and device, electronic equipment and storage medium, relates to the technical field of control systems, and comprises the following steps: constructing a single-device timing model representing the timing response characteristics of device operation based on power plant device basic performance data; constructing a hierarchical collaborative model system comprising a hierarchical device timing model and a centralized global collaborative large model in combination with industrial operation procedures and operation influence risk levels; generating a single-device operation instruction time sequence based on a target operation to be executed, determining a final operation execution path through integrated simulation deduction, cross-device conflict detection and iterative optimization, and generating and pushing a centralized control operation recommendation sequence after verification and review. The application breaks away from the dependence on artificial experience, improves the operation sequence working condition adaptability, avoids operation conflicts and operation risks, and greatly improves the efficiency and safety of power plant centralized control operation.

Description

Technical Field

[0001] This application relates to the field of control system technology, and in particular to a method, apparatus, electronic device and storage medium for generating power plant centralized control operation sequences. Background Technology

[0002] With the continuous advancement of intelligent power plant construction, centralized control operation, as the core link of power plant production control, places higher demands on the rationality, coordination, and safety of operation sequences. Many existing centralized control operation sequence schemes for power plants are driven by manual experience, employing fixed operating procedures to formulate operating strategies, which suffers from numerous technical defects: reliance on manual monitoring and verification of equipment parameters leads to significant delays in operation response, with actual operation time deviating greatly from theoretical values; static priority rules cannot adapt to the dynamic changes in equipment status and power plant operating conditions, resulting in delayed updates to operating procedures and potential misoperations; the lack of precise models characterizing the timing response characteristics of equipment operations allows for simple deductions of single-equipment operations, lacking cross-system and multi-equipment coordination mechanisms, easily leading to timing conflicts and resource competition issues; and the absence of supporting full-process operation simulation, cross-equipment conflict detection, and iterative optimization mechanisms results in insufficient operational risk control capabilities, severely impacting the efficiency and safety of centralized control operations. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and storage medium for generating operation sequences in power plant centralized control systems. It addresses the problems in related technologies caused by reliance on manual experience, lack of accurate equipment operation timing response modeling, absence of a cross-equipment global collaboration mechanism, and lack of a complete operation sequence deduction and optimization process and full-process risk control, resulting in delayed operation response, poor dynamic adaptability to operating conditions, frequent operation timing collaboration conflicts, high error rates, and insufficient operation efficiency and safety.

[0004] According to a first aspect of this application, a method for generating a power plant centralized control operation sequence is provided, comprising:

[0005] Based on the basic performance data of power plant equipment, a single-equipment time-series model is constructed to characterize the timing response characteristics of equipment operation. Based on the single-device timing model and combined with industrial operating procedures and operational impact risk levels, a hierarchical collaborative model system is constructed, which includes a hierarchical equipment timing model and a centralized control global collaborative model. Based on the target operation to be executed, the hierarchical collaborative model system is used to generate the operation instruction time series of each single device, and the operation instruction time series are integrated and simulated to obtain the operation simulation results. Based on the operation simulation results, cross-device conflict detection and iterative optimization are performed to determine the final operation execution path; After verifying and reviewing the final operation execution path, a recommended sequence of centralized control operations is generated and pushed out.

[0006] According to a second aspect of this application, a power plant centralized control operation sequence generation device is provided, comprising: The first construction module is configured to build a single-device time-series model characterizing the timing response characteristics of the equipment operation based on the basic performance data of the power plant equipment. The second construction module is configured to build a hierarchical collaborative model system that includes hierarchical equipment timing models and a centralized control global collaborative model, based on single-device timing models and combined with industrial operating procedures and operational impact risk levels. The simulation module is configured to generate the operation command time series of each single device based on the target operation to be executed, using a hierarchical collaborative model system, and to integrate and simulate the operation command time series to obtain the operation simulation results. The optimization module is configured to perform cross-device conflict detection and iterative optimization based on the operation simulation results to determine the final operation execution path; The push module is configured to generate and push a recommended sequence of centralized control operations after verifying and reviewing the final operation execution path.

[0007] According to a third aspect of this application, an electronic device is provided, comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which are executed by at least one processor to enable the at least one processor to perform the power plant centralized control operation sequence generation method of the first aspect.

[0008] According to a fourth aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the power plant centralized control operation sequence generation method of the first aspect described above.

[0009] According to a fifth aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the power plant centralized control operation sequence generation method as described in the first aspect above.

[0010] This application addresses the issue of single-equipment timing models characterizing equipment operation timing response based on power plant equipment fundamental performance data. It also establishes a hierarchical collaborative model system, incorporating hierarchical equipment timing models and a centralized control global collaborative model, by combining industrial operating procedures and operational impact risk levels. Based on the target operation to be executed, this hierarchical collaborative model system generates the operation instruction time series for each individual device, integrates and simulates the results to obtain operation simulation. Then, based on the operation simulation results, cross-equipment conflict detection and iterative optimization determine the final operation execution path. Finally, after verification and review, a recommended sequence for centralized control operation is generated and pushed. Therefore, this solution addresses the problems in related technologies where reliance on human experience and fixed operating procedures leads to significant delays in operation response and large discrepancies between actual and theoretical operation times. Static priority is thus achieved. The current system lacks a precise model to characterize the timing response of equipment operations, enabling only simple simulations of single-equipment operations. Furthermore, the absence of a cross-system, multi-equipment coordination mechanism leads to timing conflicts and resource competition. The lack of a comprehensive operational simulation, cross-equipment conflict detection, and iterative optimization mechanism further exacerbates operational risk control issues. The solution aims to eliminate reliance on manual experience, achieve precise timing modeling and global collaborative planning of power plant centralized control operation sequences, significantly improve the adaptability of operation sequences to dynamic changes in equipment status and operating conditions, effectively avoid timing conflicts and resource competition risks, improve full-process operational risk management, substantially reduce operational time and deviations from theoretical values, lower the error rate, and comprehensively enhance the efficiency and safety of power plant centralized control operations.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a method for generating a centralized control operation sequence for a power plant, provided in an embodiment of this application; Figure 2 A flowchart illustrating another method for generating a centralized control operation sequence for a power plant, provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a power plant centralized control operation sequence generation device provided in an embodiment of this application. Detailed Implementation

[0014] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0015] The following description, with reference to the accompanying drawings, describes a method, apparatus, electronic device, and storage medium for generating power plant centralized control operation sequences according to embodiments of this application.

[0016] The following will take the centralized control operation system of a 600MW supercritical coal-fired power generating unit as the application object. The unit has a rated power of 600MW and main steam parameters of 24.2MPa / 566℃ / 566℃. The implementation scenario is a complex operation sequence of load increase from 300MW to 550MW driven by grid AGC commands. It involves 14 types of core equipment in three major systems: boiler, turbine and auxiliary equipment. These are divided into 3 types of main control equipment, 7 types of subordinate equipment and 4 types of related equipment, covering 28 core operation items and 12 cross-equipment collaborative constraint rules. The implementation process of the power plant centralized control operation sequence generation method provided in this application will be explained in detail.

[0017] Figure 1 This is a flowchart illustrating a method for generating a centralized control operation sequence for a power plant, as provided in an embodiment of this application.

