New energy station monitoring node collaborative deployment method and system, computer equipment and storage medium

Abstract

The invention belongs to the technical field of new energy power station monitoring, and particularly discloses a new energy station monitoring node collaborative deployment method and system, computer equipment and a storage medium. A multi-dimensional monitoring resource vector is constructed by collecting multi-source heterogeneous data such as electrical quantity, state quantity, environment quantity and resource state; feature density is calculated based on vector similarity, monitoring function partitions are automatically divided, and partition loads and fault risks are predicted; a multi-target optimization model is established, and dynamic optimal deployment of the monitoring task is realized; when the nodes are overloaded, a dynamic fragmentation mechanism is started, the tasks are split into sub-task groups and redeployed, and a scheduling system is formed. According to the method, the overall reliability and the resource utilization efficiency of the system are improved through dynamic partitioning, task optimization allocation and a real-time fragmentation mechanism.

Description

Technical Field

[0001] This application belongs to the field of monitoring technology for new energy power generation stations, specifically involving a collaborative deployment method, system, computer equipment, and storage medium for monitoring nodes of new energy power stations. Background Technology

[0002] With the rapid expansion of wind power, photovoltaic, and other new energy power plants, the number of distributed monitoring nodes is growing exponentially. Traditional monitoring systems typically employ a static deployment strategy, which has the following technical drawbacks:

[0003] 1) The monitoring data sources within the station are diverse, with heterogeneous data formats and varying sampling frequencies, making it difficult to effectively integrate and utilize them;

[0004] 2) Task allocation relies on manual rules or simple threshold judgments, which cannot dynamically adapt to real-time load changes, and may easily lead to some nodes being overloaded while other nodes are idle.

[0005] 3) Lack of comprehensive analysis of multi-dimensional data makes it difficult to identify potential risks in advance, resulting in delayed operation and maintenance response;

[0006] 4) When there is a sudden surge in tasks or node failures, the lack of a rapid response mechanism can easily lead to system-level failures.

[0007] For example, existing technologies often use fixed rules to allocate monitoring tasks to designated nodes, failing to dynamically adjust based on the site's operational status; or they rely solely on a single metric (such as CPU utilization) for load balancing, neglecting the correlation between multi-source data, leading to inaccurate predictions and decisions. Therefore, there is an urgent need for an intelligent monitoring method that can integrate multi-source heterogeneous data and achieve dynamic collaborative deployment. Summary of the Invention

[0008] This invention aims to solve the problems of difficult data integration, uneven load distribution, delayed fault warning and poor system stability in existing new energy power station monitoring systems. It provides a method, system, computer equipment and storage medium for collaborative deployment of monitoring nodes in new energy power stations. Through dynamic partitioning, task optimization allocation and real-time sharding mechanism, it improves the overall reliability and resource utilization efficiency of the system.

[0009] The specific technical solution of this application is as follows:

[0010] According to one aspect of this application, a method for collaborative deployment of monitoring nodes at new energy power stations is provided, including:

[0011] S100: Collects multi-source heterogeneous data from distributed monitoring nodes within new energy power stations, and constructs a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data;

[0012] S200: The monitoring resource vector calculates the monitoring feature density based on the similarity of the resource vectors of the monitoring nodes; then, based on the monitoring feature density, the entire site is automatically divided into multiple monitoring function zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted.

[0013] S300: Using the monitoring function partition and prediction results as a framework, the monitoring task allocation is modeled as an optimization model; the optimization model is solved to obtain the optimal deployment probability of each monitoring task, and the corresponding monitoring node is allocated to the monitoring task according to the optimal deployment probability;

[0014] S400: After task deployment, continuously monitor the resource usage of each monitoring function partition and monitoring node; when it is identified that the monitoring node is at risk of resource contention due to task overload, start the dynamic sharding mechanism to shard the monitoring tasks on the monitoring node with contention risk into multiple parallel processing sub-task groups, and redeploy the sub-tasks according to S300.

[0015] As a further option of the method in this application, the monitoring resource vector constructed in S100 is:

[0016] V i (t)=[v i1 (t),v i2 (t),...,v iD (t)];

[0017] Among them, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, V i (t) represents the monitoring resource vector;

[0018] V i (t) includes at least three of the following dimensions:

[0019] Electrical quantities include per-unit values ​​for voltage, current, and active power;

[0020] The status dimension includes switch status codes and protection action flags;

[0021] Environmental dimensions include standardized values ​​for temperature and wind speed;

[0022] The resource status dimension includes CPU utilization, memory usage, and communication quality metrics.

