Dynamic bandwidth allocation method and system for electric power communication network

Abstract

The invention relates to the technical field of electric power system communication, in particular to a dynamic bandwidth allocation method and system for an electric power communication network, and the method comprises the steps: constructing a service traffic characteristic database, dividing service types, defining indexes such as a bandwidth threshold value, and predicting traffic fluctuation by adopting an LSTM-GARCH hybrid neural network; parameters such as a link bandwidth occupancy rate are obtained in real time, and a network state matrix is constructed; based on the prediction model and the network state, generating a bandwidth allocation scheme through a deep reinforcement learning algorithm optimized by a near-end strategy; a software-defined network controller is used to execute a strategy, and a gradient descent method is combined to optimize parameters to form a closed loop. By dynamically adapting the service flow and the network state, the bandwidth utilization rate is improved, the key service transmission quality is guaranteed, and efficient and stable operation of the electric power communication network is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system communications, and relates to a dynamic bandwidth allocation method and system for a power communication network. Background Art

[0002] Current bandwidth allocation in power communication networks primarily relies on static or rule-based semi-dynamic allocation. The first is the static allocation method: A bandwidth allocation scheme is pre-set based on historical traffic volume, allocating fixed bandwidth to different services, such as relay protection, distribution network automation, and power consumption data collection. This approach lacks the ability to detect real-time traffic fluctuations and network status changes. For example, a fixed allocation of 2 Mbps of bandwidth to relay protection services cannot be freed up for other services even during low-load periods. The second is a priority-based fixed scheduling strategy: Bandwidth allocation is prioritized based solely on service type (e.g., relay protection > distribution network automation > remote meter reading), using a first-come, first-served basis or a fixed ratio to allocate resources. This approach is unable to adapt to the dynamic bandwidth demands of emergent services (such as big data transmission during troubleshooting). The third is the manual experience-driven adjustment model: This relies on network operations and maintenance personnel to manually adjust bandwidth allocation based on real-time alerts. This results in slow response times (typically taking several to tens of minutes) and struggles to optimize overall network resources.

[0003] Existing technologies do not build a dynamic correlation model between power business characteristics and bandwidth requirements, and are unable to quantitatively evaluate the impact of bandwidth allocation on power system security. They also do not consider the impact of dynamic factors such as electromagnetic interference and terminal access / exit in power communication networks on bandwidth quality.

[0004] Specifically, existing technologies suffer from the following shortcomings: low bandwidth utilization. Static allocation leads to transmission delays for critical services during high-load periods due to insufficient bandwidth, and idle resources during low-load periods. Average bandwidth utilization across the network is typically less than 40%. Service assurance capabilities are insufficient, and the real-time and reliability requirements of services are not differentiated. When bandwidth resources are tight, real-time services such as relay protection may suffer message loss due to insufficient bandwidth allocation, impacting the safe and stable operation of the power system. Dynamic adaptability is weak, making it unable to respond to scenarios such as dynamic changes in the number of terminals in the power communication network (such as the batch access of distribution automation terminals) and sudden increases in service traffic (such as the upload of fault recording data), easily causing network congestion. The lack of intelligent optimization mechanisms and reliance on manual intervention or fixed rules make it unable to adapt to the diverse bandwidth demands of the vast number of services in smart grids (such as 5G+ power slicing and IoT terminals). Summary of the Invention

[0005] The present invention provides a method and system for dynamic bandwidth allocation in an electric power communication network, which overcomes the shortcomings of the above-mentioned existing technologies. It can effectively solve the problems that the existing technologies cannot quantitatively evaluate the impact of bandwidth allocation on the security of the power system, and do not consider the impact of dynamic factors on bandwidth quality.

