Quick start network congestion control method suitable for power system

Abstract

The invention belongs to the technical field of power systems, and more particularly relates to a quick-start network congestion control method suitable for a power system, which ensures the accuracy and reliability of data by collecting and preprocessing power equipment and network flow data. Then, a traffic prediction model is trained by using the preprocessed data, and the model can accurately predict the network traffic in a period of time in the future, thereby providing a basis for real-time decision making. In the starting process of the power system, network congestion detection is carried out in combination with real-time network state and equipment state data, and potential congestion risks are found in time. And finally, dynamically adjusting a flow scheduling and control strategy based on a network congestion detection result and a prediction result of the flow prediction model, thereby optimizing resource allocation and avoiding an improper decision. The series of steps ensure that network congestion can be timely and accurately dealt with under the condition that the network state is rapidly changed, and the efficiency and the stability of rapid starting of the power system are improved.

Description

Technical Field

[0001] The present application belongs to the technical field of power systems, and more specifically, to a fast-start network congestion control method applicable to power systems. Background Art

[0002] With the rapid development of my country's economy, the scale of the power system continues to expand and the grid structure becomes increasingly complex. In order to meet the growing demand for electricity, the rapid startup capability of the grid has become particularly important. Rapid startup means that when a grid failure occurs or maintenance is required, power supply can be quickly restored, reducing power outage time and improving the reliability and stability of the grid.

[0003] The current development trend of the power grid is highly intelligent, automated and digital. The application of rapid start technology can effectively respond to emergencies in power grid operation, such as equipment failure, line short circuit, etc., thereby ensuring the continuity and stability of power supply.

[0004] In the existing rapid start-up technology, the network congestion control of the power system usually adopts the method of manually or automatically adjusting the network resource allocation according to the degree of network congestion to alleviate the congestion; generally, the heuristic algorithm is adopted. However, the heuristic scheduling algorithm usually relies on manual experience or preset rules and lacks real-time and adaptability. In the rapid start-up phase, the rapid changes in network status and equipment status make it difficult for the heuristic algorithm to respond effectively, resulting in inappropriate resource allocation decisions. Summary of the invention

[0005] The present invention provides a fast-start network congestion control method applicable to an electric power system, and aims to solve the technical problem that in the fast-start phase, the rapid changes in network status and device status make it difficult for heuristic algorithms to effectively cope with them, resulting in inappropriate resource allocation decisions.

[0006] A fast-start network congestion control method applicable to a power system comprises the following steps:

[0007] Step 1: Collect monitoring data and network flow data of power equipment, and perform preprocessing based on the collected detection data and network flow data to obtain preprocessed detection data and network flow data of power equipment;

[0008] Historical network traffic data: including historical time series data such as bandwidth utilization, packet loss rate, and link delay.

[0009] Power load data: such as voltage, current, power changes, etc., used as external features (for example, can affect network traffic changes during certain time periods).

[0010] Step 2: Train the traffic prediction model based on the pre-processed power equipment detection data and network traffic data to obtain a trained traffic prediction model;

[0011] Step 3: Use the trained traffic prediction model and perform traffic prediction based on real-time data to predict the network traffic in a certain period of time in the future, and obtain the network traffic prediction results in the future period of time;

[0012] Step 4: During the startup of the power system, network congestion detection is performed based on real-time network status data and equipment status data of the power system;

[0013] Step 5: Schedule and control traffic based on the results of network congestion detection and the prediction results of the traffic prediction model, and output the scheduling and control strategy.

[0014] The present invention effectively solves the challenge posed to the heuristic algorithm by the rapid changes of network status and equipment status in the rapid startup phase by constructing a rapid startup network congestion control method suitable for the power system. The method first collects and preprocesses the power equipment and network flow data to ensure the accuracy and reliability of the data. Subsequently, the flow prediction model is trained using these preprocessed data, and the model can accurately predict the network flow in the future period of time, providing a basis for real-time decision-making. During the startup of the power system, network congestion detection is performed in combination with real-time network status and equipment status data to timely discover potential congestion risks. Finally, based on the results of network congestion detection and the prediction results of the flow prediction model, the flow scheduling and control strategy are dynamically adjusted to optimize resource allocation and avoid inappropriate decisions. This series of steps ensures that network congestion can be dealt with in a timely and accurate manner in the case of rapid changes in network status, thereby improving the efficiency and stability of the rapid startup of the power system.