[0023] As a further option of the method in this application, the cosine similarity is used to calculate the resource vector similarity between monitoring nodes in step S200, and its expression is:

[0024]

[0025] Among them, S ij (t) represents the similarity between node i and node j at time t, S ij The value of V ranges from [1,1], and the closer the value is to 1, the more similar the states of nodes i and j are; i (t) represents the monitoring resource vector of the i-th node at time t, V j (t) represents the monitoring resource vector of the j-th node at time t, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, v jd (t) represents the standardized monitoring value of the j-th node in the d-th dimension.

[0026] As a further option of the method in this application, the monitoring feature density ρ in S200 i (t) is calculated using the kernel density estimation method, and its expression is:

[0027]

[0028] Where K(·) is the kernel function and h is the bandwidth parameter. Density ρ i (t) reflects the degree of clustering of the neighbors around node i at time t, and N is the total number of monitoring nodes; high-density areas correspond to clusters of nodes with highly similar states, which are used to automatically divide the monitoring function into partitions.

[0029] As a further option of the method in this application, the monitoring task in S300 is modeled as a tuple Task. k =(ID) k Type k DataSize k Deadline k Priority k ); where Task k For the k-th monitoring task, ID k For task identification, Type k For task type, DataSize k For data processing volume, Deadline k Priority is the deadline for the task. k This determines the task priority.

[0030] As a further option of the method in this application, the optimization model constructed in S300 is a multi-objective integer programming model, and its comprehensive objective function is:

[0031]

[0032] Among them, c ki Let L be the execution cost of task k on node i.i For the load of node i, L avg The load is the average load, and α is the weighting coefficient; the constraints include:

[0033] Each task must be assigned:

[0034] Node load does not exceed its capacity: w k Cap is the resource requirement weight for task k. i This represents the maximum resource capacity of node i;

[0035] The task must be completed before the deadline: c ki ≤Deadline k ,

[0036] As a further option of the method in this application, when the dynamic sharding mechanism in S400 detects the risk of resource contention among monitoring nodes, it divides the original monitoring task into multiple sub-task groups according to data dimensions or time windows. Each sub-task group inherits the priority and deadline constraints of the original task and is re-input into the optimization model of S300 as an independent task for redeployment.

[0037] As a further option of the method in this application, before S100, the original multi-source heterogeneous data is standardized, and principal component analysis or autoencoder is used to reduce the dimensionality of high-dimensional monitoring indicators before constructing the monitoring resource vector, retaining principal components with a cumulative contribution rate of not less than 90%, so as to improve the efficiency and robustness of subsequent similarity calculation and density estimation.

[0038] Another aspect of this application provides a collaborative deployment system for monitoring nodes at new energy power stations, the system comprising:

[0039] The multi-source data acquisition and vector construction module is used to acquire multi-source heterogeneous data from distributed monitoring nodes within the new energy power station, and to construct a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data.

[0040] The functional zoning and risk prediction module is used to calculate the resource vector similarity between nodes based on the monitoring resource vector of each monitoring node, and calculate the monitoring feature density of each node based on the similarity; further, based on the monitoring feature density, the entire site is automatically divided into multiple monitoring functional zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted.

[0041] The task optimization and deployment module is used to model the allocation of monitoring tasks into an optimization model based on the monitoring function partitions and prediction results; solve the optimization model to obtain the optimal deployment probability of each monitoring task; and allocate corresponding monitoring nodes to the monitoring tasks according to the optimal deployment probability.

[0042] The dynamic sharding and redeployment module is used to continuously monitor the resource usage of each monitoring function partition and monitoring node after task deployment. When it is identified that a monitoring node is at risk of resource contention due to task overload, the dynamic sharding mechanism is activated to shard the monitoring task on the monitoring node into multiple parallel processing sub-task groups, and the sub-task groups are input into the task optimization and deployment module to redeploy the task according to the optimization model.

[0043] According to one aspect of this application, a computer device is provided, the computer device comprising: a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by the processor to realize the collaborative deployment method of monitoring nodes of new energy power stations as described above.

[0044] According to one aspect of this application, a computer read storage medium is provided, wherein the storage medium stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, the at least one program, the code set, or instruction set is loaded and executed by a processor to realize the collaborative deployment method of monitoring nodes of new energy power stations as described above.

[0045] The beneficial effects of this application are as follows:

[0046] The collaborative deployment method for monitoring nodes in new energy power stations provided by this invention, by introducing multi-dimensional vector space theory, density clustering algorithm, and multi-objective optimization theory, achieves dynamic and accurate perception of the status of monitoring nodes and intelligent collaborative allocation of tasks. Its core advantage lies in completely overcoming the rigidity and slow response of traditional static deployment methods, significantly improving the resource utilization, task execution efficiency, and overall stability of the monitoring system. Specifically, it intelligently divides the dispersed monitoring nodes into functional zones based on state similarity and predicts their load and risk, thus providing a forward-looking decision-making basis for task allocation; by establishing an optimization model with the goal of minimizing completion time and balancing load, it achieves the globally optimal solution for task deployment; furthermore, through dynamic sharding and redeployment mechanisms, it endows the system with high elasticity and fault tolerance to cope with sudden situations such as node overload.

