A Low-Carbon Energy-Driven Flexible Distributed Computing Resource Control Method and System

Abstract

This invention relates to a low-carbon energy-driven, flexible distributed computing power resource control method and system, belonging to the fields of computer distributed systems and green energy management technology. The method includes: acquiring load data of distributed computing power nodes, identifying server resource usage signals, and analyzing the carbon efficiency mapping relationship between nodes using an energy consumption weight detection method to construct an energy-computing power characteristic correlation matrix; extracting the unit energy consumption computing power output coefficient of each computing power node, performing low-carbon computing power resource priority scheduling through a flexible task allocation algorithm, and generating node resource allocation instructions; establishing a distributed autonomous control model, correcting the node execution state through an energy-computing power coupling feedback mechanism to obtain a distributed computing power resource balance range; and establishing a flexible distributed computing power resource closed-loop control system, adaptively correcting task allocation weights, and controlling energy-task collaborative compensation between computing power nodes based on computing power node energy consumption feedback.

Description

Technical Field

[0001] This invention belongs to the field of computer distributed systems and green energy management technology, specifically relating to a method and system for controlling elastic distributed computing resources driven by low-carbon energy. Background Technology

[0002] With the rapid development of cloud computing, edge computing, and artificial intelligence technologies, distributed computing networks have become a core infrastructure supporting large-scale data processing and intelligent applications. However, traditional distributed computing scheduling systems generally prioritize performance maximization, neglecting constraints on energy efficiency and carbon emission costs. This leads to problems such as high energy consumption, low carbon efficiency, and uneven scheduling during the use of computing resources. Especially during peak computing demand periods, a large number of nodes experience a surge in energy consumption due to excessive load, which not only affects system stability but also makes carbon emissions difficult to control, contradicting the current green computing trend of synergistic development of computing power and energy.

[0003] In existing technologies, some studies attempt to optimize energy consumption through methods such as task migration, energy consumption awareness, or virtual machine scheduling. However, these methods are mostly static or single-node control, lacking dynamic modeling and feedback control mechanisms for the energy-computing power coupling relationship between heterogeneous computing power nodes, and thus failing to achieve system-level carbon efficiency collaborative scheduling. Furthermore, traditional energy consumption control strategies struggle to cope with fluctuations in distributed energy supply and changes in multidimensional task loads, easily leading to problems such as resource allocation lag and task scheduling imbalances in actual operation.

[0004] Therefore, there is an urgent need for a distributed computing resource control method that can achieve multi-node collaboration, adaptive adjustment of energy consumption feedback, and dynamic resource migration under the drive of low-carbon energy, so as to improve the energy utilization rate of computing networks and the green and intelligent level of system operation. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a low-carbon energy-driven, flexible distributed computing resource control method.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] S1: Obtain the load data of the distributed computing power nodes, identify server resource usage signals based on the load data of the distributed computing power nodes, and analyze the carbon efficiency mapping relationship between each node through the energy consumption weight detection method to construct an energy-computing power characteristic correlation matrix.

[0008] The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix.

[0009] By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix.

[0010] S2: Based on the energy-computing power feature correlation matrix, extract the unit energy consumption computing power output coefficient of each computing power node. When a peak in resource demand is detected, use the elastic task allocation algorithm to perform low-carbon computing power resource priority scheduling on the computing power tasks of the distributed heterogeneous node cluster and generate node resource allocation instructions.

[0011] S3: Establish a distributed autonomous control model based on the node resource allocation instructions, correct the node execution state through the energy-computing power coupling feedback mechanism, deploy a negative carbon computing power network based on the server resource usage signal, analyze the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm, and obtain the distributed computing power resource balance range.

[0012] The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback.

[0013] A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology.

[0014] In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load.

[0015] S4: Establish a closed-loop control system for elastic distributed computing resources. When the energy supply is detected to deviate from the threshold by more than a set ratio, dynamic resource migration is triggered. The task allocation weight is adaptively corrected through a segmented energy consumption optimization algorithm. During the execution cycle, the energy-task collaborative compensation between computing nodes is controlled based on the energy consumption feedback of the computing nodes.