[0018] like Figure 1 As shown, the method includes the following steps: Step 101: Based on the basic performance data of power plant equipment, construct a single-equipment time-series model that characterizes the timing response characteristics of equipment operation.

[0019] In some embodiments, the basic performance data of power plant equipment includes multi-source heterogeneous data directly related to the equipment's operational response characteristics, such as equipment design ledger parameters, equipment design response delay data, equipment real-time health status data, historical operating condition response data, and equipment operation and maintenance log data. Taking a 600MW supercritical coal-fired generating unit as an example, this type of data includes equipment ledger parameters such as the rated output of the coal mill (45t / h) and the rated head of the feedwater pump (3200m), equipment response delay data such as the coal mill loading response delay (15~30s) and the wind turbine blade adjustment response delay (5~10s), and real-time health status data such as equipment bearing temperature, vibration amplitude, and health score.

[0020] After completing multi-source data acquisition, the data is first cleaned and standardized. Outlier data points are removed using the 3σ criterion, increasing the data effectiveness to 99.8%. Simultaneously, multi-system data is resampled and timestamp aligned with a 1-second acquisition cycle to eliminate data timing bias. Based on the preprocessed dataset, an attention-based LSTM timing modeling algorithm is employed. Using equipment design response latency, real-time health, and historical operating condition response data as core analysis dimensions, a timing response model for each type of core equipment is constructed, quantifying the causal latency relationship between equipment operation commands and changes in operating parameters. For edge computing nodes in the power plant, a lightweight single-equipment timing sub-model is simultaneously built, enabling 100ms-level rapid timing extrapolation at the edge. After model construction, convergence verification is performed to ensure a goodness-of-fit R² ≥ 0.98, meeting the accuracy requirements for on-site extrapolation.

[0021] The single-device timing model constructed in this step can accurately characterize the operation timing response characteristics of a single device, accurately quantify the time delay relationship between operation commands and device state changes, provide an accurate timing deduction basis for the generation of operation sequences, and effectively improve the adaptability of operation sequences to the actual response characteristics of the device.

[0022] Step 102: Based on the single-device timing model and combined with industrial operating procedures and operational impact risk levels, construct a hierarchical collaborative model system that includes a hierarchical equipment timing model and a centralized control global collaborative model.

[0023] In some embodiments, based on a single-device timing model that has completed convergence verification, and combined with current industrial operating procedures, cross-device collaboration rules, and equipment operation impact risk levels, an adaptive division of equipment classification and operation priorities is carried out to construct a hierarchical equipment timing model. Taking a 600MW supercritical coal-fired generating unit operating under a 300MW→550MW load increase condition as an example, the equipment PHM health assessment data is introduced to dynamically correct the equipment operation impact risk level matrix. Combined with the power grid AGC dispatch instructions, the equipment is divided into three levels: master control, subordinate, and associated, with corresponding operation priorities set from 1 to 3 from high to low. Among them, the risk level matrix weight of the first-level master control equipment is 1.0, the weight of the second-level subordinate equipment is 0.7, and the weight of the third-level associated equipment is 0.4. At the same time, combined with the real-time health of the equipment, the operation priority of the No. 3 coal mill with a health score of 85 is reduced by 0.5, and the operation timing buffer of the No. 2 induced draft fan with a health score of 88 is extended by 10 seconds, completing the dynamic adaptation of the hierarchical equipment timing model. Simultaneously, an operating condition adaptation submodule is built to adapt to the unit's variable load operation scenario.

[0024] After completing the construction of the time sequence models of each level of equipment, the time sequence models of all equipment at all levels are coupled and associated to build a large-scale centralized control global collaborative model. At the same time, a rule system consisting of an operation time sequence knowledge base, a cross-equipment collaborative rule base, and an operation risk base is built. A lightweight incremental update mechanism is configured for the rule system, which only performs local updates for the rules corresponding to new operating conditions and new equipment. In this load increase operating condition, only the adaptation rules are updated locally, which does not affect the basic operating efficiency of the system. Finally, a complete hierarchical collaborative model system is formed.

[0025] The hierarchical collaborative model system constructed in this step realizes dynamic adaptation of equipment operation priorities and global collaborative modeling across devices. It solves the defect that static rules cannot adapt to dynamic changes in working conditions and equipment status, and provides complete model support for global collaborative planning of operation sequences, effectively improving the compliance, collaboration and working condition adaptability of operation sequences.

[0026] Step 103: Based on the target operation to be executed, use the hierarchical collaborative model system to generate the operation instruction time series of each single device, and integrate and simulate the operation instruction time series to obtain the operation simulation results.

[0027] In some embodiments, based on the target operation to be executed, the semantic parsing and sub-operation decomposition of the target operation are first completed, and the overall target operation is decomposed into corresponding atomic operation items. A working condition similarity matching algorithm that fuses cosine similarity and weighted Mahalanobis distance is used to retrieve historical similar working conditions from the operation time sequence knowledge base of the hierarchical collaborative model system, and to match and locate the hierarchical equipment time sequence model corresponding to each atomic operation item. Taking the target operation of increasing load from 300MW to 550MW in a 600MW supercritical coal-fired power generating unit as an example, the overall load increase operation can be decomposed into 28 atomic operation items, quickly retrieving 12 sets of historical similar load increase working condition operation sequences, and completing the accurate matching of the corresponding hierarchical equipment time sequence model.

[0028] Based on the matched hierarchical equipment timing model, a time range sequence of operation instructions for each individual device is generated. Flexible buffer time intervals are set for the operation instructions based on the device health level. For devices with nonlinear response characteristics, such as coal mills, a Transformer + gated cyclic unit modeling method is used to generate nonlinear operation instruction time sequences. Simultaneously, hard time constraint intervals are set for the core operation instructions of critical equipment such as turbine control valves and boiler main control. In this load increase operation, the boiler main control fuel quantity instruction is increased in 5 stages, with a 60-second interval between each stage, and the single-stage adjustment delay must not exceed 5 seconds. The four coal mills are loaded in batches, with a 30-second interval between each loading instruction and a 15-second flexible buffer interval. The opening of the induced draft / forced draft fan blades is adjusted synchronously with the fuel quantity, with a 5-second buffer interval, strictly adhering to the hard time constraint rules.

[0029] After generating the time series of operation instructions for each individual device, the sequences are integrated with the rule base of the hierarchical collaborative model system to form a preliminary time coordination sequence. Based on the centralized control global collaborative model, and combined with the current unit operating conditions, multi-disturbance scenario simulation is carried out. Monte Carlo simulation with dynamic optimization sampling strategy is adopted, increasing the sampling number of high-risk conflict points to 2000 times, and using 500 basic sampling times for regular operation points, and finally generating complete operation simulation results.

[0030] This step can accurately generate operation instruction time sequences that adapt to equipment characteristics and operating conditions. Through global integration and multi-scenario simulation, the adaptability and feasibility of the operation sequences are fully verified, providing comprehensive simulation data support for subsequent optimization and effectively ensuring the matching degree between the operation sequences and the unit's operating objectives and equipment response characteristics.

[0031] Step 104: Based on the operation simulation results, perform cross-device conflict detection and iterative optimization to determine the final operation execution path.