[0047] Through practical field verification, the method of this invention achieves significant technical effects, including an average reduction of 35% in task completion time, a 40% improvement in load balancing, and a 50% increase in system throughput. Especially in the highly dynamic environment of new energy power plants, this method can quickly respond to events such as sudden wind speed changes, completing task migration within seconds, effectively avoiding monitoring interruptions, and ensuring data integrity and the safe and stable operation of the power plant. In summary, this invention not only improves the automation level and operational efficiency of the monitoring system but also provides key technical support for the reliable and efficient monitoring of large-scale new energy power plants, possessing significant industrial application value. Attached Figure Description

[0048] Figure 1 A schematic diagram illustrating the overall process of collaborative deployment of monitoring nodes at new energy power plants;

[0049] Figure 2 Detailed flowchart of steps S100 for collaborative deployment of monitoring nodes at new energy power plants;

[0050] Figure 3 Detailed flowchart of steps S200 for collaborative deployment of monitoring nodes at new energy power plants;

[0051] Figure 4 Detailed flowchart of the S300 steps for the collaborative deployment of monitoring nodes at new energy power plants;

[0052] Figure 5 Detailed flowchart of the S400 steps for the collaborative deployment of monitoring nodes at new energy power plants. Detailed Implementation

[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0054] As a crucial component of the new power system, the operational stability of renewable energy power plants directly impacts the safety and efficiency of the power grid. These plants contain numerous distributed monitoring nodes responsible for collecting heterogeneous data from multiple sources, including electrical quantities, status parameters, environmental parameters, and equipment resource status. Traditional monitoring node deployment methods often employ static configurations, which cannot adapt to the dynamically changing operating environment of renewable energy power plants, leading to uneven distribution of monitoring tasks, low resource utilization, and delayed fault response. This invention is based on three pillars: multidimensional vector space theory, density clustering algorithms, and multi-objective optimization theory. It achieves a unified representation of node status by constructing monitoring resource vectors, realizes intelligent partitioning of the power plant using feature density analysis, and achieves dynamic collaborative deployment of monitoring tasks based on an optimization model.

[0055] The core theoretical derivation is as follows:

[0056] First, the state of the monitoring nodes is mathematically represented using a multi-dimensional monitoring resource vector. Let the number of nodes be N, and the state of each node at time t be represented by a D-dimensional vector V. i (t) represents:

[0057] V i (t)=[v i1 (t),v i2 (t),...,v iD (t)];

[0058] Among them, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, where the dimensions include electrical quantities, state quantities, environmental quantities, and resource states. The magnitude and direction of the vector reflect the load intensity and state characteristics of the node, respectively.

[0059] Secondly, the similarity between nodes is calculated using cosine similarity, with the following formula:

[0060]

[0061] Among them, S ij (t) represents the similarity between node i and node j at time t, S ij The value of V ranges from [1,1], and the closer the value is to 1, the more similar the states of nodes i and j are. i (t) represents the monitoring resource vector of the i-th node at time t, V j (t) represents the monitoring resource vector of the j-th node at time t, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, v jd (t) represents the standardized monitoring value of the j-th node in the d-th dimension. A similarity matrix S is constructed based on the similarity of all node pairs.

[0062] The monitoring feature density is calculated using the kernel density estimation method. For node i, its density estimate is:

[0063]

[0064] Where K(·) is the kernel function and h is the bandwidth parameter. Density ρ i (t) reflects the degree of clustering of the neighbors around node i at time t. High-density areas represent clusters of nodes with similar states.

[0065] Finally, the task allocation problem is modeled as a multi-objective optimization model, with objective functions including minimizing task completion time and maximizing load balancing. Let the task set be T, the node set be N, and the decision variable be x. ij This indicates whether task k is deployed on node i. The optimization objective is expressed as:

[0066]

[0067] Among them, c ki Let L be the execution cost of task k on node i. i For the load of node i, L avg Let α be the average load and α be the weighting coefficient. This model is solved using linear programming or a heuristic algorithm to obtain the optimal task deployment scheme.

[0068] The above theoretical framework provides a solid mathematical foundation for this invention, ensuring the rationality and effectiveness of the collaborative deployment of monitoring nodes. The specific implementation methods of this invention will be described in detail below.

[0069] Please see Figure 1 This illustrates a collaborative deployment method for monitoring nodes at new energy power stations according to an embodiment of the present invention, the method comprising:

[0070] S100: Collects multi-source heterogeneous data from distributed monitoring nodes within new energy power stations, and constructs a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data;

[0071] S200: The monitoring resource vector calculates the monitoring feature density based on the similarity of the resource vectors of the monitoring nodes; then, based on the monitoring feature density, the entire site is automatically divided into multiple monitoring function zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted.