[0006] One of the technical solutions of the present invention is achieved through the following measures: A dynamic bandwidth allocation method for a power communication network includes the following steps: Step S1: Build a traffic feature database for electric power communication services, define bandwidth thresholds, maximum allowable delays, and reliability indicators for various services, and predict service traffic fluctuations; Step S2: Obtain the bandwidth occupancy rate, interference intensity, delay jitter, and bit error rate parameters of each link in real time to build a network status matrix; Step S3: Generate a bandwidth allocation plan based on the service demand prediction model and network status, with the optimization goals of maximizing bandwidth utilization, ensuring key service rates, and minimizing network latency. Establish a mathematical model that includes service demand constraints, bandwidth resource constraints, and interference avoidance constraints. Step S4: Execute the allocation strategy, dynamically adjust the service bandwidth allocation, collect actual transmission indicators and compare them with the model prediction values, optimize the model parameters, and form a closed-loop optimization mechanism.

[0007] The following is a further optimization and / or improvement of one of the above-mentioned technical solutions: In the above step S1, the business traffic fluctuation is predicted using an LSTM-GARCH hybrid neural network model with a prediction error rate of ≤10%.

[0008] The real-time acquisition of parameters in the above step S2 is achieved by deploying distributed sensing probes at power communication base stations, aggregation nodes and key terminals. The probes preprocess the data through edge computing nodes, and the preprocessing delay is ≤20ms.

[0009] In the above step S3, a bandwidth allocation scheme is generated by using a deep reinforcement learning algorithm based on proximal policy optimization, and the convergence time of the algorithm is ≤ 1000 rounds of training.

[0010] In the above step S4, the model parameters are optimized using the gradient descent method.

[0011] The closed-loop optimization mechanism formed in the above step S4 may include establishing an expert knowledge base and triggering manual intervention when the actual transmission indicator deviates from the predicted value by exceeding a threshold.

[0012] The second technical solution of the present invention is achieved through the following measures: a dynamic bandwidth allocation system for a power communication network, comprising: The service modeling module is used to build a database of power communication service traffic characteristics, define bandwidth thresholds, maximum allowable delays, and reliability indicators for various services, and predict service traffic fluctuations; The network perception module is used to obtain the bandwidth occupancy rate, interference intensity, delay jitter, and bit error rate parameters of each link in real time and build a network status matrix; The intelligent optimization module generates a bandwidth allocation plan based on the business demand prediction model and network status. It sets the optimization goals of maximizing bandwidth utilization, ensuring key business performance, and minimizing network latency. It establishes a mathematical model that includes business demand constraints, bandwidth resource constraints, and interference avoidance constraints. The execution control module is used to execute the allocation strategy, dynamically adjust the service bandwidth allocation, collect actual transmission indicators and compare them with the model prediction values, optimize the model parameters, and form a closed-loop optimization mechanism.

[0013] The following is a further optimization and / or improvement of the second technical solution of the above invention: The business modeling module may include: The service classification unit is used to classify services and define the bandwidth threshold and maximum allowable delay for each type of service; The traffic prediction unit is used to predict business traffic fluctuations.

[0014] The network awareness module may include: Distributed sensing probes are deployed at power communication base stations, aggregation nodes, and key terminals to collect link bandwidth occupancy, electromagnetic interference intensity, delay jitter, and bit error rate parameters; Edge computing nodes are used to pre-process the perception data and build the network status matrix.

[0015] The execution control module may include: Software-defined network controller for millisecond-level bandwidth slice switching; Parameter optimization unit, used to update the prediction model and optimize algorithm parameters.

