HRF+hplc communication flow prediction and load balancing method

Abstract

The application discloses a kind of HRF+HPLC communication flow prediction and load balancing method, it is related to network communication relevant technical field, method includes: to the HRF link and HPLC link in target communication network are carried out cross-media state perception;Based on cross-media feature group, link available bandwidth prediction is carried out;Real-time traffic data of target communication network is collected;In combination with real-time traffic data, HRF bandwidth prediction value, HPLC bandwidth prediction value, carry out shunt analysis based on delay cost, packet loss cost and link switching overhead;According to the shunt proportion trigger link switching, and to service message parallel copy sending.It solves the technical problems of heterogeneous communication media dynamic load imbalance, link state prediction accuracy deficiency and high cross-media switching delay and packet loss rate in the prior art, achieves the technical effects of improving network resource utilization, reducing service transmission delay and packet loss rate, and ensuring the stability of cross-media seamless switching.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, and particularly relates to a traffic prediction and load balancing method for HRF+HPLC communication. BACKGROUND

[0002] High-speed power line carrier communication (HPLC) and high-speed micro-power wireless communication (HRF) are important transmission means in power communication networks. HPLC uses existing power line infrastructure to realize data transmission, and has the advantages of low deployment cost and wide coverage. HRF provides a flexible access method through wireless radio frequency technology, and is suitable for scenarios where power lines cannot cover. However, in actual application, the two communication media face the challenges of dynamically changing channel conditions, unbalanced business load, and delay-sensitive business. Therefore, efficient traffic prediction and load balancing methods are needed to improve network performance. Traditional load balancing is mainly designed for homogeneous networks, and is difficult to adapt to the collaborative scenario of HRF and HPLC heterogeneous media. On the one hand, power line channels are easily affected by noise, impedance changes and load fluctuations, resulting in strong time-varying of HPLC link bandwidth. On the other hand, the available bandwidth of HRF link is also uncertain due to environmental interference, multipath effect and other factors. In addition, the traffic splitting of cross-media communication needs to consider the delay cost, packet loss rate and link switching overhead to avoid business jitter caused by frequent switching. The existing methods lack joint perception and dynamic prediction ability for cross-media characteristics, and it is difficult to achieve optimal resource allocation, thereby affecting the resource utilization and service quality of power communication networks.

[0003] In the related art at present, there are technical problems of dynamic load imbalance of heterogeneous communication media, insufficient link state prediction accuracy, and high switching delay and packet loss rate of cross-media. SUMMARY

[0004] The present application provides a traffic prediction and load balancing method for HRF+HPLC communication, which solves the technical problems of dynamic load imbalance of heterogeneous communication media, insufficient link state prediction accuracy, and high switching delay and packet loss rate of cross-media in the prior art, and achieves the technical effects of improving network resource utilization, reducing business transmission delay and packet loss rate, and ensuring the stability of cross-media seamless switching.

[0005] The application provides a flow prediction and load balancing method for HRF+HPLC communication, which comprises: performing cross-medium state perception on HRF links and HPLC links in a target communication network in a preset period to generate a cross-medium feature group; performing link available bandwidth prediction based on the cross-medium feature group to output HRF bandwidth prediction values and HPLC bandwidth prediction values; collecting real-time traffic data of the target communication network; performing shunting analysis based on delay cost, packet loss cost and link switching overhead in combination with the real-time traffic data, the HRF bandwidth prediction values and the HPLC bandwidth prediction values to obtain a shunting ratio; triggering link switching according to the shunting ratio, and performing parallel replication and sending of service packets on the HRF links and the HPLC links in a transition period before triggering switching.

[0006] In a possible implementation, the flow prediction and load balancing method for HRF+HPLC communication further performs the following processing: the target communication network comprises terminals and gateways simultaneously having HRF links and HPLC links.

[0007] In a possible implementation, the flow prediction and load balancing method for HRF+HPLC communication further performs the following processing: performing active probing on the HRF links based on preset HRF link indicators to collect HRF link indicator data sets; performing active probing on the HPLC links based on preset HPLC link indicators to collect HPLC link indicator data sets; performing indicator vectorization processing on the HRF link indicator data sets and the HPLC link indicator data sets to generate the cross-medium feature group.