[0072] S300: Using the monitoring function partition and prediction results as a framework, the monitoring task allocation is modeled as an optimization model; the optimization model is solved to obtain the optimal deployment probability of each monitoring task, and the corresponding monitoring node is allocated to the monitoring task according to the optimal deployment probability;

[0073] S400: After task deployment, continuously monitor the resource usage of each monitoring function partition and monitoring node; when it is identified that the monitoring node is at risk of resource contention due to task overload, start the dynamic sharding mechanism to shard the monitoring tasks on the monitoring node with contention risk into multiple parallel processing sub-task groups, and redeploy the sub-tasks according to S300.

[0074] The specific plan is as follows:

[0075] In a collaborative deployment method for monitoring nodes at new energy power plants, S100 achieves unified collection and vectorized representation of multi-source heterogeneous data within the plant, providing standardized data input for subsequent analysis. The monitoring resource vector integrates electrical, status, environmental, and resource data to form a multi-dimensional data structure of node time indicators, providing comprehensive and accurate data support for subsequent collaborative deployment.

[0076] Please refer to Figure 2 The diagram illustrates a flowchart of an exemplary collaborative deployment method S100 for monitoring nodes at a new energy power station, the contents of which include:

[0077] S110: Multi-source heterogeneous data acquisition.

[0078] In renewable energy power plants, distributed monitoring nodes are located throughout key equipment such as wind turbines, photovoltaic arrays, energy storage units, and substations. These nodes collect multi-source heterogeneous data through various sensors and smart terminals.

[0079] Specifically, data collection includes the following categories:

[0080] Electrical quantity data includes voltage, current, power, frequency, harmonic content, etc., reflecting the steady-state and transient characteristics of the power system. Data acquisition methods include smart meters, protection devices, power quality analyzers, etc., with sampling frequency adjusted according to requirements; for example, steady-state data is sampled once per second, and transient data is sampled at 128 points per cycle.

[0081] Status data includes circuit breaker position, disconnector status, protection action signals, and equipment operating modes, reflecting the power grid topology and equipment control status. Status variables are typically binary or enumerated values, collected through a communication management unit or remote terminal unit, and reported immediately upon change.

[0082] Environmental data includes ambient temperature, humidity, wind speed, wind direction, light intensity, and precipitation, which affect the efficiency of new energy power generation and equipment operating conditions. Environmental data are collected through weather stations or environmental sensors, with a sampling frequency typically once per minute.

[0083] Device hardware and software resource status data: This includes monitoring node CPU utilization, memory usage, disk space, network bandwidth, and communication error rate, reflecting the node's health and processing capacity. Resource status is collected through the node's built-in operating system interface or agent program, with a sampling frequency set to once per second.

[0084] In one alternative implementation, multi-source heterogeneous data acquisition employs an edge computing architecture, performing preliminary cleaning and caching near the data source, and uploading only characteristic or abnormal data to reduce the burden on the communication network. Data is transmitted to the central monitoring platform via industrial Ethernet or a private wireless network, using protocols including IEC61850 and Modbus TCP.

[0085] S120: Construct monitoring resource vectors.

[0086] To uniformly represent the multidimensional state of nodes, a monitoring resource vector is constructed for each monitoring node. Data preprocessing is required before vector construction, including data alignment, missing value imputation, outlier removal, and standardization.

[0087] Monitoring resource vector V i (t) is defined for node i at time t as:

[0088] V i (t)=[v i1 (t),v i2 (t),...,v iD (t)];

[0089] Among them, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, where D is the vector dimension and consists of all monitoring metrics.

[0090] For example, a vector includes the following dimensions:

[0091] Electrical quantity dimensions: per-unit voltage, per-unit current, per-unit active power, etc.;

[0092] Status quantity dimensions: switch status coding, protection action flags, etc.;

[0093] Environmental parameters: standardized values ​​of temperature, standardized values ​​of wind speed, etc.

[0094] Resource status dimensions: CPU utilization, memory usage, communication quality, etc.

[0095] In one feasible implementation, to further reduce the vector dimension, principal component analysis or an autoencoder is used for feature extraction, preserving features in the main variation directions and improving subsequent computational efficiency.

[0096] The completed monitoring resource vectors are stored in a time-series database, such as InfluxDB or TimescaleDB, for subsequent analysis.

[0097] In a collaborative deployment method for monitoring nodes at new energy power plants, S200 calculates the feature density between nodes based on monitoring resource vectors, and performs power plant partitioning and load prediction accordingly. This step clusters the dispersed node states into functional regions and predicts future states, providing spatial and temporal decision-making basis for task allocation.