[0016] The present invention constructs a traffic feature database, classifies power communication services, and defines relevant indicators. It then uses a neural network algorithm to predict traffic fluctuations, simultaneously sensing network link state parameters in real time and constructing a network state matrix. Based on the prediction model and network state, a deep reinforcement learning algorithm is used to generate an optimal bandwidth allocation scheme to maximize bandwidth utilization, improve the key service guarantee rate, and minimize network latency. Finally, a software-defined network controller executes the allocation strategy and optimizes model parameters by comparing actual transmission indicators with predicted values, forming a closed-loop optimization mechanism. The present invention can dynamically adapt to changes in traffic flow and network status in the power communication network, accurately predict traffic trends, and quickly respond to link state changes, effectively avoiding bandwidth resource waste and uneven allocation, and significantly improving network resource utilization efficiency. For critical services in power communication that require extremely high real-time performance and reliability, priority protection and dynamic adjustment strategies are used to ensure transmission stability and low latency, effectively reducing packet loss and bit error rates in data transmission, and ensuring the reliable operation of services such as power system control and relay protection. At the same time, the closed-loop optimization mechanism enables the system to continuously self-learn and improve, adapting to the complex and changeable operating environment of the power communication network, enhancing the network's anti-interference and fault response capabilities, and ultimately achieving efficient, stable, and intelligent dynamic bandwidth allocation in the power communication network, providing solid communication guarantees for the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Attachment Figure 1 The figure is a flow chart of a method for dynamic bandwidth allocation in a power communication network according to an embodiment of the present invention.

[0018] Attachment Figure 2 Schematic diagram of the structure of a dynamic bandwidth allocation system for a power communication network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is not limited to the following embodiments, and specific implementation methods can be determined based on the technical solutions of the present invention and actual conditions.

[0020] The present invention will be further described below in conjunction with the embodiments: Example 1: As shown in the attached Figure 1As shown, the dynamic bandwidth allocation method for the electric power communication network includes the following steps: Step S1: Constructing a database of electric power communication service traffic characteristics, defining bandwidth thresholds, maximum allowable delays, and reliability indicators for various services, and predicting service traffic fluctuations; Step S2: Real-time acquisition of bandwidth occupancy, interference intensity, delay jitter, and bit error rate parameters for each link to construct a network status matrix; Step S3: Generating a bandwidth allocation plan based on the service demand prediction model and network status, with the optimization objectives of maximizing bandwidth utilization, ensuring key service guarantees, and minimizing network delay. A mathematical model is established that includes service demand constraints, bandwidth resource constraints, and interference avoidance constraints; Step S4: Executing the allocation strategy, dynamically adjusting service bandwidth allocation, collecting actual transmission indicators and comparing them with model predictions, optimizing model parameters, and forming a closed-loop optimization mechanism. The method first models service characteristics and predicts traffic, then obtains real-time network status, combines the two to build a model, generates an allocation plan, and performs optimization. This allows bandwidth to be dynamically allocated based on service demand and network status, avoiding resource waste, improving network transmission efficiency, and ensuring the stable transmission of key services. In this embodiment, the LSTM-GARCH hybrid neural network model is used to predict traffic fluctuations in step S1, achieving a prediction error rate of ≤10%. Leveraging the powerful time series data processing capabilities of the LSTM-GARCH hybrid neural network model, historical traffic data is analyzed and learned to predict future traffic fluctuations. This allows for accurate understanding of traffic trends, providing a reliable forecast basis for bandwidth allocation and ensuring that bandwidth allocation is more aligned with actual needs.

[0021] In this embodiment, real-time parameter acquisition in step S2 is achieved by deploying distributed sensing probes at power communication base stations, aggregation nodes, and key terminals. These probes preprocess data via edge computing nodes, with a preprocessing latency of ≤20ms. Distributed sensing probes collect link status information in real time, and edge computing nodes rapidly preprocess the data, reducing data transmission volume and computing pressure on central nodes. This enables timely acquisition of accurate network status data, improving the real-time and accuracy of network status perception and providing timely and effective information for bandwidth allocation decisions.

[0022] In this embodiment, the bandwidth allocation plan generated in step S3 uses a deep reinforcement learning algorithm based on proximal policy optimization, with a convergence time of ≤1000 training rounds. This algorithm uses network status and service requirements as input and continuously interacts with the network environment to find the optimal bandwidth allocation strategy. This allows the automatic generation of optimal bandwidth allocation plans in complex and changing network environments, improving bandwidth utilization and network performance.

