Virtual power plant operation optimization system

Abstract

The invention discloses a virtual power plant operation optimization system, which relates to the technical field of power plant operation optimization and comprises a virtual construction module, a power plant operation monitoring module and an optimization scheduling module. The virtual construction module carries out digital modeling on accessed distributed energy resources, and covers basic characteristics, operation constraints and scheduling rules of the distributed energy resources; the power plant operation monitoring module realizes dynamic monitoring and evaluation of an operation state through real-time data acquisition and analysis, and completes tracing analysis and positioning of an abnormal working condition based on model deviation and data comparison; and the optimal scheduling module is combined with a multi-dimensional scheduling target, a scheduling strategy is generated by adopting a self-adaptive optimization algorithm, and the operation efficiency and the system stability are improved. The system has good real-time performance, intelligence and adaptability, is suitable for a smart energy management application scene in which distributed energy sources are widely accessed, and can effectively improve the operation management level and the comprehensive scheduling capability of a virtual power plant.

Description

Technical Field

[0001] The present invention relates to the technical field of power plant operation optimization, and in particular to a virtual power plant operation optimization system. Background Art

[0002] Virtual power plant operation optimization systems are primarily based on a centralized energy management architecture, relying on fixed scheduling strategies and static mathematical models to coordinate and optimize the control of connected distributed energy resources. These systems typically integrate multiple energy units, including wind power, photovoltaics, energy storage, electric vehicle charging stations, and controllable loads. Through an energy management system (EMS) or distributed energy management platform (DEMS), these systems perform power forecasting, load regulation, and energy scheduling, optimizing overall power output and demand-side response.

[0003] However, the existing virtual power plant operation system lacks static scheduling strategies and cannot respond in real time to the sharp fluctuations in the output of distributed energy sources such as wind and solar power. This causes the system to have delayed scheduling responses when facing short-cycle load changes or uncertainties in renewable energy, thereby causing problems such as output deviation, grid shocks, and inefficient resource utilization. Summary of the Invention

[0004] In response to the problems mentioned in the above background technology, a virtual power plant operation optimization system is proposed.

[0005] The objectives of the present invention can be achieved through the following technical solutions: The present invention provides a virtual power plant operation optimization system, including: a power plant operation monitoring module and an optimization scheduling module; the power plant operation monitoring module obtains key parameter data in the actual power plant operation, maps the actual power plant data to the virtual power plant model input end in a one-to-one correspondence, and establishes a parameter synchronization channel; the virtual power plant model is driven to operate by the input data; the virtual operation results are then compared with the actual operation data in real time to identify deviations; when the virtual operation results exceed the corresponding set threshold, an abnormality tracing analysis is automatically triggered and the path credibility is output; the corresponding credibility of each causal path is compared, and the most credible abnormal causal path is output; The optimization scheduling module obtains and outputs the most credible abnormal causal path for optimization processing, including process optimization and software operation optimization; process optimization includes: constructing a directed graph model containing task nodes and dependency edges to generate a legal schedulable topology; constructing a weight model to identify bottleneck tasks; for bottleneck tasks, introducing a multi-layer perceptron classification model to output process path defects; then matching corresponding optimization strategies according to different process path defect types; software operation optimization includes: calculating the proportion of deviation from normal load based on the usage of CPU, memory and I / O resources, and obtaining the CPU offset rate, memory offset rate and I / O pending offset; calculating the task bottleneck contribution index based on the impact of the usage on the overall delay of the current task, selecting the bottleneck task set, entering the optimization strategy matching, obtaining the task feature vector, inputting the task feature vector into the strategy matching unit, and matching to obtain the optimal strategy; calculating the expected comprehensive optimization benefit value of the optimization strategy for the task; if the expected benefit value is greater than the preset benefit threshold, the strategy is considered effective; if it is not satisfied, the alternative strategy reselection mechanism is triggered and the strategy matching is re-entered.