[0015] Preferably, the input layer is used to receive input data;

[0016] The LSTM layer is used to process the input data in the input layer, including a first LSTM layer and a second LSTM layer. The first LSTM layer processes the input data in the input layer and passes the output result to the second LSTM layer. The second LSTM layer processes the data and passes it to the attention mechanism layer.

[0017] The attention mechanism layer selectively focuses on important time periods to improve the prediction accuracy of the model;

[0018] The output layer includes a fully connected layer, which maps the high-dimensional features output by the LSTM layer to the final prediction result, represents the predicted value of network traffic at a certain point in the future, and uses a linear activation function to output continuous prediction values.

[0019] Preferably, the network congestion detection method is as follows:

[0020] Bandwidth utilization calculation:

[0021]

[0022] Where: U b (t) represents the bandwidth utilization at time t; C b (t) represents the bandwidth usage at time t; B b Indicates the total bandwidth of the link;

[0023] Link delay:

[0024]

[0025] Where: D(t) represents the average link delay at time t; (d i (t) represents the delay of the ith data packet at time t; d0 represents the normal delay reference value; N represents the number of observed data packets;

[0026] Packet loss rate:

[0027]

[0028] Where: P(t) represents the packet loss rate at time t; L(t) represents the number of packets lost at time t; T(t) represents the total number of packets sent at time t;

[0029] The real-time calculated bandwidth utilization, link delay and packet loss rate are compared with the set thresholds. If any one of the bandwidth utilization, link delay and packet loss rate is greater than the corresponding preset threshold, it is judged as link congestion.

[0030] Preferably, the specific steps of scheduling and controlling traffic based on the result of network congestion detection are as follows:

[0031] Prioritization: Prioritize different types of traffic in advance, where the priority classification is based on business importance, delay sensitivity, and bandwidth requirements.

[0032] Bandwidth resource allocation and traffic redirection: When network congestion is detected, a traffic redirection strategy is used to transfer part of the traffic to alternative paths or network links in space; based on network topology and traffic prediction, bandwidth resource allocation is optimized so that traffic is evenly distributed across multiple links;

[0033] Traffic speed limit and packet loss control: Limit the speed of low-priority traffic and discard some low-priority traffic packets in advance through the random early detection algorithm; dynamically adjust the congestion window size of the TCP connection to control the sending rate of data traffic.

[0034] Preferably, the specific steps of the priority division are as follows:

[0035] Calculation of business importance score I(t):

[0036]

[0037] Where: importance i represents the business importance score of flow i; max_importancei represents the maximum importance score of all flows;

[0038] Delay sensitivity score D(t) calculation:

[0039]

[0040] Where: delay_sensitivity i Indicates the sensitivity of flow i to delay; max_delay_sensitivity indicates the delay sensitivity of the flow that is most sensitive to delay among all flows;

[0041] Bandwidth demand score B(t) calculation:

[0042]

[0043] Where: bandwidth_need i Indicates the bandwidth requirement of flow i; max_bandwidth_need indicates the maximum bandwidth requirement of all flows;

[0044] Based on the business importance score, delay sensitivity score, and bandwidth requirement score, a priority score is calculated:

[0045] P score (t)=w1·I(t)+w2·D(t)+w3·B(t);

[0046] Wherein: w1, w2 and w3 all represent weighting coefficients; w1+w2+w3=1;

[0047] Based on the calculated priority score P score (t) Priority classification.

[0048] Preferably, the specific steps of bandwidth resource allocation and traffic redirection are as follows:

[0049] Dynamic optimization of bandwidth resource allocation: Calculate the bandwidth allocation for each link based on traffic priority and bandwidth requirements:

[0050]

[0051] Where: B k,i represents the bandwidth allocated to flow i on link k; B i represents the total bandwidth requirement of flow i; U k represents the current bandwidth utilization of link k; N represents the number of links in the network;

[0052] Traffic redirection strategy: When the bandwidth utilization of a link exceeds the set threshold U threshold When the traffic is redirected to other backup link B j,i :

[0053] B j,i ′=B j,i +δ·B k,i ;

[0054] Where: δ represents the redirection coefficient; B j,i ′ represents the bandwidth of traffic i on link j after redirection; B j,i represents the bandwidth of flow i on link j before redirection; the amount of redirected bandwidth is determined by the priority of the flow and the current bandwidth demand of the link;

[0055] Topology-aware bandwidth allocation: Optimizes bandwidth resource allocation based on traffic prediction results and current network topology:

[0056]

[0057] Where: represents the bandwidth demand forecast of flow i at the future time t+Δt, obtained by the flow prediction model; B optimized Indicates the optimized bandwidth allocation.