[0023] In this embodiment, the model parameters are optimized in step S4 using a gradient descent method. This method continuously adjusts the parameters of the prediction model and optimization algorithm based on the difference between actual transmission metrics and the model's predicted values, making the model and algorithm more accurate. This allows the system to continuously adapt to network changes, continuously optimize bandwidth allocation, and improve network stability and reliability.

[0024] In this embodiment, a closed-loop optimization mechanism is established in step S4, including the establishment of an expert knowledge base. Manual intervention is triggered when actual transmission metrics deviate from predicted values ​​by more than a threshold. The expert knowledge base stores historical experience and optimization strategies. When the system experiences significant deviations, manual intervention is combined with the knowledge base content to perform adjustments and optimizations. This allows for timely intervention when system anomalies occur, improving the system's ability to cope with complex situations and ensuring the effectiveness of the bandwidth allocation plan.

[0025] This paper proposes a closed-loop dynamic bandwidth allocation method based on "perception-prediction-optimization-control." By integrating power service traffic feature analysis, real-time network status perception, and intelligent optimization algorithms, this method achieves efficient utilization of bandwidth resources. The method first establishes a database of power communication service traffic characteristics and quantitatively models the bandwidth requirements, delay sensitivity, and reliability indicators of different service types. Second, a network status perception module acquires parameters such as bandwidth utilization, interference intensity, and delay jitter for each link in real time. Then, based on the service demand prediction model and network status, an improved deep reinforcement learning algorithm is used to generate an optimal bandwidth allocation plan. Finally, a software-defined network (SDN) controller executes the allocation strategy and optimizes model parameters based on actual transmission performance, forming a closed-loop optimization mechanism. This method constructs a dynamic bandwidth allocation model that integrates the real-time and reliability requirements of power services, accurately matching bandwidth resources with service demands and improving resource utilization to over 60%. An adaptive allocation algorithm based on service priority and traffic prediction is designed to ensure a minimum bandwidth limit for critical services such as relay protection while improving resource utilization for non-critical services. A real-time network status perception and interference avoidance mechanism has been established, dynamically adjusting bandwidth allocation strategies to address dynamic changes in electromagnetic interference, terminal access, and other factors. Furthermore, through quantitative analysis of power communication service traffic patterns and bandwidth usage characteristics, global optimization of bandwidth resources across the entire network is achieved.

[0026] Example 2: As shown in the attached Figure 2 As shown, this embodiment provides a dynamic bandwidth allocation system for a power communication network, including: The business modeling module is used to build a database of power communication business traffic characteristics, define the bandwidth threshold, maximum allowable delay and reliability index of each type of business, and predict business traffic fluctuations; divide power communication business into real-time type (relay protection, fault location, bandwidth requirement 2-4Mbps, delay ≤5ms), quasi-real-time type (distribution network automation, bandwidth requirement 1-2Mbps, delay ≤50ms) and non-real-time type (electricity consumption information collection, bandwidth requirement 0.5-1Mbps, delay ≤1s), define the bandwidth threshold, maximum allowable delay and reliability index (such as bit error rate ≤10⁻ 6 A dynamic demand forecasting model was established based on historical business traffic data (minute-level traffic over the past three months). An LSTM-GARCH hybrid neural network was used to predict business traffic fluctuations over the next 15 minutes, capturing burst traffic characteristics during peak electricity consumption periods (e.g., 8:00 AM to 10:00 AM). The prediction error rate was ≤10%.

[0027] The network perception module is used to obtain real-time information about bandwidth utilization, interference intensity, delay jitter, and bit error rate parameters for each link to construct a network status matrix. It also collects multi-dimensional perception indicators, including link bandwidth utilization (the percentage of bandwidth occupied per unit time), electromagnetic interference intensity (dBm), delay jitter (ms), and bit error rate, to construct a network status matrix. A distributed perception network is established, with perception probes deployed at power communication base stations, aggregation nodes, and key terminals. Edge computing nodes preprocess the perception data (such as downsampling and denoising), reducing the computing pressure on central nodes and ensuring perception latency of ≤20ms.