[0016] Specifically, the load data of the distributed computing nodes is interpolated over time using a multi-source data fusion algorithm to analyze the interference of transient load fluctuations on energy consumption calculations, and the load characteristics of the current node's operating status are obtained based on the node resource usage signal.

[0017] Specifically, the carbon efficiency mapping relationship is calculated based on the ratio of node energy consumption to computing power output to obtain a carbon efficiency factor. The carbon efficiency mapping relationship between nodes in the negative carbon computing power network is analyzed by the energy consumption weight detection method to establish a carbon efficiency correlation map between nodes.

[0018] Specifically, the elastic task allocation algorithm is based on a hierarchical multi-objective optimization mechanism, taking task response time, node energy consumption coefficient and carbon weight as optimization objectives. It dynamically allocates computing power tasks through heuristic search, prioritizes scheduling energy consumption allocation tasks of high-load nodes, and generates a computing power task allocation queue for a distributed heterogeneous node cluster.

[0019] Specifically, the method for generating the node resource allocation instruction is as follows:

[0020] The computing power tasks of the distributed heterogeneous node cluster are monitored in real time. The computing resource requirements, execution latency constraints and energy consumption sensitivity of each task are analyzed. Based on the computing power task allocation queue, the task allocation priority is dynamically sorted to generate a preliminary task priority list.

[0021] Based on the preliminary task priority list, the computing resources of each task are matched among distributed nodes through the elastic task allocation algorithm. During the allocation process, high-energy-consuming tasks are allocated to high-carbon-efficiency nodes first according to the current load, heterogeneous performance characteristics and energy status of the nodes, so as to obtain a computing resource scheduling scheme.

[0022] The computing resource scheduling scheme is executed, the matching degree between the node execution status and the expected task completion effect is verified based on historical resource data, the allocation strategy for node execution task sequence deviating from the threshold is corrected, and node resource allocation instructions are generated.

[0023] Specifically, the energy-computing power coupling feedback mechanism combines different node types and energy supply paths to normalize the collected parameters, forming a basic computing power dataset that can be used for feedback control. The basic computing power dataset is then analyzed using a nonlinear dynamic weight algorithm to calculate the impact coefficient of energy changes on computing power performance, generating a corresponding coupling feedback signal. This coupling feedback signal is then input into the control loop, and a dynamic energy-computing power coupling compensation channel is established based on the corrected task allocation weights and energy consumption priorities.

[0024] Specifically, the negative carbon computing network dynamically allocates low-carbon energy input by combining the node energy consumption status and carbon emission intensity, and performs carbon neutrality of the node cluster by connecting to external low-carbon energy sources.

[0025] Specifically, the method for generating the distributed computing resource balance interval is as follows:

[0026] Based on the real-time energy consumption data and computing load data of each node in the distributed computing network, the energy supply deviation threshold of the node is calculated by the carbon balance scheduling algorithm, and the energy utilization deviation range of the node is determined by combining the historical energy consumption curve of the node.

[0027] The resource balance critical point is calculated by a multi-node collaborative fitting algorithm. The resource balance critical point is dynamically corrected based on the distributed computing power resource closed-loop control system, the resource balance critical point is adaptively updated, and a distributed computing power resource balance interval is generated.

[0028] Specifically, the elastic distributed computing power resource closed-loop control system is based on an energy-computing power coupling feedback mechanism. When the energy supply is detected to deviate from the threshold by more than a set ratio, a resource dynamic migration mechanism is triggered. The task allocation weight is adaptively adjusted using a segmented energy consumption optimization algorithm. Based on the energy consumption feedback of the computing power nodes, the energy-task allocation parameters between nodes are collaboratively compensated.

[0029] Specifically, a low-carbon energy-driven, flexible distributed computing resource control system is characterized by comprising:

[0030] Computing power feature association module: acquires load data of distributed computing power nodes, identifies server resource usage signals based on the load data of the distributed computing power nodes, and analyzes the carbon efficiency mapping relationship between nodes through the energy consumption weight detection method to construct an energy-computing power feature association matrix;

[0031] The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix.