[0006] Furthermore, it also includes a virtual construction module: the virtual construction module is used to build a virtual scene of a power plant, traverse and match all the equipment in the standard equipment model library to obtain the number of each equipment, and generate an equipment parameter library; if the corresponding equipment dimensions and operating parameters cannot be matched, customized modeling is performed, customized modeling: parse the spatial geometric parameters and equipment logical parameters according to the actual drawings or point cloud data; standardize the input information into the platform geometric semantic format; then establish a spatial coordinate system based on the standardized structural diagram and point cloud data, and arrange each module component in the virtual space according to the actual corresponding coordinates and dimensions; for the unique equipment structure in the power plant, use modeling tools to perform high-precision modeling to simulate the equipment shape, support structure, interface connection and operating parts, and then combine the material library and texture mapping to achieve one-to-one spatial restoration; then drive the model to change in real time by accessing the real-time data source.

[0007] Furthermore, the process of outputting path credibility is: Obtain the anomaly parameters, retrieve the causal dependency graph, and use the graph traversal algorithm to obtain all influencing node paths. Based on the influencing node paths, calculate the nodes on each path, and calculate the correlation between the path nodes and the anomaly nodes using mutual information. Then, compare the calculated correlation with the correlation threshold. Nodes with correlations greater than the correlation threshold are collected as possible cause nodes, and statistics are compiled to form a possible cause set. Then, we verify whether the possible causal node mutates before the abnormal point in time. Specifically, we set the causal parameter, determine the mutation point and change rate, and if the conditions are met, we determine that the parameter is a valid suspicious factor with temporal causal consistency. We match the screened and verified suspicious factor set with the pattern in the fault rule library. If a rule exists, we mark the path as a rule-consistent path. When several suspicious factors and causal paths are output, the credibility of each causal path is calculated, and the path credibility is output by combining correlation, time consistency and rule support.

[0008] Furthermore, the most credible abnormal causal path is output, including: The correlation calculation value, time consistency calculation value and rule matching calculation value are used as three-dimensional features, and the Softmax regression model is used for classification. The probability that each path belongs to the true causal relationship is output as the credibility. The credibility corresponding to each causal path is then compared, and the causal path with the highest credibility is output as the most credible abnormal causal path.

[0009] Furthermore, the specific process of building a weight model to identify bottleneck tasks is as follows: Build a directed graph model of process dependencies, then use the directed acyclic graph to identify cyclic dependencies or illegal paths in the process, generating a legal, schedulable process topology. Based on this legal, schedulable process topology, quantitatively analyze abnormal process paths to determine the critical timing deviations that affect efficiency and stability. If the critical timing deviation is greater than zero, the task is running slower than expected; if it is less than zero, execution is ahead of schedule. Then, the anomaly propagation path diagram is constructed to calculate the anomaly sensitivity of the task. Based on the obtained offset value and anomaly sensitivity, combined with the execution complexity and historical risk score, a task priority model is constructed to obtain the bottleneck task identification factor. The bottleneck task identification factor is compared with the preset threshold. If the bottleneck task identification coefficient is greater than the preset threshold, it is a bottleneck task. The bottleneck tasks obtained by the task weight model are sorted to obtain a set. Design a process reconstruction strategy for identified bottleneck tasks. Through the collaborative execution of the strategy, make adjustments and introduce a real-time feedback mechanism: continuously monitor the task execution time and newly appeared offsets. If significant bottlenecks still exist after the scheduling adjustment, trigger the next round of software operation optimization.

[0010] Furthermore, the task bottleneck contribution index is calculated, specifically: Through a multi-stage evaluation process, we identify performance bottlenecks in the current running tasks from three dimensions, including: The instantaneous offset rate and short-term variance index are calculated for key resources such as CPU, memory, and I / O. The usage of each resource is then classified into normal, fluctuating, and abnormal states based on the instantaneous offset rate and short-term variance index. Based on the output results, each resource is assigned a label of normal, fluctuating, or abnormal. Use sliding windows to analyze task resource change trends, detect continuous change trends of key resources and trend persistence ratios; output status classifications based on judgment rules, including rising trend, fluctuating trend, and stable trend; assign each task a label of stable trend, fluctuating trend, or rising trend; Build an execution dependency graph between tasks, analyze the role of the current task in the critical path, and output the results. Label the structure as the main bottleneck, secondary bottleneck, and non-critical path; Based on resource status, task allocation and structural results, the task bottleneck contribution index is obtained by rule combination.