[0032] In some embodiments, based on the full-process execution path corresponding to the operation simulation results, cross-device, multi-dimensional conflict detection is carried out. A layered parallel conflict detection mechanism is adopted, consisting of instruction layer, resource layer, logic layer, timing layer, and risk layer. A dynamic advance judgment mechanism is introduced simultaneously to detect and warn of delay conflicts in the entire target operation process in advance, and to comprehensively identify various conflict issues such as potential instruction mutual exclusion, resource competition, logical contradictions, timing constraint violations, and risk superposition. Taking the 300MW→550MW load increase operation of a 600MW supercritical coal-fired power generating unit as an example, full conflict detection can be completed for 28 atomic operation items within 100 minutes of the entire process. It can accurately identify multiple conflict items such as feedwater header resource competition, inverted timing of coal mill loading and primary air pressure adjustment, failure of induced draft fan adjustment interval to meet hard constraints, and risk superposition of coal mill synchronous loading, and classify the conflicts according to their severity.

[0033] Based on the identified conflict issues and in conjunction with industrial operating procedures and operational risk level requirements, iterative optimization was conducted within the timeframe of the initial time coordination sequence. A multi-objective optimization algorithm was employed to balance three core objectives: operation time, risk level, and resource utilization, to complete the initial adjustments. Independent time-series isolation windows were set for high-risk operations to achieve time-series decoupling. A deep Q-network reinforcement learning algorithm was then introduced for adaptive time-series adjustment. Intelligent iteration termination conditions were set with two indicators: risk entropy and operation time. Iteration stopped when the risk entropy dropped below a preset threshold of 0.05 and the operation time met the AGC command ramp rate requirement, thus correcting all time-series deviations and conflicts and ultimately determining the compliant and safe final operation execution path. After this load increase operation iteration, the risk entropy of the entire operation sequence decreased to 0.032, and the total operation time was 100 minutes, perfectly matching the AGC command's 2.5 MW / min ramp rate requirement.

[0034] This step, through layered parallel cross-device conflict detection and multi-dimensional iterative optimization, can comprehensively eliminate coordination conflicts and operational risks in the operation sequence, ensure that the final operation execution path is fully adapted to the equipment characteristics and operating conditions, effectively avoid the risk of parameter exceeding limits during operation execution, and significantly improve the execution safety and coordination of the operation sequence.

[0035] Step 105: After verifying and reviewing the final operation execution path, generate and push the recommended sequence of centralized control operations.

[0036] In some embodiments, a full-process verification and review of the final operation execution path is conducted, employing a two-tiered verification mechanism combining automatic verification and manual hierarchical review. First, a full-scale automatic verification is performed. Based on industrial operating procedures, cross-device collaborative rule bases, and relevant standards for power safety, compliance and safety verification is conducted on each operation item within the execution path. Each verification conclusion matches the corresponding standard procedure clause, generating a traceable verification and tracing report. Simultaneously, an automatic interception module for non-compliant operations is configured to directly intercept operation paths that violate safety red lines and output corresponding correction suggestions, eliminating potential risks of non-compliant and high-risk operations at the source.

[0037] After automatic verification passes, a manual tiered review process is executed, pushing only high-risk and complex combination operation items in the path to the central control operator for manual review. Taking the execution path of a 600MW supercritical coal-fired power generating unit's 300MW→550MW load increase operation as an example, automatic verification covers all 28 operation items. After verification passes, only the 6 high-risk operation items are subject to manual review, significantly reducing the workload of manual verification.

[0038] After full verification and review, a recommended sequence of centralized control operations adapted to the current power plant operating conditions is generated. The sequence clearly marks the execution timing, constraints, and risk levels of all operation items, with a total operation time of 100 minutes, perfectly matching the AGC command's ramp rate of 2.5 MW / min. The operation intervals across all equipment in the sequence meet the hard constraints, with a maximum timing deviation of ≤3 seconds and an operation risk entropy of 0.032, far below the preset safety threshold of 0.1. Finally, the recommended sequence of centralized control operations is pushed to the power plant's centralized control monitoring interface for the shift operator to confirm and execute.

[0039] This step employs a two-layer verification and review mechanism to fully ensure the compliance and security of the operation sequence, significantly reducing the workload of manual verification. The generated recommended sequence can be directly adapted to on-site working conditions for implementation, effectively reducing the risk of misoperation and improving the execution efficiency and safety management level of centralized control operations.

[0040] Compared with related technologies, in this embodiment, a single-equipment timing model characterizing the timing response characteristics of equipment operation is constructed based on the basic performance data of power plant equipment. Based on the single-equipment timing model and combined with industrial operating procedures and operational impact risk levels, a hierarchical collaborative model system is constructed, including a hierarchical equipment timing model and a centralized control global collaborative model. Based on the target operation to be executed, the hierarchical collaborative model system is used to generate the operation instruction time series of each single equipment, and the operation instruction time series are integrated and simulated to obtain operation simulation results. Based on the operation simulation results, cross-equipment conflict detection and iterative optimization are performed to determine the final operation execution path. After verifying and reviewing the final operation execution path, a recommended sequence for centralized control operation is generated and pushed. This technology can solve the problems in related technologies caused by reliance on human experience, lack of accurate equipment operation timing response modeling, absence of cross-equipment global collaboration mechanism, lack of complete operation sequence deduction optimization and full-process risk control process, resulting in operation response delays, poor dynamic adaptability to operating conditions, frequent operation timing collaboration conflicts, high error rates, and insufficient operation efficiency and safety. It achieves precise timing modeling, global collaborative planning, intelligent deduction optimization and safety control of power plant centralized control operation sequences, significantly improves the adaptability of operation sequences to operating conditions and equipment status, effectively avoids cross-equipment operation collaboration conflicts, greatly shortens operation time and deviation from theoretical values, reduces error and operational risks, and comprehensively improves the efficiency, safety and intelligence level of power plant centralized control operation.

[0041] Figure 2 A flowchart illustrating another method for generating a centralized control operation sequence for a power plant, provided in an embodiment of this application, includes the following steps: Step 201: Integrate multi-source heterogeneous performance data, including equipment load rate, response speed, and operation and maintenance logs, and use an attention-based LSTM time series modeling algorithm to construct a time series response model for equipment status. In addition, build a lightweight single-equipment time series sub-model for the edge computing nodes in the power plant to achieve rapid time series extrapolation at the edge.

[0042] In some embodiments, the process of constructing a device status time-series response model first involves the collection and preprocessing of multi-source heterogeneous performance data. The collected data covers multiple dimensions, including real-time load rate, command response speed, full-cycle operation and maintenance logs, equipment ledger parameters, real-time health status, and historical operating data. Taking a 600MW supercritical coal-fired generating unit as an example, this specifically includes the real-time load rate of the unit during load changes, the 15-30s loading response speed of the coal mill, the 5-10s adjustment response speed of the wind turbine blades, and all data such as historical operation and maintenance records, operating parameters, and health scores. The collected data is processed using the 3σ criterion to remove outlier data points, increasing the data effectiveness from 96.2% to 99.8%. Simultaneously, multi-source data is resampled and timestamp aligned with a 1s collection cycle to eliminate time-series deviations between different systems, thus completing the construction of a standardized dataset.