[0032] By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix.

[0033] Distributed autonomous control resource allocation and scheduling module: Establishes a distributed autonomous control model based on the node resource allocation instructions, corrects the node execution state through an energy-computing power coupling feedback mechanism, deploys a negative carbon computing power network based on the server resource usage signals, and analyzes the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm to obtain the distributed computing power resource balance range.

[0034] The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback.

[0035] A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology.

[0036] In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load.

[0037] Elastic closed-loop optimization module: Establishes an elastic distributed computing power resource closed-loop control system. When the energy supply is detected to deviate from the threshold by more than a set ratio, it triggers dynamic resource migration. It adaptively corrects the task allocation weight through a segmented energy consumption optimization algorithm, and controls the energy-task collaborative compensation between computing power nodes based on the energy consumption feedback of computing power nodes during the execution cycle.

[0038] The beneficial effects of this invention are as follows:

[0039] This invention achieves deep coordination between computing resource scheduling and energy utilization by introducing a low-carbon energy-driven mechanism and an energy-computing power coupling control model into a distributed computing power system. Compared with existing technologies, this invention has the following advantages:

[0040] First, this invention constructs an energy-computing power characteristic correlation matrix, achieving a quantitative mapping between the energy consumption characteristics and computing power output efficiency of nodes. This allows for accurate identification of carbon efficiency differences among nodes, providing data support for low-carbon scheduling. Second, it employs an elastic task allocation algorithm to dynamically adjust task allocation weights based on fluctuations in computing power demand and node energy efficiency levels. During peak computing power demand periods, high-carbon-efficiency nodes are prioritized for scheduling, thereby effectively reducing the overall energy consumption of the system.

[0041] Furthermore, this invention achieves node-level adaptive energy consumption adjustment and global carbon balance control by establishing a distributed autonomous control model and an energy-computing power coupling feedback mechanism, enabling stable system operation even under fluctuating energy supply conditions. The distributed computing power resource balance range determined by the carbon balance scheduling algorithm makes task migration and resource allocation more flexible and controllable.

[0042] Finally, the elastic distributed computing power resource closed-loop control system constructed in this invention can automatically trigger dynamic migration and energy consumption optimization processes when energy consumption deviation is detected, forming a self-learning and self-correcting energy-computing power collaboration mechanism, thereby significantly improving the energy utilization efficiency and low-carbon operation capability of the computing power network, and has broad engineering application value. Attached Figure Description

[0043] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0044] Figure 1 This is a schematic diagram of the structure of a low-carbon energy-driven elastic distributed computing resource control method and system according to the present invention.

[0045] Figure 2 This is a schematic diagram illustrating the technical flow of the low-carbon energy-driven elastic distributed computing resource control method and system of the present invention. Detailed Implementation

[0046] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0047] Please see Figure 1 A method for controlling elastic distributed computing resources driven by low-carbon energy:

[0048] S1: Obtain the load data of the distributed computing power nodes, identify server resource usage signals based on the load data of the distributed computing power nodes, and analyze the carbon efficiency mapping relationship between each node through the energy consumption weight detection method to construct an energy-computing power characteristic correlation matrix.

[0049] The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix.

[0050] By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix.

[0051] S2: Based on the energy-computing power feature correlation matrix, extract the unit energy consumption computing power output coefficient of each computing power node. When a peak in resource demand is detected, use the elastic task allocation algorithm to perform low-carbon computing power resource priority scheduling on the computing power tasks of the distributed heterogeneous node cluster and generate node resource allocation instructions.

[0052] S3: Establish a distributed autonomous control model based on the node resource allocation instructions, correct the node execution state through the energy-computing power coupling feedback mechanism, deploy a negative carbon computing power network based on the server resource usage signal, analyze the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm, and obtain the distributed computing power resource balance range.

[0053] The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback.

[0054] A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology.

[0055] In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load.