[0011] Furthermore, the process of obtaining the optimal strategy is: For each task in the bottleneck task set, extract the multidimensional features of each task, integrate the multidimensional features into a task feature vector through the feature encoding module, and use it as the task portrait of the corresponding task; input the task feature vector into the strategy matching unit for matching; use the trained strategy selection agent to dynamically select the optimal strategy based on the historical optimization effect and task type; if there are multiple optimal strategies; then select the candidate strategy; after the optimization strategy matching deployment is completed, calculate the comprehensive optimization expected benefit value of the candidate strategy for the task, and select the maximum expected benefit value among the expected benefit values ​​as the optimal optimization strategy.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention breaks through the limitations of traditional virtual power plants that rely on static scheduling models, response lags, and fixed strategies by constructing a fusion mechanism of multi-source real-time perception and dynamic optimization models. It realizes continuous perception and dynamic modeling of resource status such as wind power, photovoltaics, and energy storage. It can automatically adjust the optimization strategy according to real-time data during operation, significantly improving the timeliness and accuracy of scheduling responses, effectively coping with fluctuations in new energy output, and improving the overall operating efficiency and stability of the system.

[0013] 2. The present invention introduces a multi-dimensional causal analysis and process scheduling optimization mechanism, uses causal dependency graphs and resource load analysis to accurately identify performance bottlenecks in paths, and dynamically optimizes task processes and scheduling strategies. At the same time, it also continuously adjusts the control logic based on feedback benefit indicators during operation to achieve personalized optimization of scheduling strategies and self-upgrade of system operation. This significantly improves the stability, adaptability, and overall operating efficiency of the power plant system in the face of complex operating environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0015] Figure 1 This is a schematic diagram of module connection of the present invention. DETAILED DESCRIPTION

[0016] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] It should be understood that the terms “include” and “comprising” used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0018] It should also be understood that the terminology used in this disclosure is for the purpose of describing specific embodiments only and is not intended to limit the disclosure. As used in this disclosure and the claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should be further understood that the term "and / or" as used in this disclosure and the claims refers to any and all possible combinations of one or more of the associated listed items, including and including these combinations.

[0019] See also Figure 1 As shown, a virtual power plant operation optimization system includes: a virtual construction module, a power plant operation monitoring module and an optimization scheduling module.

[0020] The virtual construction module is used to build a virtual scene of a power plant. The specific construction steps are as follows: Step 1: Collect power plant parameters on site and create information files, including: collecting power plant type (including thermal power, hydropower, wind power, etc.), land area, installed capacity, main engine model, number of units and transformer capacity, etc., as well as obtaining original data such as building structure drawings, equipment layout drawings, pipeline drawings, process flow diagrams, etc.

[0021] The power plant is divided into functional areas based on the original data, including the main plant area, boiler room, steam turbine room, power distribution room, cooling system, fuel processing area, office area, storage area, etc., and the purpose, equipment layout and entry and exit procedures of each functional area are determined and marked.

[0022] Step 2: Establish a standard equipment model library (including but not limited to transformers, steam turbines and cooling towers) based on the equipment layout drawing, match the specific models to obtain the corresponding equipment dimensions and operating parameters, then number the corresponding equipment dimensions and operating parameters, and traverse all the equipment in the standard equipment model library to match the numbers of each equipment; count all the equipment numbers and generate an equipment parameter library; if the corresponding equipment dimensions and operating parameters cannot be matched, it means there is no ready-made component, and custom modeling is performed; the custom modeling includes: parsing the spatial geometric parameters (size, coordinates and relative position, etc.) and the equipment logic according to the actual drawings / point cloud data The system can then edit parameters (such as connection sequence or control relationships), standardize input information into the platform's geometric semantic format, establish a spatial coordinate system based on the standardized structural diagram and point cloud data, and arrange each module component (such as boilers, turbines, transformers, cooling systems, etc.) in virtual space according to their actual relative coordinates and dimensions. It uses custom modeling tools to perform high-precision modeling of the unique equipment structures in power plants, simulate equipment appearance, support structures, interface connections, operating parts, etc., achieve one-to-one spatial restoration based on material libraries and texture mapping, and drive real-time model changes (including valve status, speed, temperature, etc.) by connecting to real-time data sources.