[0008] In a possible implementation, the flow prediction and load balancing method for HRF+HPLC communication further performs the following processing: the preset HRF link indicators comprise signal strength, signal quality, round-trip delay, packet loss rate and data throughput; and the preset HPLC link indicators comprise noise spectrum, subcarrier bit error rate, round-trip delay, packet loss rate and data throughput.

[0009] In a possible implementation, the flow prediction and load balancing method for HRF+HPLC communication further performs the following processing: collecting historical data of the HRF links based on the preset HRF link indicators, learning the relationship between HRF link indicators and available bandwidth, and constructing a first bandwidth predictor; collecting historical data of the HPLC links based on the preset HPLC link indicators, learning the relationship between HPLC link indicators and available bandwidth, and constructing a second bandwidth predictor; integrating the first bandwidth predictor and the second bandwidth predictor to analyze the cross-medium feature group, and generating the HRF bandwidth prediction values and the HPLC bandwidth prediction values.

[0010] In a possible implementation, the HRF+HPLC communication traffic prediction and load balancing method further performs the following processing: collecting current link usage states of the HRF link and the HPLC link; performing shunting simulation on time delay cost, packet loss cost, and link switching overhead for the HRF link according to the HRF bandwidth prediction value, the real-time traffic data, and the current link usage states, to generate a plurality of first shunting ratios and a plurality of first shunting costs; performing shunting simulation on time delay cost, packet loss cost, and link switching overhead for the HPLC link according to the HPLC bandwidth prediction value, the real-time traffic data, and the current link usage states, to generate a plurality of second shunting ratios and a plurality of second shunting costs; and performing main link selection according to the plurality of first shunting ratios and the plurality of first shunting costs, the plurality of second shunting ratios and the plurality of second shunting costs, to perform shunting ratio calculation based on a main link, and to generate the shunting ratio.

[0011] In a possible implementation, the HRF+HPLC communication traffic prediction and load balancing method further performs the following processing: performing twin modeling on the HRF link and the HPLC link respectively, to generate an HRF twin link and an HPLC twin link; loading the HRF bandwidth prediction value and the current link usage state based on the HRF twin link, performing shunting ratio iterative simulation according to the real-time traffic data, and weighting time delay cost, packet loss cost, and link switching overhead obtained through simulation, to generate the plurality of first shunting ratios and the plurality of first shunting costs.

[0012] In a possible implementation, the HRF+HPLC communication traffic prediction and load balancing method further performs the following processing: the link switching overhead is overhead generated due to session reconstruction when switching between the HRF link and the HPLC link.

[0013] In a possible implementation, the HRF+HPLC communication traffic prediction and load balancing method further performs the following processing: filtering a first optimal ratio and a first optimal cost with minimum shunting cost according to the plurality of first shunting ratios and the plurality of first shunting costs; filtering a second optimal ratio and a second optimal cost with minimum shunting cost according to the plurality of second shunting ratios and the plurality of second shunting costs; analyzing a first ratio adjustment threshold of the first optimal ratio and a second ratio adjustment threshold of the second optimal ratio based on the current link usage states; and if the first ratio adjustment threshold is smaller than the second ratio adjustment threshold, taking the HRF link as a main link, and generating the shunting ratio based on the first optimal ratio.

[0014] In a possible implementation, the HRF+HPLC communication flow prediction and load balancing method further performs the following processing: during a transition period before triggering the switching, the service message is sent in parallel on the HRF link and the HPLC link, and the parallel sending duration is not less than three times the round-trip delay.

[0015] The HRF+HPLC communication flow prediction and load balancing method provided in the present application performs cross-media state perception on the HRF link and the HPLC link in the target communication network; performs link available bandwidth prediction based on the cross-media feature group; collects real-time traffic data of the target communication network; performs flow splitting analysis based on delay cost, packet loss cost, and link switching overhead in combination with the real-time traffic data, the HRF bandwidth prediction value, and the HPLC bandwidth prediction value; triggers link switching according to the flow splitting ratio, and performs parallel replication and sending of the service message. The technical problems of dynamic load imbalance of heterogeneous communication media, insufficient link state prediction accuracy, and high cross-media switching delay and packet loss rate in the prior art are solved, and the technical effects of improving network resource utilization, reducing service transmission delay and packet loss rate, and ensuring the stability of cross-media seamless switching are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. In the present application, flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. Meanwhile, other operations can be added to these processes, or one or more steps of operations can be removed from these processes.