[0043] Based on a preprocessed standardized dataset, an attention-based LSTM temporal modeling algorithm is used to construct a device state temporal response model. In the algorithm, the attention mechanism focuses on key response periods after the device receives an operation command, strengthening the extraction weights of core response features. The LSTM network captures the parameter change dependencies of the device state over a long time series. Using device design response latency, real-time health, and historical operating condition response data as core analysis dimensions, the algorithm quantifies the causal latency relationship between device operation commands and changes in operating parameters, accurately characterizing the device's operational temporal response characteristics. After model construction, convergence verification is performed to ensure a goodness-of-fit R² ≥ 0.98, meeting the accuracy requirements for field simulations.

[0044] To address the computing power limitations of edge computing nodes in power plants, a lightweight pruning method is used to optimize the converged time-series response model. Redundant network layers and non-core feature parameters are removed, and a lightweight single-device time-series sub-model is built. This significantly reduces computing power consumption while ensuring the accuracy of core time-series inference, enabling fast time-series inference at the edge level of 100ms, and adapting to the real-time computing needs of the power plant edge.

[0045] This implementation method comprehensively covers the core factors affecting the timing response of equipment through multi-source heterogeneous data fusion. The LSTM algorithm based on the attention mechanism achieves accurate characterization of the timing response characteristics of the equipment. The lightweight sub-model takes into account the limitations of edge computing power and real-time requirements, providing an accurate and efficient model foundation for the timing deduction of operation sequences, and effectively improving the model's on-site adaptability and computational efficiency.

[0046] Step 202: Based on the single-device timing model, and combined with industrial operating procedures, equipment coordination rules, and operational impact risk levels, operational priorities are divided to construct a hierarchical equipment timing model.

[0047] In some embodiments, based on the single-device time-series model that has completed convergence verification, core feature data such as the time-series response characteristics, parameter stability duration, operational impact range, and operating condition adaptation boundary of each device are extracted. Combined with current industrial operating procedures, cross-device collaborative hard constraint rules, and equipment operation impact risk level classification standards, equipment hierarchy classification and operation priority setting are carried out. Taking the 300MW→550MW load increase operation of a 600MW supercritical coal-fired power generating unit as an example, the operation risk level classification standards are first clarified. The start-up and shutdown of the coal mill and significant adjustment of the turbine control valve are classified as high-risk operations; the adjustment of the wind turbine blades and the adjustment of the feedwater pump speed are classified as medium-risk operations; and the opening and closing of auxiliary valves are classified as low-risk operations. Combining the equipment operation impact range and risk level, the equipment is divided into three levels: master control, subordinate, and associated. Correspondingly, operation priorities are set from high to low at levels 1 to 3, where the risk level matrix weight of the first-level master control equipment is 1.0, the weight of the second-level subordinate equipment is 0.7, and the weight of the third-level associated equipment is 0.4. Match corresponding timing constraints, priority weights and compliance rules to each level of equipment, construct a hierarchical equipment timing model that corresponds one-to-one with the equipment level, and simultaneously build an operating condition adaptation submodule to adapt to the full-scenario operating condition requirements of unit variable load operation.

[0048] This step, through precise division of equipment hierarchy and operation priority, constructs a hierarchical equipment time sequence model that adapts to equipment characteristics and risk levels, achieving standardized management of equipment operation priorities. It solves the problem that static rules cannot adapt to differences in equipment characteristics and operational risks, and provides a hierarchical model foundation for subsequent global collaborative modeling.

[0049] Step 203: Couple the timing models of each hierarchical device to construct a centralized control global collaborative model, and build a three-element rule system including an operation timing knowledge base, a cross-device collaborative rule base, and an operation risk base.

[0050] In some embodiments, after completing the construction and accuracy verification of the timing models of each level of equipment, the timing models of all equipment at each level are coupled and associated, the data interaction channels between the models at each level are opened, the timing linkage logic, collaborative constraint relationship and parameter transmission rules between the master control, subordinate and associated equipment models are clarified, and a large-scale centralized control global collaborative model covering the entire system of boiler, steam turbine and auxiliary equipment is constructed to realize the global collaborative deduction of the operation timing of the entire unit equipment and the prediction of the overall effect. A three-element rule system was simultaneously built, consisting of an operation sequence knowledge base, a cross-equipment collaboration rule base, and an operation risk base. The operation sequence knowledge base stores standard operation sequences under historical operating conditions, measured data of equipment timing responses, and operating condition matching samples, and can include 12 sets of historical similar operating condition samples for load increases from 300MW to 550MW. The cross-equipment collaboration rule base stores equipment linkage rules and timing hard constraints that comply with industrial regulations, including core rules such as completing the forced draft fan linkage adjustment within 3 seconds after the induced draft fan adjustment and ensuring a matching deviation of fuel and feedwater change rates ≤5%. The operation risk base stores operation risk level classification standards, risk thresholds, and risk prevention and control rules. This three-element rule system is deeply integrated with the centralized control global collaborative model, providing comprehensive rule support for model deduction and calculation.

[0051] This step constructs a global collaborative model through the coupling of hierarchical models, and achieves deep integration of rules and models by combining it with a ternary rule system. It establishes a global collaborative mechanism across systems and multiple devices, providing complete model and rule support for the global planning and collaborative deduction of operation sequences.

[0052] Step 204 introduces equipment health assessment data to dynamically correct the risk level matrix of operational impact, combines the real-time load rate of the unit with the grid dispatch instructions to realize the hierarchical adaptive adjustment of equipment, and designs a lightweight incremental update mechanism for the three-element rule system that only performs partial updates for new operating conditions or new equipment.

[0053] In some embodiments, during the construction of the hierarchical equipment timing model and the operation of the ternary rule system, real-time access to equipment PHM health assessment data is performed. Based on the real-time equipment health score, the risk level matrix of equipment operation impact is dynamically corrected. For equipment with abnormal health, the operation priority and timing buffer are dynamically adjusted. For example, the operation priority of the No. 3 coal mill with a health score of 85 is reduced by 0.5, and the operation timing buffer of the No. 2 induced draft fan with a health score of 88 is extended by 10 seconds. Simultaneously, combined with the real-time load rate of the unit and the grid AGC / AVC scheduling instructions, adaptive adjustment of the equipment master, slave, and associated hierarchical levels is achieved to match the scheduling objectives and operational requirements of the current operating condition. For example, under the load increase condition of 300MW→550MW, combined with the AGC instruction's ramp rate requirement of 2.5MW / min, the operation priority and timing constraint boundaries of each level of equipment are dynamically adjusted. For the ternary rule system, a lightweight incremental update mechanism is specially designed. When a new operating condition, new equipment is connected, or the operating procedure is updated, only the rule entries corresponding to the new operating condition or new equipment are partially updated. The basic rules and existing valid data in the system are not changed. Only the adaptation rules for the load increase condition are partially updated, so as not to affect the basic operating efficiency of the system.

[0054] This step achieves dynamic adaptive adjustment of equipment levels through equipment health status and power grid dispatch instructions. Combined with a lightweight incremental update mechanism, it significantly improves the adaptability and update efficiency of the model and rule system, avoids the system operation burden caused by full updates, and ensures the stability and real-time performance of the system operation.