[0023] Step 3: After completing the space construction, alignment restoration and functional logic configuration, enter the 3D visualization modeling stage to enhance the realism and interactive experience of the virtual power plant; based on the texture information in the photos and video materials collected on the spot, restore the material mapping of the building and equipment surface. After the material mapping is completed, generate a realistic 3D model through the rendering engine; after completing the reproduction of the material and visual information, the 3D model has the spatial restoration and sensory realism, and can be further connected to the interactive logic and navigation system to realize the operational control of the virtual scene.

[0024] The power plant operation monitoring module obtains key parameter data in the actual power plant operation, such as equipment status, energy consumption indicators and environmental variables; maps the actual power plant data to the input end of the virtual power plant model in a one-to-one correspondence, and establishes a parameter synchronization channel; drives the operation of the virtual power plant model through the input data to generate a virtual operating state synchronized with reality; then compares the virtual operating results with the actual operating data in real time to identify deviations, anomalies or response differences; when the virtual operating results exceed the corresponding set threshold, it automatically triggers abnormality tracing analysis.

[0025] In a specific example: suppose the value of a parameter of the virtual power plant at time t is V v (t), the corresponding value of the actual power plant is V r (t), then the deviation is: ΔV(t)=|V v (t)-V r (t)∣, when ΔV(t)> , determined as an abnormal trigger point, automatically triggering abnormal retrospective analysis, where To set the threshold.

[0026] The specific process of the abnormal tracing analysis is as follows: Get the abnormal parameters, call the causal dependency graph G = (N, E), and use the graph traversal algorithm (such as breadth-first or depth-first) to obtain all the influencing node paths P = {n1, n2, ..., n k}, where each node n i is a parameter or control point that has a direct or indirect dependency relationship with the abnormal parameter; according to the influencing node path P, for each node n on the path i , calculate the correlation R(n i ), through the mutual information R(n i )=corr(V(n i ), ΔV) is calculated; then the correlation R(n i ) is compared with the correlation threshold τ, and the correlation R(n i ) are collected as possible inducement nodes, and statistics are used to form a possible inducement set.

[0027] Then verify whether the possible inducing node mutates before the abnormal point in time. Specifically, by setting the inducing parameter as x(t) and the mutation point as t x , if t x <t 异常 , and the rate of change satisfies: , then the parameter is considered to be an effective suspicious factor with consistent time causality, where t 异常 is the abnormality occurrence time, δ is the set threshold; then the set of suspicious factors that have been screened and verified is C={c1, c2, ..., c m}matches the pattern in the fault rule base, R j :{c1→Exception parameter};If there is a rule R j , then mark the path as a rule-consistent path, that is, it means that the suspicious factor c1 is supported by a known fault mechanism, and this path is more credible.

[0028] When several suspicious factors and causal paths are output, the credibility S is calculated for each causal path, combining the correlation, time consistency and rule support: R(n i )、T(n i ) and M(n i ) as a three-dimensional feature, using the Softmax regression model for classification, outputting the probability that each path belongs to the “real causal relationship” as the credibility S; where R(n i ) is the correlation calculation value, T(n i) is the time consistency calculation value, M(n i ) is the rule matching calculation value; then the credibility S corresponding to each causal path is compared, and the most credible abnormal causal path is output, that is, the causal path with the largest credibility S.

[0029] The optimization scheduling module optimizes the most credible abnormal causal path. The optimization process includes process optimization and software operation optimization. Specifically: The process optimization process is as follows: by constructing a process dependency directed graph model F=(T, D), where: T={T1, T2, ..., T p}, representing the independent and dependent task operation units in the process; D={D hg}, indicating the order constraints and resource dependencies between tasks (edges can be accompanied by constraints, such as mutual exclusion, shared resource occupancy, etc.); and each edge D hg Including: Type: sequential, parallel, conditional jump; Resource: corresponding resource occupancy type; Lock delay: expected delay time caused by lock contention; then use the directed acyclic graph (DAG) structure to model the dependency relationship between process tasks, use the loop detection algorithm to identify and eliminate circular dependencies and illegal paths in the process, and use the topological sorting algorithm to generate a legal and schedulable process topology.