[0017] Figure 1 A flowchart of the HRF+HPLC communication flow prediction and load balancing method provided in the embodiments of the present application.

[0018] Figure 2 A flowchart of cross-media state perception in the HRF+HPLC communication flow prediction and load balancing method provided in the embodiments of the present application. DETAILED DESCRIPTION

[0019] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0020] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings, and the described embodiments should not be regarded as limitations to the present application. All other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0021] In the following description, "some embodiments" are referred to, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The term "first\second" referred to only distinguishes similar objects, and does not represent a specific order of the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, product or server including a series of steps does not have to be limited to those steps clearly listed, but can include other steps not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0022] The embodiments of the present application provide a flow prediction and load balancing method for HRF+HPLC communication, as shown in Figure 1 The method comprises the following steps:

[0023] In step S100, the cross-medium state of the HRF link and the HPLC link in the target communication network is perceived in a preset period, and a cross-medium feature group is generated.

[0024] Preferably, the cross-medium state perception refers to real-time monitoring and data collection of the HRF link and the HPLC link in the target communication network in a preset period, and obtaining the running state indicators of the two types of links. Since the physical characteristics of HRF and HPLC are different, the channel parameters of each type need to be collected respectively. The preset period can be every 1 second or 5 seconds. Specifically, a probe frame such as Ping or a special signaling is sent to the HRF link, and the signal strength, signal quality, round-trip delay, packet loss rate and data throughput of the HRF link are collected. In addition, a power line channel probe signal such as impedance measurement and noise scanning is sent to the HPLC link, and the noise spectrum, subcarrier bit error rate, round-trip delay, packet loss rate and data throughput of the HPLC link are collected. These raw data are converted into structured vectors, and then a cross-medium feature group is generated, which is used for link bandwidth prediction, dynamic offloading decision and the like, to ensure efficient and reliable transmission of services in a dynamic environment.

[0025] Further, step S100 further comprises that the target communication network comprises terminals and gateways which simultaneously have HRF link and HPLC link.

[0026] Preferably, the terminal devices such as smart meters, sensors and the gateway devices such as concentrators and edge servers of the target communication network simultaneously support two communications including high-speed power line carrier communication (HPLC) and high-speed micro-power wireless communication (HRF), wherein the HPLC communication can include G3-PLC, PRIME, etc., and the HRF communication is a wireless radio frequency-based communication mode such as Wi-SUN, LoRa, ZigBee, etc. Specifically, the terminal devices are used for data acquisition and control instruction execution, and are built-in with HRF modules and HPLC modules. For example, the smart meters communicate with the gateway through the power line (HPLC) and simultaneously use the wireless (HRF) as a backup link. The gateway devices are used for data aggregation, protocol conversion, load balancing decision, support HRF and HPLC dual protocol stacks, and can simultaneously interact with the terminals through the two links. Both the terminal and the gateway support wireless (HRF) and power line (HPLC) two communication modes, and improve the network reliability, coverage range and resource utilization through dynamic cooperation.

[0027] Further, step S100 further comprises step S110 of actively probing the HRF link based on preset HRF link indexes to collect an HRF link index data set; step S120 of actively probing the HPLC link based on preset HPLC link indexes to collect an HPLC link index data set; and step S130 of performing index vectorization processing on the HRF link index data set and the HPLC link index data set to generate the cross-media feature group.

[0028] Step S100 further comprises that the preset HRF link indexes include signal strength, signal quality, round-trip delay, packet loss rate and data throughput; and the preset HPLC link indexes include noise spectrum, subcarrier bit error rate, round-trip delay, packet loss rate and data throughput.

[0029] Preferably, the real-time state data of the HRF (wireless radio frequency) and HPLC (power line carrier) links are respectively acquired through periodic active probing. Specifically, a probing frame such as a Ping packet or a special signaling frame is sent to the HRF link, the link state is analyzed through the response, and multiple HRF link indexes are acquired, including signal strength, signal quality, round-trip delay, packet loss rate and data throughput. The signal strength reflects the wireless signal receiving power; the signal quality includes signal-to-noise ratio and bit error rate, and measures the signal anti-interference capability; the round-trip delay is the time from sending the probing packet to receiving the confirmation; the packet loss rate is the proportion of the lost probing frames, for example, if 100 packets are sent and 5 packets are lost, then the packet loss rate is 5%; and the data throughput is the actual transmission rate of the current link.