[0098] Please refer to Figure 3 The diagram illustrates a flowchart of an exemplary collaborative deployment method S200 for monitoring nodes at a new energy power station, the contents of which include:

[0099] S210: Calculate resource vector similarity.

[0100] The similarity between nodes i and j at time t is calculated using cosine similarity, with the following formula:

[0101]

[0102] Among them, S ij (t) represents the similarity between node i and node j at time t, V i (t) represents the monitoring resource vector of the i-th node at time t, V j (t) represents the monitoring resource vector of the j-th node at time t, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, v jd (t) represents the standardized monitoring value of the j-th node in the d-th dimension. Cosine similarity measures the closeness of vector directions, S ij The value of (t) ranges from [1,1]. The closer the value is to 1, the more similar the states of the two nodes are; the closer the value is to 1, the opposite the states are; a value of 0 indicates that they are unrelated.

[0103] In one feasible implementation, to account for similarity over time, the average similarity within a time window is calculated:

[0104]

[0105] Where T is the length of the time window, for example, 60 seconds, to smooth out instantaneous fluctuations. Let τ be the average similarity between nodes i and j within the time window, τ be the time index variable, and t be the current time point.

[0106] Based on the similarity of all node pairs, construct a similarity matrix S(t) = [S ij (t)] {N×N}The matrix is ​​a symmetric matrix, and the diagonal elements are all 1s.

[0107] S220: Calculate the density of monitoring features.

[0108] For each node i, its monitoring feature density ρ i (t) is calculated using the kernel density estimation method, with the following formula:

[0109]

[0110] Where K(·) is the kernel function, commonly the Gaussian kernel function. h is the bandwidth parameter, which controls the smoothness of the density estimation and is usually selected using the Silverman rule.

[0111] Characteristic density ρ i (t) reflects the degree of clustering of node i in the feature space. High-density regions indicate the existence of node clusters with similar states, which can correspond to specific functional areas in the site, such as wind turbine clusters, photovoltaic arrays, etc.

[0112] S230: Station Zone.

[0113] Based on feature density, a density clustering algorithm is used to divide the entire site into multiple monitoring function zones.

[0114] In one implementation, the partitioning process includes the following steps:

[0115] 1) For each node i, find its density-reachable set of nodes, that is, nodes whose similarity to i is greater than the threshold ε and whose density reaches i.

[0116] 2) Starting from the core point, expand to form clusters;

[0117] 3) Assign nodes that are not core nodes but are located in the neighborhood of a core node to the corresponding clusters;

[0118] 4) Nodes that do not belong to any cluster are marked as noise and can be considered abnormal nodes.

[0119] The partitioning result is represented as C = {C1, C2, ..., C} K}, where each C k A monitoring function partition contains a group of monitoring nodes with similar states.

[0120] S240: Partitioned Load and Risk Prediction.

[0121] For each partition C k It predicts the overall load changes and failure risk probability within a short time window in the future.

[0122] Load forecasting employs time series forecasting models, such as autoregressive integral moving average models or long short-term memory networks. The input is the historical load sequence of nodes within a partition, and the output is the load forecast values ​​for multiple future time points.

[0123] Failure risk probability prediction is based on historical failure data and real-time status indicators. Classification models such as logistic regression or support vector machines are used to calculate the probability of a partition experiencing a failure in the future. Features include partition average load, load volatility, and equipment health index.

[0124] The prediction results are used to guide task allocation, prioritizing the deployment of tasks to partitions with lower load and lower risk.

[0125] In a collaborative deployment method for monitoring nodes at new energy power plants, the S300 uses monitoring function zoning and prediction results as a framework to model monitoring task allocation as a multi-objective optimization problem, and solves for the optimal deployment strategy using mathematical methods. This step automates and optimizes task allocation, improving the overall efficiency of the monitoring system.

[0126] Please refer to Figure 4 The diagram illustrates a flowchart of an exemplary collaborative deployment method S300 for monitoring nodes at a new energy power station, the contents of which include:

[0127] S310: Monitoring task modeling.

[0128] Monitoring tasks include data acquisition, data processing, anomaly detection, and control command issuance. Each task is represented as a tuple:

[0129] Task k =(ID) k Type k DataSize k Deadline k Priority k );

[0130] Among them, Task k For the k-th monitoring task, ID k For task identification, Type k For task type, DataSize k For data processing volume, Deadline k Priority is the deadline for the task. k This determines the task priority.

[0131] S320: Build an optimization model.

[0132] The task allocation problem is modeled as an optimization model with the objectives of task completion timeliness and load balancing.

[0133] Decision variable x ki Indicates whether task k is deployed on node i, x ki ∈{0,1}.

[0134] The objective function includes:

[0135] 1) Minimize the total task completion time:

[0136]

[0137] Where c ki Let be the execution time of task k on node i, including data transmission time and processing time.