[0055] Step 205: A working condition similarity matching algorithm that combines cosine similarity and weighted Mahalanobis distance is used to retrieve historical similar working conditions, and complex combined target operations are decomposed into sub-operations and model matched to locate the corresponding hierarchical equipment time series model.

[0056] In some embodiments, for a target operation to be executed, the operation intent is first identified through semantic parsing. Complex combined target operations are then hierarchically decomposed into several atomic operation items, clarifying the execution equipment, operation objective, and constraints corresponding to each atomic operation item. Taking the 300MW→550MW load increase operation of a 600MW supercritical coal-fired power generating unit as an example, the overall load increase operation can be decomposed into 28 atomic operation items, covering 14 types of core equipment in the three major systems of boiler, turbine, and auxiliary equipment, clarifying the equipment level and risk level corresponding to each operation.

[0057] After decomposing the sub-operations, a condition similarity matching algorithm combining cosine similarity and weighted Mahalanobis distance was used to retrieve historical similar operating conditions from the operation time series knowledge base. Cosine similarity matched the overall trend of operating conditions with the characteristics of the operation target, while weighted Mahalanobis distance quantified the differences in parameter values ​​between the current and historical operating conditions. Differential weights were also assigned to core parameters such as unit load rate and equipment health to improve matching accuracy. Based on the retrieval results, a corresponding hierarchical equipment time series model was matched for each atomic operation item. This load increase operation quickly retrieved 12 sets of historical similar operating condition samples, achieving accurate model matching for 28 atomic operation items.

[0058] This step standardizes complex target operations by breaking them down into sub-operations. The dual-dimensional fusion algorithm significantly improves the accuracy and efficiency of working condition matching, enabling precise adaptation of target operations to the time series model of hierarchical equipment, thus laying a solid foundation for the generation of subsequent operation sequences.

[0059] Step 206: Generate a single-device operation instruction time range sequence based on the corresponding hierarchical device timing model. In this step, an elastic buffer timing interval is set according to the device health status. For nonlinear response devices, a modeling method combining Transformer and gated loop unit is used to generate a nonlinear operation instruction timing sequence. And a timing hard constraint interval is set for the core operation of key devices.

[0060] In some embodiments, based on the matched hierarchical device timing model, combined with device timing response characteristics, operation priority and operating condition constraints, a time range sequence of operation instructions for a corresponding single device is generated, and the execution node, adjustment step size and execution interval of the operation instructions are specified.

[0061] Based on real-time equipment health assessment data, flexible buffer time intervals are set for equipment with different health levels. Equipment with lower health levels has a longer buffer interval to allow time for stabilization during equipment status adjustments. Taking the 300MW→550MW load increase scenario as an example, a 15-second flexible buffer interval is set for the loading operation of the No. 3 coal mill with a health score of 85, and a 10-second buffer interval is extended for the adjustment operation of the No. 2 induced draft fan with a health score of 88. For nonlinear response equipment such as coal mills, a modeling method combining Transformer and gated loop units is used to capture the nonlinear temporal dependencies of the equipment and generate a suitable nonlinear operation command timing sequence. For core operations of critical equipment such as turbine control valves and boiler main control, hard time constraint intervals are set. In this load increase operation, a hard constraint interval of 0~100min is set for the boiler main control fuel quantity command, with a single-stage adjustment delay not exceeding 5 seconds. A hard constraint requirement is set for the turbine control valve opening to deviate from the feedwater flow rate by no more than 3 seconds, locking the timing boundaries of core operations.

[0062] The sequence generated in this step is fully adapted to the characteristics of the equipment and the requirements of the operating conditions. Through flexible buffers, nonlinear modeling and hard temporal constraints, the stability and safety of the operation sequence are guaranteed in all aspects, and the operational risks caused by equipment response mismatch are effectively avoided.

[0063] Step 207: Integrate the preliminary time coordination sequence of each device by combining the three-element rule system, and conduct simulation and deduction of multiple disturbance scenarios, including minor equipment failures and small load fluctuations, through the centralized control global collaborative model, to generate operation simulation results.

[0064] In some embodiments, based on the generated time range sequence of operation instructions for each individual device, and combined with the operation procedures, cross-device collaboration rules and risk control requirements in the ternary rule system, the sequence of each individual device is time-aligned, constraint-checked and logically integrated to eliminate basic timing deviations and form a unified preliminary time coordination sequence for the entire unit.

[0065] Based on a centralized control global collaborative model and combined with the real-time operating conditions of the units, simulations were conducted to simulate multiple disturbance scenarios. These scenarios covered common power plant disturbances such as minor equipment failures, small load fluctuations, and boiler combustion fluctuations, comprehensively verifying the execution stability and anti-interference capability of the preliminary time coordination sequence. Taking the 300MW→550MW load increase scenario as an example, the simulation covered typical scenarios such as minor coal mill loading failures and small fluctuations in AGC load commands. The simulation process employed Monte Carlo simulation with a dynamically optimized sampling strategy, increasing the sampling frequency for high-risk conflict points to 2000 times and using 500 basic sampling times for routine operation points, comprehensively covering potential risk scenarios. After the simulation was completed, data on equipment status changes, parameter fluctuations, and timing deviations throughout the entire process were summarized to generate complete operation simulation results, clarifying core indicators such as the total execution time of the operation sequence and the distribution of risk points.

[0066] This step achieves coordinated unification of single-device sequences through global integration, and fully verifies the feasibility of sequence execution through multi-perturbation simulation. It provides accurate and comprehensive simulation data support for subsequent sequence optimization, effectively ensuring the field adaptability and operational safety of the operation sequence.

[0067] Step 208 adopts a layered parallel conflict detection mechanism that includes instruction layer, resource layer, logic layer, timing layer and risk layer, and introduces a dynamic lead time determination mechanism to warn of delayed conflicts in order to identify potential logical contradictions and inter-device conflict issues.

[0068] In some embodiments, based on the full-process execution path corresponding to the operation simulation results, a layered parallel conflict detection mechanism consisting of an instruction layer, resource layer, logic layer, timing layer, and risk layer is adopted to conduct synchronous conflict detection for all operation items and the entire process. Specifically, the instruction layer focuses on detecting direct conflicts such as mutually exclusive instructions and unauthorized start / stop instructions in the operation sequence; the resource layer focuses on detecting competition conflicts when multiple devices share system resources, covering conflicts arising from the occupation of core public resources such as water supply mains and shared flue gas systems; the logic layer focuses on verifying whether the operation execution order conforms to the logical requirements of industrial operating procedures, and investigating issues such as timing reversal and logical contradictions; the timing layer focuses on verifying whether the operation timing exceeds the preset hard constraint range, and investigating issues such as substandard timing intervals and excessive timing deviations; the risk layer focuses on detecting the risk superposition problem caused by the simultaneous execution of multiple risky operations, and investigating potential risks exceeding risk levels.

[0069] Based on hierarchical parallel detection, a dynamic advance judgment mechanism is introduced. This mechanism uses the linkage logic between equipment timing response characteristics and operation sequences to detect and warn of delay conflicts throughout the entire target operation process, rather than simply performing static verification on the current execution node. Taking the 300MW→550MW load increase operation of a 600MW supercritical coal-fired power generating unit as an example, this mechanism can perform full detection of 28 atomic operation items within 100 minutes of the entire process. It accurately identifies four types of potential conflict problems: competition for feedwater header resources, inverted timing of pulverizer loading and primary air pressure adjustment, failure of induced draft fan adjustment intervals to meet hard constraints, and superimposed risks of synchronous pulverizer loading. Simultaneously, it completes the timing location and severity classification of conflict points.