[0030] Preferably, the HPLC link is actively probed by sending a probe signal of a specific frequency band, such as a sweep signal or an impedance measurement signal, to obtain a plurality of HPLC link indicators, including a noise spectrum, a subcarrier bit error rate, a round-trip delay, a packet loss rate, and a data throughput, wherein the noise spectrum is an analysis of the noise types of each frequency band of the power line, such as narrowband noise and impulse noise; the subcarrier bit error rate is the bit error rate of each subcarrier in OFDM modulation, reflecting the channel quality; the round-trip delay is the transmission delay of the power line channel; the packet loss rate is the data loss in the power line environment; and the data throughput is the effective bandwidth of the power line link. Then, the collected raw link indicator data is preprocessed, including normalization and missing value filling, and is converted into a structured feature vector, and then a cross-medium feature group is formed, i.e., the physical layer indicators of the wireless (HRF) and the power line (HPLC) are uniformly expressed, realizing the quantitative perception of the HRF and HPLC links, reflecting the state of the dual-link in real time, supporting the subsequent bandwidth prediction and shunt ratio calculation, so as to facilitate the load balancing of the heterogeneous network.

[0031] In step S200, based on the cross-medium feature group, link available bandwidth prediction is performed, and HRF bandwidth prediction value and HPLC bandwidth prediction value are output.

[0032] Further, as shown in Figure 2 step S200 further includes step S210 of collecting historical data of the HRF link based on the preset HRF link indicators, learning the relationship between the HRF link indicators and the available bandwidth, and constructing a first bandwidth predictor; step S220 of collecting historical data of the HPLC link based on the preset HPLC link indicators, learning the relationship between the HPLC link indicators and the available bandwidth, and constructing a second bandwidth predictor; and step S230 of integrating the first bandwidth predictor and the second bandwidth predictor to analyze the cross-medium feature group, and generating the HRF bandwidth prediction value and the HPLC bandwidth prediction value.

[0033] Preferably, based on the cross-medium feature set, an index-bandwidth mapping model of the HRF and HPLC link is established by machine learning to perform link available bandwidth prediction. Specifically, according to the preset HRF link index, historical data of the HRF link is collected, and the signal strength, signal quality, round-trip delay, packet loss rate, and data throughput of the HRF link are continuously recorded to form an HRF historical index dataset as input features. A bandwidth prediction model is established based on a time series model or a regression model, the HRF historical index dataset is input, the available bandwidth measured by the iperf tool is taken as an output label, the relationship between the HRF link index and the available bandwidth is learned, a first bandwidth predictor is constructed for processing HRF features, and an HRF bandwidth prediction value is output. According to the preset HPLC link index, historical data of the HPLC link is collected, and the noise spectrum, subcarrier bit error rate, round-trip delay, packet loss rate, and data throughput of the HPLC link are continuously recorded to form an HPLC historical index dataset as input features. A bandwidth prediction model is constructed based on a random forest anti-noise model, the HPLC historical index dataset is input, the measured available bandwidth is taken as an output label, the relationship between the HPLC link index and the available bandwidth is learned, a second bandwidth predictor is constructed for processing HPLC features, and an HPLC bandwidth prediction value is output. Then, the first bandwidth predictor and the second bandwidth predictor are integrated, and the cross-medium feature set is input for prediction analysis to generate the HRF bandwidth prediction value and the HPLC bandwidth prediction value. Through the dual-predictor architecture, the link characteristics and bandwidth relationship of the HRF (wireless) and the HPLC (power line) are learned respectively to achieve accurate cross-medium resource prediction.

[0034] In step S300, real-time traffic data of the target communication network is collected.