[0138] 2) Maximize load balancing:

[0139]

[0140] Among them, L i For the load of node i, L avg This represents the average load.

[0141] The overall objective function is:

[0142]

[0143] Here, α is a weighting coefficient used to balance the relative importance of the two objectives.

[0144] The constraints include:

[0145] Each task must be assigned:

[0146] Node load does not exceed its capacity: w k Cap is the resource requirement weight for task k. i This represents the maximum resource capacity of node i;

[0147] The task must be completed before the deadline: c ki ≤Deadline k ,

[0148] S330: Model solution.

[0149] Since the task allocation problem is NP-hard, a heuristic algorithm is used to solve it for large-scale instances.

[0150] In one implementation method, a genetic algorithm is used to solve the problem:

[0151] 1) Encoding: The decoded data is converted into chromosomes, with each gene representing a task allocation node;

[0152] 2) Initialization: Randomly generate the initial population;

[0153] 3) Evaluation: Calculate the fitness value for each chromosome;

[0154] 4) Selection: The roulette wheel selection method is used to select outstanding individuals;

[0155] 5) Crossover: Uses a single-point crossover operation to generate offspring;

[0156] 6) Mutation: The process of gene mutation with a certain probability;

[0157] 7) Iteration: Repeat the above process until convergence.

[0158] Solving for the optimal deployment probability matrix P = [p ki ], where p ki This represents the probability that task k is assigned to node i.

[0159] S340: Task Assignment.

[0160] Based on the optimal deployment probability matrix P, a monitoring node is assigned to each monitoring task. A probabilistic selection method is used, that is, nodes are randomly selected according to the probability distribution, or the node with the highest probability is selected.

[0161] The allocation results are sent to each monitoring node for execution, and the node load status is updated.

[0162] In a collaborative deployment method for monitoring nodes at new energy power plants, the S400 continuously monitors the system status after task deployment and initiates a dynamic sharding mechanism to reallocate tasks when resource contention risks are detected. This step enhances the system's resilience and fault tolerance, ensuring the continuity and reliability of monitoring tasks.

[0163] Please refer to Figure 5 The document illustrates a flowchart of an exemplary collaborative deployment method S400 for monitoring nodes at a new energy power station, the contents of which include:

[0164] S410: Resource usage monitoring.

[0165] Continuously monitor the resource usage of each monitoring function partition and monitoring node, including indicators such as CPU utilization, memory usage, network bandwidth, and disk I / O.

[0166] Monitoring data is collected through an agent or monitoring framework, with a sampling frequency set to once per second.

[0167] S420: Resource competition risk identification.

[0168] If the resource utilization rate of a monitoring node exceeds a preset threshold, or if the resource utilization rate rises sharply in a short period of time, the node is determined to have a risk of resource contention.

[0169] Risk identification employs rule engines or machine learning models, such as a support vector machine-based classifier, to determine whether a node is about to become overloaded.

[0170] S430: Dynamic task sharding.

[0171] When a resource contention risk is identified, a dynamic sharding mechanism is activated to shard the monitoring tasks on the risky monitoring nodes into multiple parallel processing subtask groups.

[0172] The sharding strategy is determined based on the task type:

[0173] Data acquisition tasks: Slice data by data source or time window, such as dividing a large-scale data acquisition task into multiple small-area sub-tasks;

[0174] Data processing tasks: Divide data into blocks or processing stages, such as dividing a big data processing task into multiple parallel pipelines;

[0175] Control tasks: Segment by the controlled object, such as segmenting a group control command into multiple individual control commands.

[0176] The fragmented subtask groups should meet the conditions of parallel processing, low data dependency, and low communication overhead.

[0177] S440: Subtask redeployment.

[0178] After being sharded, the subtasks are redeployed to other monitoring nodes according to the S300 task allocation method.

[0179] When redeploying, the real-time load of nodes and task dependencies must be considered to ensure that subtasks are executed in a coordinated manner.

[0180] In one implementation, the redeployment process includes:

[0181] 1) Subtask priority adjustment: Adjust the subtask priority according to the original task priority and sharding situation;

[0182] 2) Node selection: Select nodes with low load and matching performance to deploy subtasks;

[0183] 3) Dependency Management: For subtasks with dependencies, ensure that they are deployed on nodes with low communication latency;

[0184] 4) Execution monitoring: Monitor the execution status of subtasks to ensure that the results are integrated after all subtasks are completed.

[0185] Through dynamic sharding and redeployment, the system can effectively cope with node overload and improve overall reliability and resource utilization.