[0058] Preferably, the specific steps of the flow rate limiting and packet loss control are as follows:

[0059] Traffic speed limit:

[0060] R adjusted =R current ·(1―α·U);

[0061] Where: R adjusted Represents the adjusted traffic sending rate; R current represents the sending rate of the current traffic; α represents the traffic adjustment coefficient; U represents the bandwidth utilization of the current link;

[0062] Packet loss control:

[0063]

[0064] Where: P drop Indicates the probability of packet loss; Q avgIndicates the average queue length; maxth indicates the maximum packet loss threshold; minth indicates the minimum packet loss threshold;

[0065] Congestion window adjustment:

[0066]

[0067] Where: MSS represents the maximum segment size, that is, the maximum payload of each TCP data packet; cwnd old Indicates the current congestion window size; cwnd new Indicates the adjusted congestion window size.

[0068] The beneficial effects of the present invention include:

[0069] The present invention effectively solves the challenge posed to the heuristic algorithm by the rapid changes of network status and equipment status in the rapid startup phase by constructing a rapid startup network congestion control method suitable for the power system. The method first collects and preprocesses the power equipment and network flow data to ensure the accuracy and reliability of the data. Subsequently, the flow prediction model is trained using these preprocessed data, and the model can accurately predict the network flow in the future period of time, providing a basis for real-time decision-making. During the startup of the power system, network congestion detection is performed in combination with real-time network status and equipment status data to timely discover potential congestion risks. Finally, based on the results of network congestion detection and the prediction results of the flow prediction model, the flow scheduling and control strategy are dynamically adjusted to optimize resource allocation and avoid inappropriate decisions. This series of steps ensures that network congestion can be dealt with in a timely and accurate manner in the case of rapid changes in network status, thereby improving the efficiency and stability of the rapid startup of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0071] Figure 1 An overall step block diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0073] See also Figure 1 As shown, the best embodiment of the present invention is further described;

[0074] A fast-start network congestion control method applicable to a power system comprises the following steps:

[0075] Step 1: Collect monitoring data and network flow data of power equipment, and perform preprocessing based on the collected detection data and network flow data to obtain preprocessed detection data and network flow data of power equipment;

[0076] Step 2: Train the traffic prediction model based on the pre-processed power equipment detection data and network traffic data to obtain a trained traffic prediction model;

[0077] Step 3: Use the trained traffic prediction model and perform traffic prediction based on real-time data to predict the network traffic in a certain period of time in the future, and obtain the network traffic prediction results in the future period of time;

[0078] The prediction module includes an input layer, an LSTM layer, an attention mechanism layer and an output layer;

[0079] Preferably, the prediction module includes an input layer, an LSTM layer, an attention mechanism layer and an output layer;

[0080] Input Layer:

[0081] The input data include: historical network traffic data (bandwidth utilization, packet loss rate, link delay) and external power data (such as voltage, current) as well as time characteristics.

[0082] Input size: Set the data of each time step to X_t = [flow data, power data, time feature], where the dimension of X_t is D, representing a multidimensional feature vector.

[0083] LSTM Layer:

[0084] LSTM1: This is the first LSTM layer, which receives the time step data of the previous step and processes the long-term dependencies of historical traffic. The output of LSTM will be a hidden state vector of dimension H1, capturing the dynamic changes of network traffic.

[0085] LSTM2: Based on LSTM1, a second layer of LSTM can be added to further improve the model's ability to capture complex temporal relationships. The output dimension of the second layer of LSTM is H2.

[0086] The hidden state h_t and cell state c_t of each LSTM layer are updated at each time step to capture the relationship between long and short-term memory.

[0087] An attention mechanism is added after the LSTM layer so that the model can focus on the time step features that are most important for flow prediction. The attention mechanism can dynamically weight the impact of different time steps based on historical flow data and power system status.

[0088] Through the attention mechanism, the model can selectively focus on important time periods, thereby improving prediction accuracy.