[0056] S4: Establish a closed-loop control system for elastic distributed computing resources. When the energy supply is detected to deviate from the threshold by more than a set ratio, dynamic resource migration is triggered. The task allocation weight is adaptively corrected through a segmented energy consumption optimization algorithm. During the execution cycle, the energy-task collaborative compensation between computing nodes is controlled based on the energy consumption feedback of the computing nodes.

[0057] Specifically, the load data of the distributed computing nodes is interpolated over time using a multi-source data fusion algorithm to analyze the interference of transient load fluctuations on energy consumption calculations, and the load characteristics of the current node's operating status are obtained based on the node resource usage signal.

[0058] Specifically, the carbon efficiency mapping relationship is calculated based on the ratio of node energy consumption to computing power output to obtain a carbon efficiency factor. The carbon efficiency mapping relationship between nodes in the negative carbon computing power network is analyzed by the energy consumption weight detection method to establish a carbon efficiency correlation map between nodes.

[0059] In this embodiment, low-carbon autonomous operation of the distributed computing power system is achieved through four stages: energy consumption acquisition, carbon efficiency modeling, elastic scheduling, and closed-loop optimization, combined with energy-computing power coupling feedback equations and dynamic constraint optimization algorithms. This negative-carbon computing power network not only reduces energy consumption intensity but also adaptively maintains the system's carbon balance under multiple energy supply conditions, realizing a full-link closed-loop control mechanism of "carbon sensing—computing power scheduling—energy consumption optimization." Specifically, the negative-carbon computing power network is based on a cloud-edge collaborative architecture, achieving dynamic balance and adaptive control of distributed computing power resources under negative carbon constraints through computing node energy consumption sensing, carbon efficiency modeling, scheduling optimization, and closed-loop feedback.

[0060] The real-time operating parameters of the distributed computing node ni are monitored through the node energy consumption acquisition module, including power consumption P. i (t) Computing power output C i (t), temperature coefficient T i (t) and energy type weight λ i The energy consumption weight detection algorithm is used to calculate the output coefficient η of computing power per unit energy consumption. i :

[0061] ,

[0062] Where, λ i Representing the energy carbon intensity weights, the system obtains the carbon efficiency mapping matrix M through feature normalization and time sliding window smoothing. ce =[η1,η2,...,η n ].

[0063] Based on the node energy consumption and carbon efficiency characteristics, the system constructs an energy-computing power characteristic correlation matrix A. es Its elements represent the computing power contribution of the i-th node to the j-th task per unit energy consumption:

[0064] ,

[0065] Among them, w j The task energy consumption weight is determined by a combination of task complexity, latency constraints, and energy sensitivity. The scheduling engine is based on an improved multi-objective elastic task allocation algorithm to minimize the energy cost function E. total And constrain carbon emission levels:

[0066] ,

[0067] Where, δ i Assign a proportion to the node tasks, C req Λ is the global computing power requirement threshold. max This is the upper limit for the system's carbon emissions.

[0068] Once the node allocation scheme is determined, the system establishes a negative carbon computing power network topology G based on the carbon efficiency matrix and energy consumption weight calculation results. nc =(N,E), where the node set N represents computing power nodes, and the edge set E represents the energy-task interaction links. Each node's energy feedback signal ƒ i (t) is defined as:

[0069] ,

[0070] Where η is the average carbon efficiency of the network, Λ cur Given the current carbon emissions, α and β are adjustment weights. The system dynamically adjusts the node task weights δ based on feedback signals. i (t+1)=δi(t)+γfi(t) to achieve node-level adaptive energy balance.

[0071] The system periodically executes a carbon balance scheduling algorithm to determine whether network energy consumption deviates from a threshold. When the deviation ΔE(t) exceeds a set range... When this occurs, the energy optimization function is triggered:

[0072] ,

[0073] Among them, E opt (t) represents the optimal energy consumption estimate, and μ is the learning rate parameter. Through this closed-loop correction mechanism, the negative carbon computing network can achieve coordinated adjustment of energy and computing power, enabling the system to maintain a carbon-neutral or carbon-negative balance under dynamic load.