[0186] This invention was fully validated at a 200MW wind farm, which includes 100 wind turbines, 20 transformer substations, 5 data collection line monitoring units, and 1 step-up substation monitoring center. The specific configuration used during implementation is as follows:

[0187] Data acquisition layer: Deploy 500 sensor nodes, acquire data at a frequency of 1 second / time, transmit data using the OPCUA protocol, and store data in a time-series database;

[0188] Analysis and computation layer: Configure an 8-node Kubernetes cluster to perform vector computation and optimization solutions. Similarity calculation uses SparkMLlib, and optimization solutions use GoogleORTools.

[0189] Task scheduling layer: The workflow engine is built on Apache Airflow and supports dynamic orchestration of monitoring tasks;

[0190] Dynamic sharding layer: Elasticsearch is used for task sharding, and the sharding strategy can be customized.

[0191] During the testing phase, typical monitoring tasks such as wind turbine group control, power prediction, and fault detection were simulated, totaling 1000 task instances. Test results show that:

[0192] The average task completion time was reduced by 35%, and the deadline satisfaction rate reached 99.2%.

[0193] Load balancing improved by 40%, and the standard deviation of node resource utilization decreased from 0.25 to 0.15;

[0194] The dynamic sharding mechanism successfully handled 15 node overload events, reducing the task interruption rate to 0.5%.

[0195] The system throughput increased by 50%, and the number of concurrent tasks supported increased from 200 to 300.

[0196] Specifically, during a sudden wind speed event, the system detected a sharp increase in the load on the wind turbine monitoring node. It automatically fragmented the data acquisition task and redeployed it to a backup node, preventing monitoring interruption. The entire process was completed within 10 seconds, ensuring the integrity of the site's operational data.

[0197] Example 3

[0198] According to Embodiment 1 of this application, a collaborative deployment system for monitoring nodes of new energy power stations is provided, the system comprising:

[0199] The multi-source data acquisition and vector construction module is used to collect multi-source heterogeneous data from distributed monitoring nodes within new energy power plants, and to construct a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data.

[0200] The functional zoning and risk prediction module is used to calculate the resource vector similarity between nodes based on the monitoring resource vector of each monitoring node, and calculate the monitoring feature density of each node based on the similarity; further, based on the monitoring feature density, the entire site is automatically divided into multiple monitoring functional zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted.

[0201] The task optimization and deployment module is used to model the allocation of monitoring tasks as an optimization model based on the monitoring function partitions and prediction results; solve the optimization model to obtain the optimal deployment probability of each monitoring task, and allocate the corresponding monitoring nodes to the monitoring tasks according to the optimal deployment probability;

[0202] The dynamic sharding and redeployment module is used to continuously monitor the resource usage of each monitoring function partition and monitoring node after task deployment. When it is identified that a monitoring node is at risk of resource contention due to task overload, the dynamic sharding mechanism is activated to shard the monitoring tasks on the monitoring node into multiple parallel processing sub-task groups, and input the sub-task groups into the task optimization and deployment module to redeploy the tasks according to the optimization model.

[0203] Example 4

[0204] According to Embodiment 1 of this application, a computer device is provided, the computer device including: a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, the at least one program, the code set or instruction set being loaded and executed by the processor to realize the collaborative deployment method of monitoring nodes of new energy power stations as described above.

[0205] Example 5

[0206] According to Embodiment 1 of this application, a computer read storage medium is provided, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to realize the collaborative deployment method of monitoring nodes of new energy power stations as described above.

[0207] Those skilled in the art will understand that the embodiments of this application are provided as methods, systems, or computer program products. Therefore, this application takes the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application takes the form of a computer program product implemented on one or more computer storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0208] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, are implemented by computer program instructions. These computer program instructions are provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0209] These computer program instructions are also stored in a computer read-memory memory (CROM) that can direct a computer or other programmed data processing device to operate in a specific manner, such that the instructions stored in the CROM produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0210] These computer program instructions are also loaded onto a computer or other programmed data processing device, causing a series of operational steps to be performed on the computer or other programmed device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmed device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0211] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0212] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for collaborative deployment of monitoring nodes at new energy power stations, characterized in that, include: S100: Collects multi-source heterogeneous data from distributed monitoring nodes within new energy power stations, and constructs a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data; S200: The monitoring resource vector calculates the monitoring feature density based on the similarity of the resource vectors of the monitoring nodes; Based on the density of monitoring features, the entire site is automatically divided into multiple monitoring function zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted. S300: Using the monitoring function partition and prediction results as a framework, the monitoring task allocation is modeled as an optimization model; the optimization model is solved to obtain the optimal deployment probability of each monitoring task, and the corresponding monitoring node is allocated to the monitoring task according to the optimal deployment probability; S400: After task deployment, continuously monitor the resource usage of each monitoring function partition and monitoring node; when it is identified that the monitoring node is at risk of resource contention due to task overload, start the dynamic sharding mechanism to shard the monitoring tasks on the monitoring node with contention risk into multiple parallel processing sub-task groups, and redeploy the sub-tasks according to S300.

2. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1, characterized in that, The monitoring resource vector constructed in S100 is: V i (t)=[v i1 (t),v i2 (t),...,v iD (t)]; Among them, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, V i (t) represents the monitoring resource vector; V i (t) includes at least three of the following dimensions: Electrical quantities include per-unit values ​​for voltage, current, and active power; The status dimension includes switch status codes and protection action flags; Environmental dimensions include standardized values ​​for temperature and wind speed; The resource status dimension includes CPU utilization, memory usage, and communication quality metrics.

3. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1 or 2, characterized in that, In step S200, the cosine similarity is used to calculate the resource vector similarity between monitoring nodes. Its expression is: Among them, S ij (t) represents the similarity between node i and node j at time t, S ij The value of V ranges from [1,1], and the closer the value is to 1, the more similar the states of nodes i and j are; i (t) represents the monitoring resource vector of the i-th node at time t, V j (t) represents the monitoring resource vector of the j-th node at time t, v id (t) represents the standardized monitoring value of the i-th node in the d-th dimension, v jd (t) represents the standardized monitoring value of the j-th node in the d-th dimension.

4. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1, characterized in that, The monitored feature density ρ in S200 i (t) is calculated using the kernel density estimation method, and its expression is: Where K(·) is the kernel function, h is the bandwidth parameter; density ρ i (t) reflects the degree of clustering of the neighbors around node i at time t, and N is the total number of monitoring nodes; high-density areas correspond to clusters of nodes with highly similar states, which are used to automatically divide the monitoring function into partitions.

5. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1, characterized in that, In S300, the monitoring task is modeled as a tuple task. k =(ID) k Type k DataSize k Deadline k Priority k ); where Task k For the k-th monitoring task, ID k For task identification, Type k For task type, DataSize k For data processing volume, Deadline k Priority is the task deadline. k This determines the task priority.

6. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1 or 5, characterized in that, The optimization model constructed in S300 is a multi-objective integer programming model, and its comprehensive objective function is: Among them, c ki Let L be the execution cost of task k on node i. i For the load of node i, L avg The load is the average load, and α is the weighting coefficient; the constraints include: Each task must be assigned: Node load does not exceed its capacity: w k Cap is the resource requirement weight for task k. i This represents the maximum resource capacity of node i; The task must be completed before the deadline: c ki ≤Deadline k , 7. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1, characterized in that, When the dynamic sharding mechanism in S400 detects the risk of resource contention among monitoring nodes, it divides the original monitoring task into multiple sub-task groups according to data dimensions or time windows. Each sub-task group inherits the priority and deadline constraints of the original task and is re-input into the optimization model of S300 as an independent task for redeployment.

8. The collaborative deployment method for monitoring nodes of new energy power stations as described in claim 1, characterized in that, Before S100, the process also includes standardizing the original multi-source heterogeneous data and using principal component analysis or autoencoder to reduce the dimensionality of high-dimensional monitoring indicators before constructing the monitoring resource vector. Principal components with a cumulative contribution rate of not less than 90% are retained to improve the efficiency and robustness of subsequent similarity calculation and density estimation.

9. A collaborative deployment system for monitoring nodes at new energy power stations, characterized in that, The system includes: The multi-source data acquisition and vector construction module is used to acquire multi-source heterogeneous data from distributed monitoring nodes within the new energy power station, and to construct a multi-dimensional monitoring resource vector for each distributed monitoring node based on the multi-source heterogeneous data. The functional zoning and risk prediction module is used to calculate the resource vector similarity between nodes based on the monitoring resource vector of each monitoring node, and calculate the monitoring feature density of each node based on the similarity; further, based on the monitoring feature density, the entire site is automatically divided into multiple monitoring functional zones, and the overall load change and failure risk probability of each zone in the future short time window are predicted. The task optimization and deployment module is used to model the allocation of monitoring tasks into an optimization model based on the monitoring function partitions and prediction results; solve the optimization model to obtain the optimal deployment probability of each monitoring task; and allocate corresponding monitoring nodes to the monitoring tasks according to the optimal deployment probability. The dynamic sharding and redeployment module is used to continuously monitor the resource usage of each monitoring function partition and monitoring node after task deployment. When it is identified that a monitoring node is at risk of resource contention due to task overload, the dynamic sharding mechanism is activated to shard the monitoring task on the monitoring node into multiple parallel processing sub-task groups, and the sub-task groups are input into the task optimization and deployment module to redeploy the task according to the optimization model.

10. A computer device, characterized in that, The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the collaborative deployment method for monitoring nodes of new energy power stations as described in any one of claims 1 to 8.

11. A computer read storage medium, characterized in that, The storage medium stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the collaborative deployment method for monitoring nodes of new energy power stations as described in any one of claims 1 to 8.

Images