[0030] Based on the generation of a legal schedulable process topology, a quantitative analysis of abnormal process paths is carried out to obtain the key timing deviations that affect efficiency and stability. The offset calculation process for each task is as follows: Output task T h Time offset value ;in, For task T h The actual execution completion time, For task T h The optimal completion time under ideal scheduling conditions (can also be understood as the expected benchmark time); if Δt h > 0, it means that the task is slower than expected; if Δt h <0, it means that the execution is ahead of schedule.

[0031] Construct an anomaly propagation path diagram and deduce the degree of anomaly impact propagation through the formula: Get the abnormal sensitivity of the task , where succ(h) is the task T h All direct and indirect downstream tasks (i.e. tasks that depend on it); For each successor task T h Time offset; For T h to T g The depth of a path in a dependency graph represents the distance it has traveled.

[0032] According to the offset value and abnormal sensitivity , and combined with execution complexity and historical risk scores, a task priority model is constructed, including: The execution time duration_h, dependency complexity degree_h, risk score risk_h (i.e. historical anomaly rate × current offset value) and anomaly sensitivity of each task Interval standardization was performed separately to obtain normalized indices Dh0, Gh0, Rh0 and Sh0.

[0033] Take the maximum value of each normalized indicator to construct the bottleneck task identification factor: Wh0=max{Dh0, Gh0, Rh0, Sh0}, where Dh0 represents the normalized value of task execution time, Gh0 represents the normalized value of dependency complexity, Rh0 represents the normalized value of historical risk score, and Sh0 represents the normalized value of abnormal sensitivity; Wh0 represents the bottleneck task identification factor.

[0034] Compare the bottleneck task identification factor with the preset threshold. If the bottleneck task identification factor is greater than the preset threshold, it is a bottleneck task. Sort the bottleneck tasks obtained by the task weight model to obtain the set {T b}, as the core object of scheduling optimization.

[0035] For the identified bottleneck tasks, a multi-layer perceptron classification model is introduced as a defect identification model to output process path defects; process path defects include: delay defects, conflict defects, blocking defects, fault-tolerant defects and serial inefficiency defects, etc.; then, corresponding optimization strategies are matched according to different process path defect types: for delay tasks: split tasks or preload resources to reduce initialization and calculation time; for conflict tasks: adjust task start sequence or remap resources to reduce resource crowding; for blocking tasks: optimize dependency path structure, introduce intermediate cache or predictive execution of predecessor tasks; for fault-tolerant tasks: strengthen input verification mechanism and fault rollback strategy; for serial inefficiency tasks: adjust scheduling logic, enable concurrent execution mechanism or automatic reordering of flowcharts.

[0036] During the optimization strategy execution process, a real-time feedback mechanism is introduced after adjustments are made: task execution time and newly occurring offsets are continuously monitored. If bottlenecks still exist after scheduling adjustments, the next round of software operation optimization is triggered. Software operation optimization: Based on the usage of resources such as CPU, memory, and I / O, calculate the deviation from the normal load ratio to obtain the CPU offset rate, memory offset rate, and I / O wait offset; based on the impact of the usage on the overall delay of the current task, calculate the task bottleneck contribution index B h , and is divided into three levels: high, medium and low.

[0037] In a preferred example, a multi-stage evaluation process is used to identify performance bottlenecks in the current task from three dimensions (including indicator deviation, trend evolution status, and path dependency), including: For key resources such as CPU, memory, and I / O, the instantaneous offset rate and short-term variance index are calculated. Based on these two indicators, resource usage is classified as normal, fluctuating, or abnormal. The status determination process is as follows: when the instantaneous offset rate is less than or equal to 1.2 times the original offset rate, and the variance index is less than or equal to 1.5 times the original variance index, the status is normal; when the instantaneous offset rate is between 1.2 and 1.5 times the original instantaneous offset rate, and the variance index is between 1.5 and 3 times the original variance index, the status is fluctuating; when the instantaneous offset rate is greater than 1.5 times the original instantaneous offset rate, and the variance index is greater than 3 times the original variance index, the status is abnormal. Based on the output results, each resource (such as CPU, memory, etc.) is assigned a "normal / fluctuating / abnormal" label.