[0035] Preferably, the real-time traffic data refers to the state information of the service flow being transmitted in the target communication network, which is used to quantify the load and service demand of the current network, analyze the service type through deep packet inspection, collect traffic statistics information using SNMP or NetFlow protocol, or embed a lightweight probe in the gateway or terminal to record the service characteristics in real time, and obtain real-time traffic data including service traffic characteristics, service type and priority, delay and jitter information, and queue state. Specifically, the service traffic characteristics include data throughput, the data rate being transmitted on each link at present, such as 5 Mbps for HRF link and 8 Mbps for HPLC link; the number of concurrent connections, the number of simultaneously active communication sessions between the terminal and the gateway; the packet size distribution, the proportion of large packets (such as file transfer) and small packets (such as heartbeat signal). The service classification identification includes key services such as power grid fault alarm and remote control instruction, ordinary services such as regular power meter data reporting, or delay-sensitive and bandwidth-sensitive services. The delay and jitter information includes end-to-end delay, the actual delay of service data from sending to receiving; delay jitter, the variation amplitude of the delay of consecutive data packets, reflecting the stability of the link. The queue state includes the cache queue length, the backlog of data packets to be forwarded in the gateway or terminal; and the packet loss record, the packet loss statistics due to queue overflow or link error.

[0036] In step S400, the real-time traffic data, the HRF bandwidth prediction value, and the HPLC bandwidth prediction value are combined to perform flow splitting analysis based on delay cost, packet loss cost, and link switching overhead, and the flow splitting ratio is obtained.

[0037] Further, step S400 further includes that the link switching overhead is the overhead generated due to session reconstruction when switching between the HRF link and the HPLC link.

[0038] Preferably, the link switching overhead refers to the additional resource consumption generated due to network protocol layer session reconstruction when the service flow is switched from the HRF link to the HPLC link or vice versa, mainly including signaling interaction, control packets required for re-establishing connection, such as TCP three-way handshake and TLS key negotiation; state synchronization, delay caused by transmitting session parameters such as sequence number and window size; service interruption, temporary data transmission stagnation during switching; for example, in the smart grid, if the power meter is switched from HPLC to HRF, the communication protocol needs to be re-negotiated, which may introduce an additional delay of 10-50 ms. This overhead needs to be optimized by weighing the link quality in the flow splitting decision.

[0039] The step S400 further comprises a step S410 of collecting current link usage states of the HRF link and the HPLC link; a step S420 of performing flow splitting simulation on latency cost, packet loss cost and link switching overhead for the HRF link according to the HRF bandwidth prediction value, the real-time traffic data and the current link usage states, generating a plurality of first flow splitting ratios and a plurality of first flow splitting costs; a step S430 of performing flow splitting simulation on latency cost, packet loss cost and link switching overhead for the HPLC link according to the HPLC bandwidth prediction value, the real-time traffic data and the current link usage states, generating a plurality of second flow splitting ratios and a plurality of second flow splitting costs; and a step S440 of performing master link selection according to the plurality of first flow splitting ratios and the plurality of first flow splitting costs, the plurality of second flow splitting ratios and the plurality of second flow splitting costs, performing flow splitting ratio calculation based on the master link, and generating the flow splitting ratio.

[0040] Preferably, intelligent flow distribution is achieved between the HRF (wireless) and the HPLC (power line) dual links through dynamic simulation and cost evaluation, i.e., flow splitting analysis based on latency cost, packet loss cost and link switching overhead. Specifically, the current link usage states of the HRF link and the HPLC link are collected, including real-time signal strength, queue occupancy, current transmitted traffic type and proportion of the HRF link, and power line noise level, cache queue depth, subcarrier availability of the HPLC link, for example, HRF link signal strength -72dBm, queue occupancy 70%, 3-way video monitoring data being transmitted; HPLC link noise level "medium", queue occupancy 40%, carrying 5-way power meter data.

[0041] Preferably, for the HRF link, the HRF bandwidth prediction value, real-time traffic data and current link usage states are inputted, and then flow splitting simulation on latency cost, packet loss cost and link switching overhead is performed. Specifically, different flow splitting ratios are simulated, such as HRF carrying 10% / 30% / 50% flow, and then the three costs of latency cost, packet loss cost and link switching overhead are calculated. The latency cost refers to the predicted latency change after flow splitting based on the current RTT and queue state of the HRF; the packet loss cost refers to the estimated packet loss probability after flow splitting according to the HRF packet loss rate model; and the link switching overhead refers to the additional overhead of session reconstruction when switching from the HPLC to the HRF, such as signaling interaction time. Then, a plurality of first flow splitting costs corresponding to different first flow splitting ratios are simulated and outputted, such as the costs corresponding to HRF carrying 10% / 30% / 50% flow, which are [latency=50ms, packet loss=2%, overhead=5ms], [latency=80ms, packet loss=5%, overhead=5ms], [latency=120ms, packet loss=8%, overhead=5ms].