[0089] The output layer includes a fully connected layer, which maps the high-dimensional features output by the LSTM layer to the final prediction result, that is, the predicted value of network traffic at a certain point in the future (such as bandwidth demand, packet loss rate, etc.), and uses a linear activation function to output continuous prediction values.

[0090] Step 4: During the startup of the power system, network congestion detection is performed based on real-time network status data and equipment status data of the power system;

[0091] The network congestion detection method is as follows:

[0092] Bandwidth utilization calculation:

[0093]

[0094] Where: U b (t) represents the bandwidth utilization at time t; C b (t) represents the bandwidth usage at time t; B b Indicates the total bandwidth of the link;

[0095] Link delay:

[0096]

[0097] Where: D(t) represents the average link delay at time t; (d i (t) represents the delay of the ith data packet at time t; d0 represents the normal delay reference value; N represents the number of observed data packets;

[0098] Packet loss rate:

[0099]

[0100] Where: P(t) represents the packet loss rate at time t; L(t) represents the number of packets lost at time t; T(t) represents the total number of packets sent at time t;

[0101] The bandwidth utilization, link delay and packet loss rate calculated in real time are compared with the set thresholds. If any one of the bandwidth utilization, link delay and packet loss rate is greater than the corresponding preset threshold, it is determined that the link is congested.

[0102] Step 5: Schedule and control traffic based on the results of network congestion detection and the prediction results of the traffic prediction model, and output the scheduling and control strategy.

[0103] Preferably, the specific steps of scheduling and controlling traffic based on the result of network congestion detection are as follows:

[0104] Prioritization: Prioritize different types of traffic in advance, where the priority classification is based on business importance, delay sensitivity, and bandwidth requirements.

[0105] Bandwidth resource allocation and traffic redirection: When network congestion is detected, a traffic redirection strategy is used to transfer part of the traffic to alternative paths or network links in space; based on network topology and traffic prediction, bandwidth resource allocation is optimized so that traffic is evenly distributed across multiple links;

[0106] Traffic speed limit and packet loss control: Limit the speed of low-priority traffic and discard some low-priority traffic packets in advance through the random early detection algorithm; dynamically adjust the congestion window size of the TCP connection to control the sending rate of data traffic.

[0107] The entire execution sequence of frame resource allocation and traffic redirection, traffic speed limiting and packet loss control dynamically adjusts the execution sequence and parallelism of these steps according to the real-time performance and response requirements of the system.

[0108] For example, traffic rate limiting and packet loss control are performed after bandwidth resource allocation and traffic redirection to ensure that network congestion is effectively alleviated. Traffic rate limiting targets low-priority traffic and reduces network congestion by limiting its sending rate. Packet loss control is usually implemented through a random early detection (RED) algorithm, which discards some low-priority traffic packets in advance to avoid further network congestion. These two sub-steps can be performed in parallel because traffic rate limiting and packet loss control are independent operations and can be performed simultaneously for traffic of different priorities.

[0109] Preferably, the specific steps of the priority division are as follows:

[0110] Calculation of business importance score I(t):

[0111]

[0112] Where: importance i Indicates the business importance score of flow i; max_importance indicates the maximum importance score of all flows;

[0113] Delay sensitivity score D(t) calculation:

[0114]

[0115] Where: delay_sensitivity i Indicates the sensitivity of flow i to delay; max_delay_sensitivity indicates the delay sensitivity of the flow that is most sensitive to delay among all flows;

[0116] Bandwidth demand score B(t) calculation:

[0117]

[0118] Where: bandwidth_need i Indicates the bandwidth requirement of flow i; max_bandwidth_need indicates the maximum bandwidth requirement of all flows;

[0119] Based on the business importance score, delay sensitivity score, and bandwidth requirement score, a priority score is calculated:

[0120] P score (t)=w1·I(t)+w2·D(t)+w3·B(t);

[0121] Wherein: w1, w2 and w3 all represent weighting coefficients; w1+w2+w3=1;

[0122] Based on the calculated priority score P score (t) Priority classification.