[0038] Then, a sliding window is used to analyze the changing trends of task resources, detecting the continuous changing trends of key resources (whether resource usage has continued to rise, fall, or fluctuate repeatedly over the past period of time) and the trend persistence ratio (the proportion of time that a single trend (such as an increase) persists within the observation window). According to the judgment rules, the output status classification includes rising trend, fluctuating trend, and stable trend. Similarly, each task is assigned a label of "stable trend / fluctuating trend / rising trend"

[0039] The judgment rules are as follows: when the continuous change trend is rising for more than 70% of the time and the current trend persistence ratio is abnormal, it is an upward trend; when the continuous change trend is alternating between rising and falling, and the trend direction switches frequently, it is a fluctuating trend; when the continuous change trend is less than the preset change value within a fixed time and is in the normal or fluctuating range, it is a stable trend.

[0040] Build an execution dependency graph (task dependency directed graph) between tasks and analyze the role of the current task in the critical path: if the task is located in the most consuming period of the critical path and the resource status is "abnormal", it is marked as a "primary bottleneck point"; if there is an abnormal node in the dependent task and the current task has not yet recovered, it is marked as a "secondary bottleneck"; if the current task has no dependent bottleneck task, or is located on a non-critical path, it is a "non-bottleneck path"; similarly, based on the output results, the structure is judged and labeled as "primary bottleneck / secondary bottleneck / non-critical path".

[0041] Based on resource status, task allocation and structural results, determine the task bottleneck contribution index B according to the rule combination h; When the task tag has "item resource abnormality", "trend rising" and "main bottleneck", B h High; when the task tag has "item resource fluctuation", "trend fluctuation" and "secondary bottleneck or main bottleneck", B h When the task labels are "Normal Resources", "Stable Trend" and "Non-critical Path", B h is low.

[0042] According to the task bottleneck contribution index B h The sorting result; select the bottleneck task set T B , and enter the optimization strategy matching stage: specifically: For each task t in the bottleneck task set TB h , extract each task t h Multidimensional features, including but not limited to: 1: Task type (computation-intensive / I / O-intensive / network call, etc.), 2: Average execution time; 3: Historical anomaly rate; 4: Depth of the dependency path; 5: System resource usage (CPU, memory); 6: The waiting time of the previous and next tasks; 7: Historical optimization strategy response performance; integrate the multi-dimensional features into the task feature vector through the feature encoding module k={ 1, 2, ..., 7}, as the corresponding task t h mission portrait.

[0043] The task feature vector k is input to the strategy matching unit for matching; the trained strategy selection agent π( k)→Lp, dynamically select the optimal strategy Lp0 based on historical optimization results and task type; if there are multiple optimal strategies Lp0; then the candidate strategy Lp1.

[0044] An optimization strategy library L={L1, L2, ..., Lu} is established in the strategy matching unit. Each strategy Lp corresponds to a set of matching conditions Cp, scope of application and optimization objectives, such as scheduling priority adjustment (applicable to light blocking tasks), delayed execution merging (applicable to batch computing tasks), concurrent decomposition processing (applicable to highly coupled and time-consuming tasks), asynchronous processing substitution (applicable to waiting tasks), resource tilt optimization (applicable to resource bottleneck tasks) and exception isolation and rapid recovery (applicable to high error rate tasks), etc.