[0070] This step utilizes a layered, parallel, multi-dimensional detection mechanism to achieve comprehensive screening of various operational conflicts. Combined with a dynamic lead time determination mechanism, it provides early warning of delayed conflicts, effectively avoiding the missed detection problems of traditional static detection and providing accurate positioning basis for the optimization and adjustment of subsequent operation sequences.

[0071] Step 209: Use a multi-objective optimization algorithm to adjust the instruction execution order and range, and set up an independent timing isolation window for high-risk operations to achieve timing decoupling.

[0072] In some embodiments, based on various potential problems identified through conflict detection, and combined with industrial operating procedures and equipment operation risk level requirements, a multi-objective optimization algorithm is used to adjust the execution order and scope of instructions. The algorithm uses three core optimization objectives—operation time, risk level, and resource utilization rate—to perform timing misalignment, step size adjustment, and sequence correction on the operation instructions corresponding to conflict points without breaking the hard constraints of equipment timing or changing the core operational objectives. This eliminates conflicts while simultaneously considering operational efficiency and unit operational stability.

[0073] For high-risk operations in the sequence, an independent time-series isolation window is set up to completely decouple high-risk operations from other medium- and low-risk operations in terms of timing. This avoids the risk superposition problem caused by the simultaneous execution of multiple risky operations, while reserving sufficient equipment stabilization time for high-risk operations. Taking the load increase operation of a 600MW supercritical coal-fired power generation unit from 300MW to 550MW as an example, through a multi-objective optimization algorithm, the speed adjustment sequence of the two steam-driven feedwater pumps is staggered by 15 seconds to eliminate resource competition conflicts caused by feedwater header pressure fluctuations; the loading command of the No. 3 coal mill is delayed by 20 seconds to correct the logical timing reversal problem; the adjustment interval between the induced draft fan and the forced draft fan is extended to 4 seconds to meet the hard constraints of the regulations; and the operation sequence of the two synchronously loaded coal mills is staggered by 30 seconds. With the independent time-series isolation window, the timing decoupling of medium-risk operations is achieved, ultimately forming a preliminary time coordination sequence without basic conflicts.

[0074] This step achieves precise correction of the conflict problem through a multi-objective optimization algorithm, taking into account the safety, efficiency and stability of the operation sequence. The setting of the time isolation window avoids the hidden danger of risk superposition from the root and greatly reduces the safety risks in the operation execution process.

[0075] Step 210 introduces a deep Q-network reinforcement learning algorithm to adaptively iteratively adjust the conflict points, and sets a dual-indicator intelligent termination condition of risk entropy and operation time to determine the final operation execution path.

[0076] In some embodiments, based on the initially adjusted time coordination sequence, a deep Q-network reinforcement learning algorithm is introduced to address the timing deviations and collaborative adaptation problems identified in the simulation. This algorithm combines historical conflict handling experience, equipment timing response characteristics, and operating condition adaptation data to adaptively iteratively adjust the remaining conflict points and timing deviation terms. The algorithm uses the lowest risk entropy and optimal operation time of the operation sequence as its reward mechanism, continuously optimizing the timing nodes and adjustment step size of the operation commands. This achieves precise correction of conflict points and collaborative optimization of the entire sequence, adapting to the optimization needs of different operating conditions without requiring manual pre-setting of adjustment rules.

[0077] A dual-indicator intelligent termination condition is set for risk entropy and operation time. Iteration automatically terminates only when the overall risk entropy of the sequence drops below a preset safety threshold, and the total operation time meets the requirements of the power grid dispatch instructions and the unit's operational goals. This avoids inadequate optimization due to insufficient iteration or efficiency loss caused by excessive iteration. Taking the 300MW→550MW load ramp-up operation of a 600MW supercritical coal-fired power unit as an example, a deep Q-network reinforcement learning algorithm is used to adaptively correct three issues: timing deviations in fuel quantity and feedwater flow, coordination deviations between the blower and furnace negative pressure, and timing deviations between the coal mill and primary air pressure. The iteration termination condition is set to a risk entropy below 0.05 and a total operation time matching the AGC instruction's ramp-up rate requirement of 2.5MW / min. After the final iteration, the overall sequence risk entropy drops to 0.032, and the total operation time is 100 minutes, fully meeting the preset requirements. This determines the final operation execution path.

[0078] This step utilizes a deep Q-network reinforcement learning algorithm to achieve adaptive intelligent optimization of the operation sequence. The dual-index termination condition balances the safety of the sequence with execution efficiency, significantly improving the accuracy and automation level of optimization adjustments, and ensuring that the final operation execution path is fully adapted to the on-site working conditions and equipment operating status.

[0079] Step 211: Establish a two-layer verification mechanism that combines automatic verification with manual graded review. Automatic verification covers all operation paths and generates verification traceability reports that match industry standards and power plant regulations. Manual review is only for high-risk operations or complex combination operations. At the same time, the automatic interception module for violations directly intercepts operation paths that touch the safety red line.

[0080] In some embodiments, a two-layer verification mechanism combining automatic verification and manual hierarchical review is established for the final determined operation execution path to achieve dual control over compliance and security throughout the entire operation process. The automatic verification process covers all operation items within the final operation execution path. Based on an operation risk database, a cross-device collaborative rule database, and industry standards such as the "Power Safety Work Regulations" and "Unit Operation Regulations," as well as internal power plant management regulations, it verifies the execution sequence, adjustment step size, collaborative constraints, and risk level of each operation instruction item by item. Each verification conclusion matches the corresponding standard procedure clause, generating a complete and traceable verification and tracing report. A synchronized automatic violation interception module is configured to identify violations and high-risk operations that cross safety red lines in the operation path in real time, directly intercepting non-compliant operation paths and simultaneously outputting corresponding correction suggestions to eliminate safety hazards in operation execution from the source. Taking the final execution path of the 300MW→550MW load increase operation of a 600MW supercritical coal-fired power generating unit as an example, the automatic verification covers all 28 atomic operation items, including 6 high-risk, 14 medium-risk, and 8 low-risk operations. The verification process fully matches all the requirements of the unit operation procedures, generates a complete verification traceability report, and simultaneously completes the safety red line verification, with no operation items touching the safety red line.

[0081] After automatic verification, a tiered manual review process is implemented, abandoning the traditional model of full manual verification. Only high-risk and complex combined operation items in the operation path are pushed to the power plant's centralized control operator for manual review, significantly reducing the workload of manual monitoring and verification while ensuring the absolute safety of core high-risk operations. In this load increase operation, only 6 high-risk operation items were manually reviewed, while the remaining 22 medium- and low-risk operation items did not require manual verification one by one. This significantly improved the overall efficiency of verification and review while ensuring operational safety.

[0082] This step employs a two-layer verification and review mechanism, which not only achieves automated compliance verification of all operation items, intercepting non-compliant operations at the source, but also ensures the safety of high-risk operations through tiered manual review, significantly reducing the workload of manual verification. At the same time, the verification and traceability report enables traceable control of the entire process, effectively reducing the risk of misoperation and comprehensively ensuring the compliance, safety, and on-site executability of the final operation execution path.