[0074] Specifically, the method for generating the node resource allocation instruction is as follows:

[0075] The computing power tasks of the distributed heterogeneous node cluster are monitored in real time. The computing resource requirements, execution latency constraints and energy consumption sensitivity of each task are analyzed. Based on the computing power task allocation queue, the task allocation priority is dynamically sorted to generate a preliminary task priority list.

[0076] Based on the preliminary task priority list, the computing resources of each task are matched among distributed nodes through the elastic task allocation algorithm. During the allocation process, high-energy-consuming tasks are allocated to high-carbon-efficiency nodes first according to the current load, heterogeneous performance characteristics and energy status of the nodes, so as to obtain a computing resource scheduling scheme.

[0077] The computing resource scheduling scheme is executed, the matching degree between the node execution status and the expected task completion effect is verified based on historical resource data, the allocation strategy for node execution task sequence deviating from the threshold is corrected, and node resource allocation instructions are generated.

[0078] Specifically, the energy-computing power coupling feedback mechanism combines different node types and energy supply paths to normalize the collected parameters, forming a basic computing power dataset that can be used for feedback control. The basic computing power dataset is then analyzed using a nonlinear dynamic weight algorithm to calculate the impact coefficient of energy changes on computing power performance, generating a corresponding coupling feedback signal. This coupling feedback signal is then input into the control loop, and a dynamic energy-computing power coupling compensation channel is established based on the corrected task allocation weights and energy consumption priorities.

[0079] Specifically, the negative carbon computing network dynamically allocates low-carbon energy input by combining the node energy consumption status and carbon emission intensity, and performs carbon neutrality of the node cluster by connecting to external low-carbon energy sources.

[0080] This embodiment provides a method and system for controlling elastic distributed computing resources driven by low-carbon energy. It is deployed in cloud computing centers, edge nodes and green energy computing network convergence platforms. Through distributed computing data collection, energy consumption weight analysis, carbon efficiency modeling and closed-loop scheduling control, it realizes low-carbon priority allocation of computing tasks and energy-computing power collaborative optimization.

[0081] like Figure 1 The system shown consists of a data acquisition layer, an energy consumption analysis layer, a computing power scheduling layer, and an autonomous control layer. The data acquisition layer acquires real-time data on CPU utilization, GPU power consumption, energy type, and temperature sensing data of distributed computing nodes through multi-source monitoring interfaces (such as SNMP collectors, Prometheus node exporters, and energy consumption sensing modules). It also performs data normalization, noise filtering, and time series synchronization using streaming computing frameworks (such as Apache Flink and Kafka Streams). The energy consumption analysis layer calculates the carbon efficiency mapping relationship between nodes based on energy consumption weight detection algorithms and carbon efficiency assessment models, generating an energy-computing power characteristic correlation matrix to quantify the computing power output per unit of energy consumption of each node.

[0082] In the computing power scheduling layer, the system uses an elastic task allocation algorithm (based on an improved genetic algorithm and a multi-objective constrained programming model) to extract task load characteristics and energy consumption sensitivity parameters, and combines the energy-computing power characteristic correlation matrix to generate node resource allocation instructions. This process is executed by a scheduling engine (such as Kubernetes Scheduler or Ray Tune) to achieve low-carbon priority scheduling and dynamic load balancing of distributed heterogeneous nodes.

[0083] The autonomous control layer, based on a distributed autonomous control model and an energy-computing power coupling feedback mechanism, uses message queues (such as RabbitMQ or ZeroMQ) to achieve state synchronization and energy consumption feedback transmission between nodes. The system employs a carbon balance scheduling algorithm to dynamically calculate energy supply deviation thresholds, generating a distributed computing power resource balance interval. When an energy deviation exceeds a set proportion, a dynamic resource migration module is triggered. This module calls a segmented energy consumption optimization algorithm (based on distributed gradient descent and energy consumption gradient backtracking mechanisms) to adaptively adjust task allocation weights, achieving coordinated compensation between energy and tasks and closed-loop carbon balance control of the system.