[0123] Preferably, the specific steps of bandwidth resource allocation and traffic redirection are as follows:

[0124] Dynamic optimization of bandwidth resource allocation: Calculate the bandwidth allocation for each link based on traffic priority and bandwidth requirements:

[0125]

[0126] Where: B k,i represents the bandwidth allocated to flow i on link k; B i represents the total bandwidth requirement of flow i; U k represents the current bandwidth utilization of link k; N represents the number of links in the network;

[0127] Traffic redirection strategy: When the bandwidth utilization of a link exceeds the set threshold U threshold When the traffic is redirected to other backup link B j,i :

[0128] B j,i ′=Bj,i +δ·B k,i ;

[0129] Where: δ represents the redirection coefficient; B j,i ′ represents the bandwidth of traffic i on link j after redirection; B j,i represents the bandwidth of flow i on link j before redirection; the amount of redirected bandwidth is determined by the priority of the flow and the current bandwidth demand of the link;

[0130] Topology-aware bandwidth allocation: Optimizes bandwidth resource allocation based on traffic prediction results and current network topology:

[0131]

[0132] Where: represents the bandwidth demand forecast of flow i at the future time t+Δt, obtained by the flow prediction model; B optimized Indicates the optimized bandwidth allocation.

[0133] Preferably, the specific steps of the flow rate limiting and packet loss control are as follows:

[0134] Traffic speed limit:

[0135] R adjusted =R current ·(1―α·U);

[0136] Where: R adjusted Represents the adjusted traffic sending rate; R current represents the sending rate of the current traffic; α represents the traffic adjustment coefficient; U represents the bandwidth utilization of the current link;

[0137] Packet loss control:

[0138]

[0139] Where: P drop Indicates the probability of packet loss; Q avg Indicates the average queue length; maxth indicates the maximum packet loss threshold; minth indicates the minimum packet loss threshold;

[0140] Congestion window adjustment:

[0141]

[0142] Where: MSS represents the maximum segment size, that is, the maximum payload of each TCP data packet; cwnd old Indicates the current congestion window size; cwnd new Indicates the adjusted congestion window size.

[0143] The present invention effectively solves the challenge posed to the heuristic algorithm by the rapid changes of network status and equipment status in the rapid startup phase by constructing a rapid startup network congestion control method suitable for the power system. The method first collects and preprocesses the power equipment and network flow data to ensure the accuracy and reliability of the data. Subsequently, the flow prediction model is trained using these preprocessed data, and the model can accurately predict the network flow in the future period of time, providing a basis for real-time decision-making. During the startup of the power system, network congestion detection is performed in combination with real-time network status and equipment status data to timely discover potential congestion risks. Finally, based on the results of network congestion detection and the prediction results of the flow prediction model, the flow scheduling and control strategy are dynamically adjusted to optimize resource allocation and avoid inappropriate decisions. This series of steps ensures that network congestion can be dealt with in a timely and accurate manner in the case of rapid changes in network status, thereby improving the efficiency and stability of the rapid startup of the power system.

[0144] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A fast start network congestion control method applicable to a power system, characterized in that: The following steps are involved: Step 1: Collect monitoring data and network flow data of power equipment, and perform preprocessing based on the collected detection data and network flow data to obtain preprocessed detection data and network flow data of power equipment; Step 2: Train the traffic prediction model based on the pre-processed power equipment detection data and network traffic data to obtain a trained traffic prediction model; Step 3: Use the trained traffic prediction model and perform traffic prediction based on real-time data to predict the network traffic in a certain period of time in the future, and obtain the network traffic prediction results in the future period of time; Step 4: During the startup of the power system, network congestion detection is performed based on real-time network status data and equipment status data of the power system; Step 5: Schedule and control traffic based on the results of network congestion detection and the prediction results of the traffic prediction model, and output the scheduling and control strategy.

2. A fast start network congestion control method applicable to a power system according to claim 1, characterized in that: The prediction module includes an input layer, an LSTM layer, an attention mechanism layer and an output layer; The input layer is used to receive input data; The LSTM layer is used to process the input data in the input layer, including a first LSTM layer and a second LSTM layer. The first LSTM layer processes the input data in the input layer and passes the output result to the second LSTM layer. The second LSTM layer processes the data and passes it to the attention mechanism layer. The attention mechanism layer selectively focuses on important time periods to improve the prediction accuracy of the model; The output layer includes a fully connected layer, which maps the high-dimensional features output by the LSTM layer to the final prediction result, represents the predicted value of network traffic at a certain point in the future, and uses a linear activation function to output continuous prediction values.