[0042] Preferably, the same is true for the HPLC link. The input HPLC bandwidth prediction value, real-time traffic data and current link usage state are used to perform simulation of the split ratio under different conditions, such as HPLC bearing 60% / 80% / 100% traffic, and then calculate the three costs, i.e. the delay cost, the packet loss cost and the link switching overhead, which are the effects of power line noise on the delay and packet loss, such as the error rate soaring under high noise. Finally, the simulation outputs multiple second split costs corresponding to different second split ratios, such as the costs corresponding to HPLC bearing 60% / 80% / 100% traffic are [delay = 30ms, packet loss = 1%, overhead = 10ms], [delay = 40ms, packet loss = 3%, overhead = 10ms] and [delay = 60ms, packet loss = 6%, overhead = 10ms].

[0043] Preferably, the main link is selected according to the multiple first split ratios and the multiple first split costs, the multiple second split ratios and the multiple second split costs. Specifically, for each candidate ratio, the total cost is calculated by weighted summation of the costs, such as the total cost of HRF candidate 30% ratio is 80x0.5+5x0.3+5x0.2=44.5, and the total cost of HPLC candidate 80% ratio is 40x0.5+3x0.3+10x0.2=22.9. Then the link with the minimum total cost is selected as the main link, and the split ratio calculation based on the main link is performed, i.e. on the basis of HPLC as the main link, the optimal ratio (80%) is selected, and the remaining 20% is allocated to HRF. Finally, the split ratio is generated, and dynamic traffic distribution is realized on the dual link of HRF and HPLC, ensuring the reliability of communication transmission.

[0044] Further, step S420 further comprises step S421 of performing twin modeling on the HRF link and the HPLC link respectively to generate an HRF twin link and an HPLC twin link; and step S422 of loading the HRF bandwidth prediction value and the current link usage state based on the HRF twin link, performing split ratio iterative simulation according to the real-time traffic data and the preset ratio compensation, and weighting the delay cost, the packet loss cost and the link switching overhead obtained by simulation to generate the multiple first split ratios and the multiple first split costs.

[0045] Preferably, by constructing a digital twin model for the HRF and HPLC link, the dynamic characteristics of the real link are reproduced in the simulation environment, including channel behavior, transmission delay, packet loss mechanism, etc., wherein the input of the twin model is the real-time collected signal strength, noise spectrum, etc. Link state and bandwidth prediction value, the output is the simulated shunt effect, delay, packet loss, etc. Cost, take the HRF twin link as an example, model the HRF link, generate the HRF twin link, load the HRF bandwidth prediction value and the current link usage state into the HRF twin link, and compensate according to the preset proportion, that is, according to the business priority, preset the compensation weight of the high latency sensitive flow, such as the video flow delay weight +20%; Then perform shunt proportion iterative simulation according to real-time traffic data, including simulating business transmission under different shunt proportions (10% / 30% / 50%), recording the delay cost, packet loss cost and link switching overhead of each proportion, and finally dynamically adjusting the weight according to the business demand, such as delay weight 0.6, packet loss weight 0.3, link switching overhead weight 0.1, calculating the comprehensive first shunt cost, and finally output multiple first shunt proportions and their corresponding multiple first shunt costs. Similarly, for the HPLC link, multiple second shunt proportions and their corresponding multiple second shunt costs are generated.

[0046] Further, step S440 further comprises step S441 of screening the first optimal proportion and the first optimal cost with the minimum shunt cost according to the multiple first shunt proportions and the multiple first shunt costs; step S442 of screening the second optimal proportion and the second optimal cost with the minimum shunt cost according to the multiple second shunt proportions and the multiple second shunt costs; step S443 of analyzing the first proportion adjustment threshold of the first optimal proportion and the second proportion adjustment threshold of the second optimal proportion based on the current link usage state; and step S444 of taking the HRF link as the main link and generating the shunt proportion with the first optimal proportion if the first proportion adjustment threshold is less than the second proportion adjustment threshold.