[0045] After the optimization strategy matching deployment is completed, for the candidate strategy Lp1, the delay optimization model, resource utilization optimization model and abnormal control optimization model are constructed respectively to independently evaluate the candidate strategy in task t h Each model outputs the corresponding optimization indicators, such as task execution time, CPU usage and exception rate. Based on the Pareto non-dominated sorting method in multi-objective optimization, these three indicators are used as different dimensions to construct a multi-dimensional optimization space. By screening out the Pareto optimal solution set, eliminating the non-optimal strategies that are superior to other strategies in all indicators, a set of candidate strategies that take into account delay, resource and exception control performance are output. As the task t h The final optimization strategy Lp0; Lp0 is applied to the bottleneck task t h And continuously collect its execution delay, resource consumption and abnormal rate changes through the real-time monitoring module; adopt the fuzzy comprehensive evaluation model, and design fuzzy membership functions for each indicator: task execution delay change, CPU resource consumption change, and abnormal rate change, respectively, map the indicator value to a fuzzy evaluation level (such as: excellent, good, medium, and poor), and calculate its corresponding fuzzy evaluation vector to form a fuzzy evaluation matrix RJ; operate the weight vector W and the fuzzy evaluation matrix RJ to obtain the comprehensive fuzzy evaluation result BJ of each strategy, defuzzify the comprehensive fuzzy evaluation result BJ (such as the centroid method) and convert it into a specific score value as a monitoring indicator ; If the monitoring indicators meet the profit threshold ≥θ1 (θ1 is the preset profit threshold), the strategy is considered effective; if not met, the alternative strategy reselection mechanism is triggered and the strategy matching is re-entered.

[0046] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A virtual power plant operation optimization system, comprising: Power plant operation monitoring module and optimization scheduling module; characterized by: the power plant operation monitoring module obtains key parameter data from actual power plant operation, maps the actual power plant data to the input end of the virtual power plant model in a one-to-one correspondence, and establishes a parameter synchronization channel; drives the virtual power plant model operation through the input data; then compares the virtual operation results with the actual operation data in real time to identify deviations; when the virtual operation results exceed the corresponding set threshold, it automatically triggers abnormal tracing analysis and outputs the path credibility; compares the corresponding credibility of each causal path, and outputs the most credible abnormal causal path; The optimization scheduling module obtains and outputs the most credible abnormal causal path for optimization processing, including process optimization and software operation optimization; process optimization includes: constructing a directed graph model containing task nodes and dependency edges to generate a legal schedulable topology; constructing a weight model to identify bottleneck tasks; for bottleneck tasks, introducing a multi-layer perceptron classification model to output process path defects; then matching corresponding optimization strategies according to different process path defect types; software operation optimization includes: calculating the proportion of deviation from normal load based on the usage of CPU, memory and I / O resources, and obtaining the CPU offset rate, memory offset rate and I / O pending offset; calculating the task bottleneck contribution index based on the impact of the usage on the overall delay of the current task, selecting the bottleneck task set, entering the optimization strategy matching, obtaining the task feature vector, inputting the task feature vector into the strategy matching unit, and matching to obtain the optimal strategy; calculating the expected comprehensive optimization benefit value of the optimization strategy for the task; if the expected benefit value is greater than the preset benefit threshold, the strategy is considered effective; if it is not satisfied, the alternative strategy reselection mechanism is triggered and the strategy matching is re-entered.

2. A virtual power plant operation optimization system according to claim 1, characterized in that: It also includes a virtual construction module: the virtual construction module is used to build a virtual scene of the power plant, traverse and match all the equipment in the standard equipment model library to obtain the number of each equipment, and generate an equipment parameter library; if the corresponding equipment dimensions and operating parameters cannot be matched, customized modeling is performed. Customized modeling: parse spatial geometric parameters and equipment logical parameters according to actual drawings or point cloud data; standardize input information into the platform geometric semantic format; then establish a spatial coordinate system based on the standardized structural diagram and point cloud data, and arrange each module component in the virtual space according to the actual corresponding coordinates and dimensions; For the unique equipment structures in power plants, high-precision modeling is performed using modeling tools to simulate the equipment shape, support structure, interface connections and operating components, and then combined with material libraries and texture mapping to achieve one-to-one spatial restoration; Then, by accessing real-time data sources, the model can be driven to change in real time.