[0084] The technology stack includes:

[0085] Distributed monitoring and data collection: Prometheus, Kafka, InfluxDB;

[0086] Energy consumption analysis and modeling: Python (NumPy, Pandas, SciPy), TensorFlow, XGBoost;

[0087] Scheduling and Optimization: Kubernetes, Ray, Optuna, Ceres Solver;

[0088] Control and feedback: gRPC, RabbitMQ, Redis, ZeroMQ;

[0089] Visualization and report generation: Grafana, Plotly, ReportLab, Jinja2.

[0090] The specific technical flow is as follows: node energy consumption collection → feature normalization → energy consumption weight analysis → energy-computing power matrix construction → elastic task scheduling → distributed autonomous control → energy-computing power feedback → dynamic migration and optimization closed loop.

[0091] Based on this, such as Figure 2 The system shown can adaptively correct its operating status through energy consumption anomaly detection and carbon efficiency regression analysis modules. When the energy consumption of a node deviates from the threshold, it triggers an iterative optimization process. Based on rolling updates and reinforcement learning algorithms, it dynamically adjusts the carbon efficiency model weights and scheduling strategy parameters to achieve continuous optimization and green autonomous operation of the low-carbon distributed computing system.

[0092] Specifically, the method for generating the distributed computing resource balance interval is as follows:

[0093] Based on the real-time energy consumption data and computing load data of each node in the distributed computing network, the energy supply deviation threshold of the node is calculated by the carbon balance scheduling algorithm, and the energy utilization deviation range of the node is determined by combining the historical energy consumption curve of the node.

[0094] The resource balance critical point is calculated by a multi-node collaborative fitting algorithm. The resource balance critical point is dynamically corrected based on the distributed computing power resource closed-loop control system, the resource balance critical point is adaptively updated, and a distributed computing power resource balance interval is generated.

[0095] Specifically, the elastic distributed computing power resource closed-loop control system is based on an energy-computing power coupling feedback mechanism. When the energy supply is detected to deviate from the threshold by more than a set ratio, a resource dynamic migration mechanism is triggered. The task allocation weight is adaptively adjusted using a segmented energy consumption optimization algorithm. Based on the energy consumption feedback of the computing power nodes, the energy-task allocation parameters between nodes are collaboratively compensated.

[0096] Specifically, a low-carbon energy-driven, flexible distributed computing resource control system is characterized by comprising:

[0097] Computing power feature association module: acquires load data of distributed computing power nodes, identifies server resource usage signals based on the load data of the distributed computing power nodes, and analyzes the carbon efficiency mapping relationship between nodes through the energy consumption weight detection method to construct an energy-computing power feature association matrix;

[0098] The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix.

[0099] By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix.

[0100] Resource allocation and scheduling module: Based on the energy-computing power feature correlation matrix, extract the unit energy consumption computing power output coefficient of each computing power node. When a peak in resource demand is detected, use the elastic task allocation algorithm to perform low-carbon computing power resource priority scheduling on the computing power tasks of the distributed heterogeneous node cluster and generate node resource allocation instructions.

[0101] Distributed autonomous control resource allocation and scheduling module: Establishes a distributed autonomous control model based on the node resource allocation instructions, corrects the node execution state through an energy-computing power coupling feedback mechanism, deploys a negative carbon computing power network based on the server resource usage signals, and analyzes the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm to obtain the distributed computing power resource balance range.

[0102] The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback.

[0103] A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology.

[0104] In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load.

[0105] Elastic closed-loop optimization module: Establishes an elastic distributed computing power resource closed-loop control system. When the energy supply is detected to deviate from the threshold by more than a set ratio, it triggers dynamic resource migration. It adaptively corrects the task allocation weight through a segmented energy consumption optimization algorithm, and controls the energy-task collaborative compensation between computing power nodes based on the energy consumption feedback of computing power nodes during the execution cycle.