[0028] The intelligent optimization module generates bandwidth allocation plans based on the service demand prediction model and network status. The optimization objectives are maximizing bandwidth utilization, ensuring critical services, and minimizing network latency. A mathematical model is established that incorporates service demand constraints, bandwidth resource constraints, and interference avoidance constraints. A multi-objective optimization model is also established, with the optimization objectives of maximizing bandwidth utilization (objective function 1), ensuring critical services (objective function 2, such as achieving ≥99.9% bandwidth compliance for relay protection services), and minimizing network latency (objective function 3). A deep reinforcement learning framework and a bandwidth allocation algorithm based on Proximal Policy Optimization (PPO) are designed. The algorithm takes the network state matrix and service demand vector as input and outputs bandwidth allocation plans and priority weights for each service. Through interactive training with the power communication network environment, the algorithm adapts to dynamic changes in service traffic, achieving convergence in ≤1000 training rounds.

[0029] The execution control module is responsible for executing allocation policies, dynamically adjusting service bandwidth allocation, collecting actual transmission metrics and comparing them with model predictions, optimizing model parameters, and forming a closed-loop optimization mechanism. A real-time control mechanism dynamically adjusts bandwidth allocation for each service through the SDN controller, supporting millisecond-level bandwidth slice switching (for example, when a relay protection service burst occurs, bandwidth can be borrowed from non-real-time services within 5ms). Performance evaluation and parameter optimization are performed, collecting actual transmission metrics (such as service bandwidth utilization, latency, and bit error rate) and comparing them with model predictions. Gradient descent is used to optimize the parameters of the prediction model and optimization algorithm, with the model updated every 10 minutes.

[0030] In this embodiment, the service modeling module includes a service classification unit, which is used to classify services and define bandwidth thresholds and maximum allowable latency for each type of service; and a traffic prediction unit, which is used to predict service traffic fluctuations. The service classification unit classifies services and determines metrics, while the traffic prediction unit uses data to predict traffic flow, providing a basis for subsequent allocation. This allows the demand characteristics of different services to be clearly defined, creating the prerequisite for rational bandwidth allocation.

[0031] In this embodiment, the network perception module includes distributed perception probes deployed at power communication base stations, aggregation nodes, and key terminals to collect link bandwidth utilization, electromagnetic interference intensity, latency jitter, and bit error rate parameters; and edge computing nodes to preprocess the perception data and construct a network status matrix. The distributed perception probes collect data, and the edge computing nodes preprocess and construct the matrix, enabling real-time perception of the network status. This allows for comprehensive and timely acquisition of network operating status information, providing an accurate basis for bandwidth allocation.

[0032] In this embodiment, the execution control module includes a software-defined network controller (SDN) for implementing millisecond-level bandwidth slice switching and a parameter optimization unit for updating prediction models and optimizing algorithm parameters. The SDN controller rapidly executes allocation policies, while the parameter optimization unit optimizes parameters based on actual conditions. This ensures rapid execution of bandwidth allocation policies and enables continuous system optimization to adapt to network changes.

[0033] The dynamic bandwidth allocation system for the electric power communication network in this embodiment adopts a five-layer distributed architecture, integrating edge computing, SDN, and intelligent decision-making technologies. Specifically, it includes: Physical perception layer. (1) Core function: Real-time collection of traffic data and link status of the power communication network to build a physical layer perception network. (2) Hardware components: Traffic monitoring probes, deployed at the communication interfaces of substations and distribution rooms, use dedicated network chips (such as the Intel 800 series) to achieve traffic collection for links below 10Gbps, support the IPFIX protocol, and collect indicators including: source / destination IP, service type (based on DPI deep packet inspection), bandwidth occupancy, and delay jitter. Electromagnetic interference sensor, using a broadband electromagnetic detection module (detection range 30MHz-3GHz), monitors the electromagnetic interference spectrum generated by substation equipment (such as transformers and circuit breakers) in real time, outputs interference power spectrum density, and has a resolution of ≤100kHz. Synchronous clock module, using GPS / Beidou timing, ensures that the time synchronization accuracy of the entire network perception data is ≤100μs, meeting the stringent time synchronization requirements of the power business.