3. A virtual power plant operation optimization system according to claim 1, characterized in that: The process of outputting path credibility is: Obtain the anomaly parameters, retrieve the causal dependency graph, and use the graph traversal algorithm to obtain all influencing node paths. Based on the influencing node paths, calculate the nodes on each path, and calculate the correlation between the path nodes and the anomaly nodes using mutual information. Then, compare the calculated correlation with the correlation threshold. Nodes with correlations greater than the correlation threshold are collected as possible cause nodes, and statistics are compiled to form a possible cause set. Then, we verify whether the possible causal node mutates before the abnormal point in time. Specifically, we set the causal parameter, determine the mutation point and change rate, and if the conditions are met, we determine that the parameter is a valid suspicious factor with temporal causal consistency. We match the screened and verified suspicious factor set with the pattern in the fault rule library. If a rule exists, we mark the path as a rule-consistent path. When several suspicious factors and causal paths are output, the credibility of each causal path is calculated, and the path credibility is output by combining correlation, time consistency and rule support.

4. A virtual power plant operation optimization system according to claim 3, characterized in that: Output the most credible abnormal causal path, including: The correlation calculation value, time consistency calculation value and rule matching calculation value are used as three-dimensional features, and the Softmax regression model is used for classification. The probability that each path belongs to the true causal relationship is output as the credibility. The credibility corresponding to each causal path is then compared, and the causal path with the highest credibility is output as the most credible abnormal causal path.

5. A virtual power plant operation optimization system according to claim 4, characterized in that: The specific process of building a weight model to identify bottleneck tasks is as follows: Build a directed graph model of process dependencies, then use the directed acyclic graph to identify cyclic dependencies or illegal paths in the process, generating a legal, schedulable process topology. Based on this legal, schedulable process topology, quantitatively analyze abnormal process paths to identify critical timing deviations that affect efficiency and stability. If the critical timing deviation is greater than zero, the task is slower than expected; If the critical timing deviation is less than zero, the execution is ahead of schedule; Then, the abnormal propagation path diagram is constructed to calculate the abnormal sensitivity of the task; Based on the obtained offset value and abnormal sensitivity, combined with the execution complexity and historical risk score, a task priority model is constructed to obtain the bottleneck task identification factor; the bottleneck task identification factor is compared with the preset threshold. If the bottleneck task identification coefficient is greater than the preset threshold, it is a bottleneck task; the bottleneck tasks obtained by the task weight model are sorted to obtain a set; Design process refactoring strategies for identified bottleneck tasks. Through collaborative execution of these strategies, adjustments are made and a real-time feedback mechanism is introduced to continuously monitor task execution times and emerging offsets. If significant bottlenecks still exist after scheduling adjustments, the next round of software operation optimization will be triggered.

6. A virtual power plant operation optimization system according to claim 5, characterized in that: Calculate the task bottleneck contribution index, specifically: Through a multi-stage evaluation process, we identify performance bottlenecks in the current running tasks from three dimensions, including: The instantaneous offset rate and short-term variance index are calculated for key resources such as CPU, memory, and I / O. The usage of each resource is then classified into normal, fluctuating, and abnormal states based on the instantaneous offset rate and short-term variance index. Based on the output results, each resource is assigned a label of normal, fluctuating, or abnormal. Use sliding windows to analyze task resource change trends, detect continuous change trends of key resources and trend persistence ratios; output status classifications based on judgment rules, including rising trend, fluctuating trend, and stable trend; assign each task a label of stable trend, fluctuating trend, or rising trend; Build an execution dependency graph between tasks, analyze the role of the current task in the critical path, and output the results. Label the structure as the main bottleneck, secondary bottleneck, and non-critical path; Based on resource status, task allocation and structural results, the task bottleneck contribution index is obtained by rule combination.

7. A virtual power plant operation optimization system according to claim 6, characterized in that: The process of obtaining the optimal strategy is: For each task in the bottleneck task set, extract the multidimensional features of each task, integrate the multidimensional features into a task feature vector through the feature encoding module, and use it as the task portrait of the corresponding task; input the task feature vector into the strategy matching unit for matching; use the trained strategy selection agent to dynamically select the optimal strategy based on the historical optimization effect and task type; if there are multiple optimal strategies; then select the candidate strategy; after the optimization strategy matching deployment is completed, calculate the comprehensive optimization expected benefit value of the candidate strategy for the task, and select the maximum expected benefit value among the expected benefit values ​​as the optimal optimization strategy.

Images