[0106] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for controlling flexible distributed computing resources driven by low-carbon energy, characterized in that, include: S1: Obtain the load data of the distributed computing power nodes, identify server resource usage signals based on the load data of the distributed computing power nodes, and analyze the carbon efficiency mapping relationship between each node through the energy consumption weight detection method to construct an energy-computing power characteristic correlation matrix. The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix. By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix. S2: Based on the energy-computing power feature correlation matrix, extract the unit energy consumption computing power output coefficient of each computing power node. When a peak in resource demand is detected, use the elastic task allocation algorithm to perform low-carbon computing power resource priority scheduling on the computing power tasks of the distributed heterogeneous node cluster and generate node resource allocation instructions. S3: Establish a distributed autonomous control model based on the node resource allocation instructions, correct the node execution state through the energy-computing power coupling feedback mechanism, deploy a negative carbon computing power network based on the server resource usage signal, analyze the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm, and obtain the distributed computing power resource balance range. The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback. A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology. In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load. S4: Establish a closed-loop control system for elastic distributed computing resources. When the energy supply is detected to deviate from the threshold by more than a set ratio, dynamic resource migration is triggered. The task allocation weight is adaptively corrected through a segmented energy consumption optimization algorithm. During the execution cycle, the energy-task collaborative compensation between computing nodes is controlled based on the energy consumption feedback of the computing nodes.

2. The method according to claim 1, characterized in that, The load data of the distributed computing nodes is interpolated over time using a multi-source data fusion algorithm to analyze the interference of transient load fluctuations on energy consumption calculations, and the load characteristics of the current node's operating status are obtained based on the node resource usage signals.

3. The method according to claim 1, characterized in that, The carbon efficiency mapping relationship is based on the carbon efficiency factor calculated by the ratio of node energy consumption to computing power output, and the carbon efficiency mapping relationship between nodes in the negative carbon computing power network is analyzed by the energy consumption weight detection method to establish a carbon efficiency correlation map between nodes.

4. The method according to claim 2, characterized in that, The elastic task allocation algorithm is based on a hierarchical multi-objective optimization mechanism. It takes task response time, node energy consumption coefficient and carbon weight as optimization objectives, and dynamically allocates computing power tasks through heuristic search. It prioritizes scheduling energy consumption allocation tasks of high-load nodes and generates a computing power task allocation queue for a distributed heterogeneous node cluster.

5. The method according to claim 4, characterized in that, The method for generating the node resource allocation instruction is as follows: The computing power tasks of the distributed heterogeneous node cluster are monitored in real time. The computing resource requirements, execution latency constraints and energy consumption sensitivity of each task are analyzed. Based on the computing power task allocation queue, the task allocation priority is dynamically sorted to generate a preliminary task priority list. Based on the preliminary task priority list, the computing resources of each task are matched among distributed nodes through the elastic task allocation algorithm. During the allocation process, high-energy-consuming tasks are allocated to high-carbon-efficiency nodes first according to the current load, heterogeneous performance characteristics and energy status of the nodes, so as to obtain a computing resource scheduling scheme. The computing resource scheduling scheme is executed, the matching degree between the node execution status and the expected task completion effect is verified based on historical resource data, the allocation strategy for node execution task sequence deviating from the threshold is corrected, and node resource allocation instructions are generated.

6. The method according to claim 2, characterized in that, The energy-computing power coupling feedback mechanism combines different node types and energy supply paths to normalize the collected parameters, forming a basic computing power dataset that can be used for feedback control. It then uses a nonlinear dynamic weighting algorithm to analyze the basic computing power dataset, calculate the impact coefficient of energy changes on computing power performance, generate a corresponding coupling feedback signal, and input the coupling feedback signal into the control loop. Based on the corrected task allocation weight and energy consumption priority, a dynamic energy-computing power coupling compensation channel is established.

7. The method according to claim 1, characterized in that, The negative carbon computing network dynamically allocates low-carbon energy input by combining the node's energy consumption status and carbon emission intensity, and performs carbon neutrality of the node cluster by connecting to an external low-carbon energy source.