3. A fast start network congestion control method applicable to a power system according to claim 1, characterized in that: The network congestion detection method is as follows: Bandwidth utilization calculation: Where: U b (t) represents the bandwidth utilization at time t; C b (t) represents the bandwidth usage at time t; B b Indicates the total bandwidth of the link; Link delay: Where: D(t) represents the average link delay at time t; (d i (t) represents the delay of the ith data packet at time t; d0 represents the normal delay reference value; N represents the number of observed data packets; Packet loss rate: Where: P(t) represents the packet loss rate at time t; L(t) represents the number of packets lost at time t; T(t) represents the total number of packets sent at time t; The real-time calculated bandwidth utilization, link delay and packet loss rate are compared with the set thresholds. If any one of the bandwidth utilization, link delay and packet loss rate is greater than the corresponding preset threshold, it is judged as link congestion.

4. A fast start network congestion control method applicable to a power system according to claim 1, characterized in that: The specific steps for scheduling and controlling traffic based on the results of network congestion detection are as follows: Prioritization: pre-prioritize different types of traffic, where the priority classification is based on business importance, delay sensitivity, and bandwidth requirements. Bandwidth resource allocation and traffic redirection: When network congestion is detected, a traffic redirection strategy is used to shift part of the traffic to an alternate path or space network link; Based on network topology and traffic prediction, optimize the allocation of bandwidth resources so that traffic is evenly distributed across multiple links; Traffic speed limit and packet loss control: Limit the speed of low-priority traffic and discard some low-priority traffic packets in advance through the random early detection algorithm; dynamically adjust the congestion window size of the TCP connection to control the sending rate of data traffic.

5. A fast start network congestion control method applicable to a power system according to claim 4, characterized in that: The specific steps of the priority division are as follows: Calculation of business importance score I(t): Where: importance i represents the business importance score of flow i; max_importance represents the maximum importance score among all flows; Delay sensitivity score D(t) calculation: Where: delay_sensitivity i Indicates the sensitivity of flow i to delay; max_delay_sensitivity indicates the delay sensitivity of the flow that is most sensitive to delay among all flows; Bandwidth demand score B(t) calculation: Where: bandwidth_need i Indicates the bandwidth requirement of flow i; max_bandwidth_need indicates the maximum bandwidth requirement of all flows; Based on the business importance score, delay sensitivity score, and bandwidth requirement score, a priority score is calculated: P score (t)=w1·I(t)+w2·D(t)+w3·B(t); Wherein: w1, w2 and w3 all represent weighting coefficients; w1+w2+w3=1; Based on the calculated priority score P score (t) Priority classification.

6. A fast-start network congestion control method applicable to a power system according to claim 4, characterized in that: The specific steps of bandwidth resource allocation and traffic redirection are as follows: Dynamic optimization of bandwidth resource allocation: Calculate the bandwidth allocation for each link based on traffic priority and bandwidth requirements: Where: B k,i represents the bandwidth allocated to flow i on link k; B i represents the total bandwidth requirement of flow i; U k represents the current bandwidth utilization of link k; N represents the number of links in the network; Traffic redirection strategy: When the bandwidth utilization of a link exceeds the set threshold U threshold When the traffic is redirected to other backup link B j,i : B j,i ′=B j,i +δ·B k,i ; Where: δ represents the redirection coefficient; B j,i ′ represents the bandwidth of traffic i on link j after redirection; B j,i represents the bandwidth of flow i on link j before redirection; The amount of bandwidth redirected is determined by the priority of the traffic and the current bandwidth requirements of the link; Topology-aware bandwidth allocation: Optimizes bandwidth resource allocation based on traffic prediction results and current network topology: Where: represents the bandwidth demand forecast of flow i at the future time t+Δt, obtained by the flow prediction model; B optimized Indicates the optimized bandwidth allocation.

7. A fast-start network congestion control method applicable to a power system according to claim 4, characterized in that: The specific steps of flow rate limiting and packet loss control are as follows: Traffic speed limit: R adjusted =R current ·(1―α·U); Where: R adjusted Represents the adjusted traffic sending rate; R current represents the sending rate of the current traffic; α represents the traffic adjustment coefficient; U represents the bandwidth utilization of the current link; Packet loss control: Where: P drop Indicates the probability of packet loss; Q avg Indicates the average queue length; maxth indicates the maximum packet loss threshold; minth indicates the minimum packet loss threshold; Congestion window adjustment: Where: MSS represents the maximum segment size, that is, the maximum payload of each TCP data packet; cwnd old Indicates the current congestion window size; cwnd new Indicates the adjusted congestion window size.

Images