[0034] Edge processing layer. (1) Core functions: Implement local preprocessing, feature extraction, and edge decision-making of sensor data to reduce the load on the central node. (2) Key modules: Data preprocessing unit, using FPGA (such as Xilinx UltraScale+) to implement traffic data filtering (eliminating power frequency interference), aggregation (statistical bandwidth requirements by service type), and compression (using LZ4 algorithm, compression ratio 8:1), with data processing delay ≤ 5ms. Service feature extractor, based on edge servers (such as NVIDIA Jetson AGX Xavier), analyzes real-time traffic, extracts burst traffic features (such as pulsed traffic of relay protection services), and delay sensitivity levels (calculating the quantile of message transmission delay through sliding windows). Local decision engine, preset emergency response rules (such as triggering the local bandwidth preemption mechanism when detecting relay protection service delay ≥ 5ms), supports millisecond-level emergency decision-making, and avoids service failure caused by central node communication interruption.

[0035] Central control layer. (1) Core function: Build a global bandwidth resource management and intelligent optimization center to achieve cross-regional resource collaborative allocation. (2) System composition: Bandwidth resource database, using a time series database (InfluxDB) to store historical traffic data (time granularity 1s) and business demand logs (minute-level sampling), supporting queries based on time and space dimensions (such as querying the bandwidth usage pattern of a substation during peak power consumption). Intelligent optimization engine, deploying a GPU server cluster (such as NVIDIAA100×8), running the PPO algorithm model, input includes: global network state matrix (dimension is the number of links × number of monitoring nodes), cross-regional business demand vector (integrating the real-time business priority of each substation), historical allocation strategy effect evaluation (such as the bandwidth utilization curve of the past 24 hours). SDN controller cluster, using ONOS controller, supports OpenFlow1.5 protocol, realizes bandwidth slicing and priority queue configuration of switches, and the control instruction transmission delay is ≤30ms. (3) Algorithm parameters: State space: network status indicators (bandwidth occupancy, interference intensity) + service demand parameters (bandwidth, latency), a total of 15 dimensions; Action space: bandwidth allocation plan (10 service types × 5 bandwidth levels), a total of 50 dimensions; Reward function: comprehensive bandwidth utilization (weight 0.4), key service guarantee rate (weight 0.5), network latency (weight 0.1).

[0036] Application management layer. (1) Core functions: Provides human-computer interaction interface, policy configuration and system operation and maintenance management, and supports visual monitoring of power communication services. (2) Functional modules: Visual monitoring platform, based on ECharts to develop a real-time monitoring interface, displaying bandwidth heat map (bandwidth occupancy of different services is displayed in color gradient), business traffic topology map (real-time flow of data between substations), and optimization strategy execution effect curve (such as bandwidth utilization in the past hour and key business latency). Policy configuration center, supports operation and maintenance personnel to customize business priority rules (such as setting the lower limit of relay protection business bandwidth to 2Mbps) and bandwidth allocation constraints (such as non-real-time business bandwidth ratio ≤30%). Fault diagnosis module, based on historical data training anomaly detection model (such as isolation forest algorithm), automatically identifies bandwidth allocation anomalies (such as continuous congestion of a link causing business interruption), and generates fault location report (locating to specific switch or business type).