8. The method according to claim 1, characterized in that, The method for generating the distributed computing power resource balance interval is as follows: Based on the real-time energy consumption data and computing load data of each node in the distributed computing network, the energy supply deviation threshold of the node is calculated by the carbon balance scheduling algorithm, and the energy utilization deviation range of the node is determined by combining the historical energy consumption curve of the node. The resource balance critical point is calculated by a multi-node collaborative fitting algorithm. The resource balance critical point is dynamically corrected based on the distributed computing power resource closed-loop control system, the resource balance critical point is adaptively updated, and a distributed computing power resource balance interval is generated.

9. The method according to claim 1, characterized in that, The elastic distributed computing power resource closed-loop control system is based on an energy-computing power coupling feedback mechanism. When the energy supply is detected to deviate from the threshold by more than a set ratio, a resource dynamic migration mechanism is triggered. The task allocation weight is adaptively adjusted using a segmented energy consumption optimization algorithm, and the energy-task allocation parameters between nodes are collaboratively compensated based on the energy consumption feedback of the computing power nodes.

10. A low-carbon energy-driven, flexible distributed computing resource control system, characterized in that, include: Computing power feature association module: acquires load data of distributed computing power nodes, identifies server resource usage signals based on the load data of the distributed computing power nodes, and analyzes the carbon efficiency mapping relationship between nodes through the energy consumption weight detection method to construct an energy-computing power feature association matrix; The method for constructing the energy-computing power characteristic correlation matrix is ​​as follows: based on the load data of the distributed computing power nodes, analyze the computing power utilization rate of each node in different time windows, and calculate the carbon efficiency mapping value between nodes using the energy consumption weight detection method according to the normalized computing power utilization rate and energy consumption index, so as to form a preliminary carbon efficiency matrix. By combining the energy consumption characteristics of distributed nodes, the preliminary carbon efficiency matrix is ​​weighted and corrected, multi-dimensional energy-computing power feature vectors of nodes are extracted, and the multi-dimensional feature vectors are arranged in a matrix according to the correspondence between nodes and energy to generate an energy-computing power feature correlation matrix. Resource allocation and scheduling module: Based on the energy-computing power feature correlation matrix, extract the unit energy consumption computing power output coefficient of each computing power node. When a peak in resource demand is detected, use the elastic task allocation algorithm to perform low-carbon computing power resource priority scheduling on the computing power tasks of the distributed heterogeneous node cluster and generate node resource allocation instructions. Distributed autonomous control resource allocation and scheduling module: Establishes a distributed autonomous control model based on the node resource allocation instructions, corrects the node execution state through an energy-computing power coupling feedback mechanism, deploys a negative carbon computing power network based on the server resource usage signals, and analyzes the energy supply deviation threshold of the negative carbon computing power network based on the carbon balance scheduling algorithm to obtain the distributed computing power resource balance range. The method for constructing the distributed autonomous control model is as follows: each computing node in the distributed computing power network is extracted as an autonomous unit, and an energy-computing power coupling architecture is established based on the computing power performance indicators and task execution status of the autonomous unit. The response capability of task scheduling is analyzed through a node consensus algorithm to obtain node energy consumption feedback. A distributed autonomous collaborative control mechanism is constructed based on the node energy consumption feedback. The energy-computing power coupling architecture combines information interaction and state synchronization between nodes, and coordinates the global carbon balance constraint by introducing dynamic carbon efficiency weights to form a multi-node autonomous control topology. In the distributed autonomous control process, the execution status of nodes is corrected in real time through the distributed autonomous collaborative control mechanism. When the carbon efficiency of a node deviates from the set threshold, the global carbon balance control parameters are dynamically corrected, and a distributed autonomous control model is constructed based on the computing load. Elastic closed-loop optimization module: Establishes an elastic distributed computing power resource closed-loop control system. When the energy supply is detected to deviate from the threshold by more than a set ratio, it triggers dynamic resource migration. It adaptively corrects the task allocation weight through a segmented energy consumption optimization algorithm, and controls the energy-task collaborative compensation between computing power nodes based on the energy consumption feedback of computing power nodes during the execution cycle.

Images