[0083] Step 212: After generating and pushing the recommended sequence of centralized control operations, synchronously collect execution feedback data and real-time data of unit operation status, iterate backward to the centralized control global collaborative big model and the three-element rule system, and establish a version management system that supports version rollback.

[0084] In some embodiments, after the recommended sequence of centralized control operations is generated, pushed, and executed on-site, the entire process execution feedback data of the operation sequence and real-time data of unit operating status are collected simultaneously to construct a closed-loop optimization mechanism for data feedback and to establish a version-based management system for the model and rule system. The execution feedback data collection phase covers all dimensions of effective data, including the actual execution delay of operation instructions, changes in equipment parameters, fluctuations in unit operating conditions, operation execution effects, and anomaly handling. After cleaning and standardizing the collected data, it is iterated back to the centralized control global collaborative model and the ternary rule system to achieve optimized updates of model parameters and iterative improvements to the rule system. Taking the 300MW→550MW load increase operation as an example, after the operation was completed, full data was collected within 100 minutes showing that the unit load steadily increased to 550MW, the main steam parameter fluctuation was ≤±0.5MPa / ±5℃, and the execution accuracy of 28 operation items was 100%. After reverse iteration of the effective data, the time sequence coordination rules were corrected, and hard constraint rules for the fuel quantity-feedwater flow time sequence deviation ≤3s under the load increase condition were added. The cross-equipment coordination rule library was updated. The parameters of the single equipment time sequence model were optimized, and the actual response delay data of the coal mill and flue gas system were added, increasing the model fit R² from 0.98 to 0.992. The operation risk level matrix was adjusted, and the risk weight of the coal mill loading operation under the load increase condition was corrected from 0.7 to 0.75, improving the priority of risk prevention and control. At the same time, the operation sequence, execution effect, and conflict handling plan were stored in the operation time sequence knowledge base, and 12 sets of similar operating condition matching samples were added to improve the speed and accuracy of subsequent operating condition matching.

[0085] For each iteration and optimization of the centralized control global collaborative model and the three-element rule system, a complete version management system is established. This system fully records and backs up each optimized model and rule base version, clearly indicating the content of the version update, the applicable operating conditions, and the optimization effect. It also supports one-click rollback of historical versions. When a new version encounters operational anomalies or compatibility issues, it can quickly roll back to a stable historical version, preventing system malfunctions from affecting the normal operation of the power plant's centralized control system.

[0086] This step constructs a complete closed-loop optimization mechanism by performing reverse iterations of execution data, realizing continuous self-optimization of the model and rule system. It solves the problem that historical operation data cannot effectively feed back into system optimization in traditional technologies, and significantly shortens the intelligent evolution cycle of the system. The version management system fully ensures the stability and reliability of system operation, avoids operational risks caused by iterative updates, and achieves an effective balance between system optimization and stable operation.

[0087] Figure 3 This is a schematic diagram of the structure of a power plant centralized control operation sequence generation device provided in an embodiment of this application, as shown below. Figure 3As shown, it includes: a first construction module 301, a second construction module 302, a deduction module 303, an optimization module 304, and a push module 305.

[0088] The first construction module 301 is configured to construct a single-device time-series model characterizing the timing response characteristics of the equipment operation based on the basic performance data of the power plant equipment. The second construction module 302 is configured to construct a hierarchical collaborative model system that includes a hierarchical equipment timing model and a centralized control global collaborative model, based on a single device timing model and combined with industrial operating procedures and operational impact risk levels. The simulation module 303 is configured to generate the operation instruction time series of each single device based on the target operation to be executed, using a hierarchical collaborative model system, and to integrate and simulate the operation instruction time series to obtain the operation simulation results. Optimization module 304 is configured to perform cross-device conflict detection and iterative optimization based on operation simulation results to determine the final operation execution path; The push module 305 is configured to generate and push a recommended sequence of centralized control operations after verifying and reviewing the final operation execution path.

[0089] In some examples of this embodiment, the first construction module 301 is specifically configured to integrate multi-source heterogeneous performance data, including equipment load rate, response speed, and operation and maintenance logs, and to construct a device state time-series response model using an LSTM time-series modeling algorithm based on an attention mechanism. It also builds a lightweight single-device time-series sub-model for the power plant's edge computing nodes to achieve rapid time-series extrapolation at the edge.

[0090] In some examples of this embodiment, the second construction module 302 is specifically configured to construct a hierarchical equipment timing model by dividing operation priorities based on a single device timing model, combined with industrial operating procedures, equipment coordination rules, and operation impact risk levels; couple each hierarchical equipment timing model to construct a centralized control global coordination model, and build a three-element rule system including an operation timing knowledge base, a cross-device coordination rule base, and an operation risk base; wherein, the operation impact risk level matrix is ​​dynamically corrected by introducing equipment health assessment data, and the equipment hierarchical adaptive adjustment is realized by combining the unit's real-time load rate and the grid dispatch instructions, and a lightweight incremental update mechanism is designed for the three-element rule system to only locally update new operating conditions or new equipment.

[0091] In some examples of this embodiment, the deduction module 303 is specifically configured to use a working condition similarity matching algorithm that combines cosine similarity and weighted Mahalanobis distance to retrieve historical similar working conditions, and to perform sub-operation decomposition and model matching for complex combined target operations to locate the corresponding hierarchical equipment time series model; generate a single-equipment operation instruction time range sequence based on the corresponding hierarchical equipment time series model, wherein an elastic buffer time series interval is set according to the equipment health, a nonlinear operation instruction time series sequence is generated for nonlinear response equipment using a modeling method combining Transformer and gated loop unit, and a time series hard constraint interval is set for the core operations of key equipment; integrate the preliminary time coordination sequence of each equipment with the ternary rule system, and perform multi-disturbance scenario simulation deduction covering minor equipment failures and small load fluctuations through a centralized global collaborative large model to generate operation simulation results.

[0092] In some examples of this embodiment, the optimization module 304 is specifically configured to employ a layered parallel conflict detection mechanism that includes an instruction layer, a resource layer, a logic layer, a timing layer, and a risk layer, and to introduce a dynamic lead time determination mechanism to warn of delayed conflicts in order to identify potential logical contradictions and inter-device conflict issues; to use a multi-objective optimization algorithm to adjust the instruction execution order and range, and to set independent timing isolation windows for high-risk operations to achieve timing decoupling; to introduce a deep Q-network reinforcement learning algorithm to adaptively iteratively adjust conflict points, and to set a dual-indicator intelligent termination iteration condition of risk entropy and operation time to determine the final operation execution path.

[0093] In some examples of this embodiment, the push module 305 is specifically configured to establish a two-layer verification mechanism that combines automatic verification with manual hierarchical review. The automatic verification covers all operation paths and generates a verification traceability report that matches industry standards and power plant regulations. The manual review is only for high-risk operations or complex combination operations. At the same time, the automatic interception module for violations directly intercepts operation paths that touch the safety red line. After generating and pushing the recommended sequence of centralized control operations, the execution feedback data and real-time data of unit operation status are collected synchronously. The data is then iterated backward to the centralized control global collaborative big model and the three-element rule system, and a version management system that supports version rollback is established.