[0037] Execution and scheduling layer. (1) Core function: Implement the physical execution of bandwidth allocation strategies, including switch configuration updates and service traffic scheduling. (2) Hardware module: Smart switches, using industrial-grade SDN switches (such as Arista7050X), support flow table-based bandwidth slicing (accuracy ≤100kbps) and priority queues (SP / WRR scheduling algorithm, supporting 8 priority queues). Service scheduling gateway, integrated into the aggregation node, supports traffic shaping for different service types (such as rate limiting for non-real-time services) and QoS marking (based on the DSCP field) to ensure that real-time services are forwarded first. (3) Control process: The central control layer issues allocation instructions (such as allocating 3Mbps bandwidth to the relay protection service of a substation, priority queue 1); the smart switch updates the flow table, marks the service traffic as high priority, and limits the bandwidth upper limit of other services; the edge processing layer monitors the service indicators after control in real time. If the bandwidth is insufficient (such as the actual available bandwidth <2.5Mbps), the local reallocation mechanism is triggered (with a higher priority than the non-emergency instructions of the central layer).

[0038] The interactive process of the dynamic bandwidth allocation system of the power communication network in this embodiment is as follows: (1) Real-time perception stage: The physical perception layer collects traffic and interference data every 50ms, and the edge processing layer simultaneously analyzes the service characteristics and generates a feature vector (including 8 network indicators + 4 service parameters). (2) Prediction and optimization stage: The edge layer uploads the feature vector to the central layer every 1s. The central layer combines the future traffic predicted by LSTM and inputs the PPO model to generate an allocation plan. The calculation time is ≤150ms. (3) Execution feedback stage: The execution control layer completes the switch configuration update within 10ms, the edge layer continuously monitors the effect, and feeds back the control results to the central layer every 200ms. (4) Closed-loop optimization stage: The central layer updates the model parameters based on the feedback data, adopts the experience replay mechanism (caches 500,000 historical decision data), and performs model iterative training once a day.

[0039] The present invention innovatively combines the delay sensitivity and reliability requirements of the power business with the bandwidth allocation model to establish a business demand-bandwidth resource mapping relationship, breaking through the allocation model that only considers bandwidth capacity in the traditional communication field. The present invention is based on an intelligent optimization algorithm based on deep reinforcement learning, and introduces the PPO framework to achieve adaptive optimization of bandwidth allocation, so that the system can adapt to the dynamic changes in traffic in the power communication network through autonomous learning, without the need for manual pre-setting of complex rules. The present invention designs an anti-interference allocation mechanism under multi-dimensional constraints, introduces electromagnetic interference characteristic parameters into the optimization model, and reduces the impact of interference on power communications through bandwidth avoidance and priority adjustment strategies, thereby ensuring the transmission quality of key services.

[0040] Compared with the existing technology, the advantages of the present invention are as follows: (1) Significant improvement in bandwidth utilization: By dynamically adapting service traffic, the bandwidth utilization of the entire network can be increased by more than 40% compared with the traditional static allocation method, and the advantage is more obvious during peak power consumption periods (for example, during the emergency repair of distribution network automation faults, the bandwidth utilization of burst data transmission is increased to 80%). (2) Enhanced key business guarantee capability: For real-time services such as relay protection, the service bandwidth compliance rate is increased from 95% of the traditional method to 99.99% through a priority-driven bandwidth reservation mechanism, meeting the safe operation requirements of the power system. (3) Improved dynamic adaptability: The system can respond to dynamic changes such as terminal batch access and sudden increase in service traffic within 50ms, which is more than 100 times faster than manual adjustment, effectively avoiding network congestion. (4) High level of intelligence: Without manual intervention, the system can autonomously learn the power service traffic pattern and bandwidth usage rules, adapt to the access requirements of new services such as 5G slicing and Internet of Things terminals in the smart grid, and reduce operation and maintenance costs by more than 30%.

[0041] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Non-essential technical features can be added or removed according to actual needs to meet the requirements of different situations.