[0047] Preferably, the optimal solution is selected from the candidate offloading schemes of the HRF and HPLC link by cost comparison and dynamic threshold adjustment, and the main link and offloading ratio are finally determined, that is, the comprehensive cost of delay, packet loss and switching overhead is minimized under the premise of meeting the service demand. Specifically, from the multiple offloading ratios simulated by the HRF, the option with the minimum comprehensive offloading cost is selected as the first optimal ratio and the first optimal cost; similarly, from the candidate ratios simulated by the HPLC, the option with the minimum comprehensive offloading cost is selected as the second optimal ratio and the second optimal cost. Then, based on the current link usage state, the first ratio adjustment threshold of the first optimal ratio and the second ratio adjustment threshold of the second optimal ratio are analyzed respectively, wherein the ratio adjustment threshold refers to the maximum offloading ratio deviation that the link can tolerate in the current state, which is used to prevent overload or resource waste. The first ratio adjustment threshold is calculated according to the real-time queue depth and signal stability of the HRF link, for example, if the current HRF queue occupancy rate has reached 70%, the first ratio adjustment threshold may be set to ±10%, that is, the offloading ratio cannot exceed 40%; the second ratio adjustment threshold is dynamically set based on the power line noise level and buffer capacity, for example, if the noise level is "medium", the second ratio adjustment threshold may be ±15%; then the first ratio adjustment threshold and the second ratio adjustment threshold are compared, if the first ratio adjustment threshold of the HRF is less than the second ratio adjustment threshold of the HPLC, it means that the state of the HRF link is more stable, then the HRF is preferentially selected as the main link, and the offloading ratio is generated with the first optimal ratio, for example, 30% of the optimal ratio of the HRF is taken as the main link, and the remaining 70% is allocated to the HPLC, thereby realizing intelligent load distribution of heterogeneous links and ensuring decision accuracy.

[0048] Step S500 triggers link switching according to the offloading ratio, and during the transition period before triggering the switching, the service packets are sent in parallel on the HRF link and the HPLC link.

[0049] Step S500 further includes sending the service packets in parallel on the HRF link and the HPLC link during the transition period before triggering the switching, and the duration of parallel sending is not less than three times the round-trip delay.

[0050] Preferably, according to the shunt ratio triggered link switching, by transmitting the service message on the HRF and HPLC link in parallel during the transition period (not less than 3 times the round-trip delay) before switching, it ensures that the data is not lost and the service is not aware of the switching. Specifically, parallel copying means that the gateway copies the to-be-sent message to the HRF and HPLC link at the same time before triggering the switching, and the same data is transmitted on the two links at the same time; the parallel copying transmission time is not less than three times the round-trip delay, which ensures that the receiving end can at least obtain complete data from one link, even if the other link is interrupted due to switching; after the transition period ends, the switching is completed, the redundant transmission is stopped, and only the selected main link is used to transmit the subsequent message, avoiding the waiting time of session reconstruction in traditional switching, which can improve the network resource utilization, reduce the service transmission delay and packet loss rate, guarantee the stability of cross-media seamless switching, ensure the transmission reliability, and is suitable for high-sensitive services such as smart grid control instructions.

[0051] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. 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. In some cases, the actions or steps described in the present application can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or may be advantageous.

Claims

1. A method for flow prediction and load balancing using HRF+HPLC communication, characterized in that, include: Within a preset period, cross-media state sensing is performed on the HRF link and HPLC link in the target communication network to generate cross-media feature groups; Based on the cross-media feature set, perform link available bandwidth prediction and output HRF bandwidth prediction value and HPLC bandwidth prediction value; Collect real-time traffic data of the target communication network; By combining the real-time traffic data, the HRF bandwidth prediction value, and the HPLC bandwidth prediction value, a traffic splitting analysis based on latency cost, packet loss cost, and link switching overhead is performed to obtain the traffic splitting ratio. According to the aforementioned traffic splitting ratio, link switching is triggered, and during the transition period before the switching is triggered, service packets are copied and sent in parallel on the HRF link and the HPLC link. Based on the aforementioned cross-media feature set, perform link available bandwidth prediction and output HRF bandwidth prediction values ​​and HPLC bandwidth prediction values, including: Based on preset HRF link metrics, historical data of the HRF link is collected to learn the relationship between the HRF link metrics and available bandwidth, and a first bandwidth predictor is constructed. Based on preset HPLC link indicators, historical data of the HPLC link are collected, the relationship between the HPLC link indicators and available bandwidth is learned, and a second bandwidth predictor is constructed. The first bandwidth predictor and the second bandwidth predictor are integrated to analyze the cross-media feature set and generate the HRF bandwidth prediction value and the HPLC bandwidth prediction value.

2. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 1, characterized in that, The target communication network includes terminals and gateways that simultaneously have HRF and HPLC links.

3. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 1, characterized in that, Within a preset period, cross-media state sensing is performed on the HRF link and HPLC link in the target communication network to generate cross-media feature sets, including: Actively probe the HRF link based on the preset HRF link indicators and collect the HRF link indicator dataset. Actively probe the HPLC link based on preset HPLC link indicators and collect HPLC link indicator dataset. The HRF link index dataset and the HPLC link index dataset are vectorized to generate the cross-media feature group.

4. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 3, characterized in that, The preset HRF link metrics include signal strength, signal quality, round-trip time, packet loss rate, and data throughput; the preset HPLC link metrics include noise spectrum, subcarrier bit error rate, round-trip time, packet loss rate, and data throughput.

5. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 1, characterized in that, Combining the real-time traffic data, the HRF bandwidth prediction value, and the HPLC bandwidth prediction value, a traffic splitting analysis based on latency cost, packet loss cost, and link switching overhead is performed to determine the traffic splitting ratio, including: Collect the current usage status of the HRF link and the HPLC link; For the HRF link, based on the predicted HRF bandwidth, the real-time traffic data, and the current link usage status, a traffic splitting simulation is performed to consider latency cost, packet loss cost, and link switching overhead, generating multiple first splitting ratios and multiple first splitting costs. For the HPLC link, based on the HPLC bandwidth prediction value, the real-time traffic data, and the current link usage status, a traffic splitting simulation is performed to generate multiple second traffic splitting ratios and multiple second traffic splitting costs. Based on the plurality of first traffic splitting ratios and plurality of first traffic splitting costs, the plurality of second traffic splitting ratios and plurality of second traffic splitting costs, a primary link is selected, and a traffic splitting ratio calculation based on the primary link is performed to generate the traffic splitting ratio.

6. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 5, characterized in that, For the HRF link, based on the predicted HRF bandwidth, the real-time traffic data, and the current link usage status, a traffic splitting simulation is performed to consider latency cost, packet loss cost, and link switching overhead, generating multiple first splitting ratios and multiple first splitting costs, including: The HRF link and the HPLC link are modeled using twins to generate HRF twin links and HPLC twin links respectively; Based on the HRF twin link, the predicted HRF bandwidth and the current link usage status are loaded, and compensation is performed according to a preset ratio. The traffic splitting ratio iterative simulation is performed according to the real-time traffic volume data, and the latency cost, packet loss cost and link switching overhead obtained from the simulation are weighted to generate the multiple first traffic splitting ratios and their respective multiple first traffic splitting costs.

7. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 6, characterized in that, The link switching overhead is the overhead incurred due to session reconstruction when switching between the HRF link and the HPLC link.

8. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 5, characterized in that, Based on the plurality of first traffic splitting ratios and plurality of first traffic splitting costs, the plurality of second traffic splitting ratios and plurality of second traffic splitting costs, a primary link is selected, and a traffic splitting ratio calculation based on the primary link is performed to generate the traffic splitting ratio, including: Based on the multiple first diversion ratios and multiple first diversion costs, select the first optimal ratio and the first optimal cost with the minimum diversion cost; Based on the multiple second diversion ratios and multiple second diversion costs, select the second optimal ratio and the second optimal cost that have the lowest diversion cost; Based on the current link usage status, analyze the first ratio adjustment threshold of the first optimal ratio and the second ratio adjustment threshold of the second optimal ratio; If the first ratio adjustment threshold is less than the second ratio adjustment threshold, the HRF link is used as the main link, and the traffic splitting ratio is generated according to the first optimal ratio.

9. The method for flow prediction and load balancing using HRF+HPLC communication as described in claim 1, characterized in that, During the transition period before the switchover is triggered, service packets are copied and sent in parallel on the HRF link and the HPLC link, and the duration of parallel copying and sending is not less than three times the round-trip delay.

Images