[0094] It should be noted that other corresponding descriptions of the functional units involved in the power plant centralized control operation sequence generation device provided in this embodiment can be found in [reference]. Figure 1 , Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0095] Based on the above, Figure 1 , Figure 2The embodiment illustrates a method for generating a centralized control operation sequence for a power plant. Correspondingly, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 , Figure 2 This illustrates a method for generating operation sequences in a power plant centralized control system.

[0096] Based on the above, Figure 1 , Figure 2 The embodiment illustrates a method for generating a centralized control operation sequence for a power plant. Correspondingly, this embodiment also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described... Figure 1 , Figure 2 This illustrates a method for generating operation sequences in a power plant centralized control system.

[0097] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0098] Based on the above, Figure 1 , Figure 2 A method for generating a power plant centralized control operation sequence is shown, and Figure 3 To achieve the above objectives, the present application also provides an electronic device, such as a personal computer or a server, in the illustrated virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 , Figure 2 This illustrates a method for generating operation sequences in a power plant centralized control system.

[0099] In some embodiments, the aforementioned physical device may further include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit such as a keyboard, etc., and optionally, a USB interface, a card reader interface, etc. In some embodiments, the network interface may include a standard wired interface, a wireless interface (such as a Wi-Fi interface), etc.

[0100] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0101] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0102] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for generating a power plant centralized control operation sequence, characterized in that, include: Based on the basic performance data of power plant equipment, a single-equipment time-series model is constructed to characterize the timing response characteristics of equipment operation. Based on the single-device timing model and combined with industrial operating procedures and operational impact risk levels, a hierarchical collaborative model system is constructed, which includes a hierarchical equipment timing model and a centralized control global collaborative model. Based on the target operation to be executed, the hierarchical collaborative model system is used to generate the operation instruction time series of each single device, and the operation instruction time series are integrated and simulated to obtain the operation simulation results. Based on the operation simulation results, cross-device conflict detection and iterative optimization are performed to determine the final operation execution path; After verifying and reviewing the final operation execution path, a recommended sequence of centralized control operations is generated and pushed.

2. The method for generating a power plant centralized control operation sequence according to claim 1, characterized in that, The construction of a single-equipment time-series model characterizing the operational timing response characteristics of the equipment, based on the fundamental performance data of power plant equipment, includes: By integrating multi-source heterogeneous performance data, including equipment load rate, response speed, and operation and maintenance logs, an attention-based LSTM time series modeling algorithm is used to construct a time series response model for equipment status. A lightweight single-equipment time series sub-model is also built for the edge computing nodes in the power plant to achieve rapid time series extrapolation at the edge.

3. The method for generating a power plant centralized control operation sequence according to claim 1, characterized in that, The hierarchical collaborative model system, based on the single-device timing model and combined with industrial operating procedures and operational impact risk levels, is constructed, including a hierarchical equipment timing model and a centralized control global collaborative model. Based on the single-device timing model, combined with industrial operating procedures, equipment coordination rules, and operational impact risk levels, operational priorities are divided to construct a hierarchical equipment timing model. Couple the time-series models of each hierarchical device to construct a centralized global collaborative model, and build a three-element rule system including an operation time-series knowledge base, a cross-device collaborative rule base, and an operation risk base; Among them, the risk level matrix of operation impact is dynamically corrected by introducing equipment health assessment data, and the equipment is classified and adaptively adjusted by combining the real-time load rate of the unit and the grid dispatch instructions. A lightweight incremental update mechanism is designed for the three-element rule system to only locally update new operating conditions or new equipment.

4. The method for generating a power plant centralized control operation sequence according to claim 1, characterized in that, Based on the target operation to be executed, the hierarchical collaborative model system is used to generate the operation instruction time series of each individual device, and the operation instruction time series are integrated and simulated to obtain the operation simulation results, including: A working condition similarity matching algorithm that combines cosine similarity and weighted Mahalanobis distance is used to retrieve historical similar working conditions, and complex combined target operations are decomposed into sub-operations and model matched to locate the corresponding time series model of the hierarchical equipment. Based on the corresponding hierarchical equipment timing model, a single equipment operation instruction time range sequence is generated. Among them, an elastic buffer timing interval is set according to the equipment health. For nonlinear response equipment, a modeling method combining Transformer and gated loop unit is used to generate a nonlinear operation instruction timing sequence. And a timing hard constraint interval is set for the core operation of key equipment. By integrating the preliminary time coordination sequences of each device into the ternary rule system, and using the centralized control global collaborative model, simulations and deductions are performed on multiple disturbance scenarios, including minor equipment failures and small load fluctuations, to generate operational simulation results.

5. The method for generating a power plant centralized control operation sequence according to claim 1, characterized in that, The step of performing cross-device conflict detection and iterative optimization based on the operation simulation results to determine the final operation execution path includes: A layered parallel conflict detection mechanism is adopted, which includes instruction layer, resource layer, logic layer, timing layer and risk layer, and a dynamic advance judgment mechanism is introduced to warn of delayed conflicts in order to identify potential logical contradictions and inter-device conflict issues. A multi-objective optimization algorithm is used to adjust the instruction execution order and range, and an independent timing isolation window is set for high-risk operations to achieve timing decoupling. A deep Q-network reinforcement learning algorithm is introduced to adaptively and iteratively adjust the conflict points, and a dual-indicator intelligent termination condition of risk entropy and operation time is set to determine the final operation execution path.

6. The method for generating a power plant centralized control operation sequence according to claim 1, characterized in that, After verifying and reviewing the final operation execution path, a recommended sequence of centralized control operations is generated and pushed, including: Establish a two-tiered verification mechanism that combines automatic verification with tiered manual review. Automatic verification covers all operation paths and generates verification traceability reports that match industry standards and power plant regulations. Manual review is only for high-risk operations or complex combination operations. At the same time, an automatic interception module for violations directly intercepts operation paths that touch the safety red line. After generating and pushing the recommended sequence of centralized control operations, the system synchronously collects execution feedback data and real-time data on unit operating status, iterates backward to the centralized control global collaborative big model and the three-element rule system, and establishes a versioned management system that supports version rollback.

7. A power plant centralized control operation sequence generation device, characterized in that, include: The first construction module is configured to build a single-device time-series model characterizing the timing response characteristics of the equipment operation based on the basic performance data of the power plant equipment. The second construction module is configured to construct a hierarchical collaborative model system, which includes a hierarchical equipment timing model and a centralized control global collaborative model, based on the single device timing model and combined with industrial operating procedures and operational impact risk levels. The simulation module is configured to generate the operation instruction time series of each single device based on the target operation to be executed, using the hierarchical collaborative model system, and to integrate and simulate the operation instruction time series to obtain the operation simulation results. The optimization module is configured to perform cross-device conflict detection and iterative optimization based on the operation simulation results to determine the final operation execution path; The push module is configured to generate and push a recommended sequence of centralized control operations after verifying and reviewing the final operation execution path.

8. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the power plant centralized control operation sequence generation method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the power plant centralized control operation sequence generation method according to any one of claims 1-6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the power plant centralized control operation sequence generation method according to any one of claims 1-6.

Images