Claims

1. A method for dynamic bandwidth allocation in a power communication network, characterized in that: The following steps are involved: Step S1: Build a traffic feature database for electric power communication services, define bandwidth thresholds, maximum allowable delays, and reliability indicators for various services, and predict service traffic fluctuations; Step S2: Obtain the bandwidth occupancy rate, interference intensity, delay jitter, and bit error rate parameters of each link in real time to build a network status matrix; Step S3: Generate a bandwidth allocation plan based on the service demand prediction model and network status, with the optimization goals of maximizing bandwidth utilization, ensuring key service rates, and minimizing network latency. Establish a mathematical model that includes service demand constraints, bandwidth resource constraints, and interference avoidance constraints. Step S4: Execute the allocation strategy, dynamically adjust the service bandwidth allocation, collect actual transmission indicators and compare them with the model prediction values, optimize the model parameters, and form a closed-loop optimization mechanism.

2. The method for dynamic bandwidth allocation of a power communication network according to claim 1, characterized in that: In step S1, the business traffic fluctuation is predicted using an LSTM-GARCH hybrid neural network model with a prediction error rate of ≤10%.

3. The method for dynamic bandwidth allocation of a power communication network according to claim 1 or 2, characterized in that: The real-time acquisition of parameters in step S2 is achieved by deploying distributed sensing probes at power communication base stations, aggregation nodes and key terminals. The probes preprocess the data through edge computing nodes, and the preprocessing delay is ≤20ms.

4. The method for dynamic bandwidth allocation of a power communication network according to claim 1 or 2, characterized in that: In step S3, a bandwidth allocation scheme is generated by adopting a deep reinforcement learning algorithm based on proximal policy optimization, and the convergence time of the algorithm is ≤ 1000 rounds of training.

5. The method for dynamic bandwidth allocation of a power communication network according to claim 1 or 2, characterized in that: In step S4, the model parameters are optimized using a gradient descent method.

6. The method for dynamic bandwidth allocation of a power communication network according to claim 1 or 2, characterized in that: In step S4, a closed-loop optimization mechanism is formed, including establishing an expert knowledge base, and triggering manual intervention when the actual transmission index deviates from the predicted value by more than a threshold.

7. A dynamic bandwidth allocation system for an electric power communication network, characterized in that: include: The service modeling module is used to build a database of power communication service traffic characteristics, define bandwidth thresholds, maximum allowable delays, and reliability indicators for various services, and predict service traffic fluctuations; The network perception module is used to obtain the bandwidth occupancy rate, interference intensity, delay jitter, and bit error rate parameters of each link in real time and build a network status matrix; The intelligent optimization module generates a bandwidth allocation plan based on the business demand prediction model and network status. It sets the optimization goals of maximizing bandwidth utilization, ensuring key business performance, and minimizing network latency. It establishes a mathematical model that includes business demand constraints, bandwidth resource constraints, and interference avoidance constraints. The execution control module is used to execute the allocation strategy, dynamically adjust the service bandwidth allocation, collect actual transmission indicators and compare them with the model prediction values, optimize the model parameters, and form a closed-loop optimization mechanism.

8. The dynamic bandwidth allocation system for electric power communication network according to claim 7, characterized in that: The business modeling module includes: The service classification unit is used to classify services and define the bandwidth threshold and maximum allowable delay for each type of service; The traffic prediction unit is used to predict business traffic fluctuations.

9. The dynamic bandwidth allocation system for the electric power communication network according to claim 7 or 8, characterized in that: The network perception module includes: Distributed sensing probes are deployed at power communication base stations, aggregation nodes, and key terminals to collect link bandwidth occupancy, electromagnetic interference intensity, delay jitter, and bit error rate parameters; Edge computing nodes are used to pre-process the perception data and build the network status matrix.

10. The dynamic bandwidth allocation system for electric power communication network according to claim 7 or 8, characterized in that: The execution control module includes: Software-defined network controller for millisecond-level bandwidth slice switching; Parameter optimization unit, used to update the prediction model and optimize algorithm parameters.

Images