Power data release path planning method and system based on load balancing

Abstract

The invention belongs to the technical field of power data transmission, provides a power data release path planning method and system based on load balancing, and provides a multi-path transmission strategy optimization method based on LSTM prediction and MPC, and the method comprises the steps: firstly constructing a multi-path data transmission state prediction model through LSTM; acquiring a data transmission state prediction value of each path at the next moment; then, inputting the predicted value at the next moment into the LSTM model so as to obtain the data transmission predicted quantity of a period of time in the future; and finally, on the basis of taking the set minimum release time as a target function, adopting rolling optimization and feedback correction of MPC to obtain each path transmission control quantity meeting the target function. The efficiency that the electricity consumption information acquisition system publishes the total acquisition data to the marketing business application system is improved, and a load balancing strategy is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power data transmission, and in particular to a method and system for planning a power data release path based on load balancing. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The current electricity consumption information collection system serves as the basic data platform for marketing business applications. In order to improve the efficiency of the collection system in publishing the full amount of collected data to the marketing business application system, the electricity consumption information collection system is currently integrated with the marketing business application system. That is, massive data is collected. After the pre-data collection is completed, it is first cached in the collection system's unstructured data storage, and then the data is synchronized with the marketing system's unstructured data storage through the unstructured data synchronization mechanism, so as to further shorten the release time of the full amount of meter reading data, meet the practical needs of the main station software functions of the electricity consumption information collection system, and fully support the in-depth application work such as electricity fee recovery and marketing and distribution basic data governance.

[0004] Traditional single-path transmission technology can no longer meet the needs of publishing full power data, and multi-path transmission technology has become a new option. Multi-path transmission technology refers to establishing multiple transmission paths between the communicating parties through multiple network interfaces on the device to achieve bandwidth aggregation, load balancing, path redundancy, increase transmission throughput, and improve reliability. Since multi-paths have differences in path states such as bandwidth, latency, and packet loss rate, multi-path transmission protocols can provide customized and differentiated transmission services for applications based on different path states by selecting different mechanism algorithms or parameters.

[0005] Multipath transmission technology also faces some challenges and problems. For example, how to select and plan the transmission path, how to achieve load balancing and congestion control, how to ensure the security and reliability of transmission, etc. Some of the currently public solutions include: "Xiao Ting'e. Research on congestion control algorithm based on multipath TCP [D]. Central China Normal University, 2023. DOI: 10.27159 / d.cnki.ghzsu.2023.001478", which proposes a multipath congestion control optimization algorithm DL-LIA based on packet loss differentiation. Different sending window adjustment methods are adopted according to the differentiated network packet loss types, so that the window can be restored to a high state faster and the sending rate can be improved. "Zhou Jieyu. Research on Multi-source Multi-path Congestion Control Mechanism for NDN[D]. Central South University, 2022. DOI: 10.27661 / d.cnki.gzhnu.2022.001834" designed a multi-source multi-path identification mechanism and a nearest data source search algorithm, which can be used to identify different sub-paths of all data sources and deal with the cache invalidation problem of the identification mechanism. "Li X, Zhou X. Reliability Measurement in A Multi-path Transmission Network using SMP-BP Algorithm[J]. International Journal of Online and Biomedical Engineering, 2019, 15(11): 16-28" proposed an improved algorithm using a single minimum path-based backup path (SSMP-BP), which aims to ensure data transmission when the second path is not working. However, the above methods have shortcomings such as poor adaptability to dynamic network environments, slow transmission path switching, and load imbalance. Summary of the invention

[0006] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a power data release path planning method and system based on load balancing, which proposes a multi-path selection release strategy based on LSTM prediction combined with model predictive control (MPC), realizes load balancing and congestion control of multi-path transmission, further shortens the release time of full meter reading data from the electricity consumption information collection system to the marketing business application system, and ensures the stability and security of meter reading data release operation.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A first aspect of the present invention provides a method for planning a power data publishing path based on load balancing, comprising the following steps:

[0009] Get the data transmission status of each path at the current moment;

[0010] Based on the data transmission status of each path at the current moment and the constructed short-term multi-path data transmission status prediction model, the data transmission status prediction value of each path at the next moment is obtained, and the data transmission status prediction value of each path at the next moment is used as the input of the short-term multi-path data transmission status prediction model, and the multi-path data transmission status prediction value within the specified time period is iteratively obtained;

[0011] Taking the minimum overall data transmission time as the objective function and path selection as the constraint condition, the MPC rolling optimization is adopted based on the predicted value of the multi-path data transmission state within the specified time period to perform multi-step predictive control on the multi-path selection. The path selection sequence of the first moment in the obtained optimal path selection path sequence is used as the path selection transmission operation of each moment, and the operation of the current moment is applied to the data transmission path selection at the next moment, until the selection operation of each path transmission at each moment in the specified time period is obtained.

[0012] Furthermore, the calculation process of the multipath data transmission state prediction value for the specified time period includes:

[0013] According to the bandwidth, transmission delay and packet loss rate data of each path at the current moment and the constructed short-term multi-path data transmission state prediction model, the transmission delay and packet loss rate of each path at the next moment are predicted;

[0014] The transmission time per unit time of each path at the next moment is calculated based on the transmission delay and packet loss rate of each path at the next moment;

[0015] The transmission delay and packet loss rate prediction values ​​of each path at the next moment are input into the short-term multipath data transmission state prediction model, and the unit time transmission time of each path within a period of time is obtained by repeated iteration.

[0016] Furthermore, the calculation formula for the transmission time per unit time of each path at the next moment is:

[0017]

[0018] Among them, MSS is the maximum segment length, RTT i (t+1) is the round trip time, which is numerically equal to 2d i (t+1); d i (t+1) and l i (t+1) is the transmission delay and packet loss rate of each path at time t+1.

[0019] Furthermore, the objective function is:

[0020]

[0021] The constraints are:

[0022]

[0023] x i (k)∈{0,1},

[0024] Among them, T total (k) is the total transmission time predicted in the kth step, α is the weight coefficient of different paths, P is the total number of predicted steps, x i (k) indicates whether the i-th path is selected in the k-th step, 1 indicates selection, 0 indicates no selection, and n indicates the total number of paths.

[0025] Furthermore, the short-term multi-path data transmission state prediction model adopts an LSTM model.

[0026] Furthermore, after obtaining the actual path data transmission selection operation, the data release time at the current moment is obtained, the prediction error of the short-term multipath data transmission state prediction model is calculated, and then the prediction error is used to correct the prediction value of the model, thereby realizing feedback correction.

[0027] Furthermore, the calculation formula of the prediction error is:

[0028] e(t)=|y(t|t)-y m (t-1)|,

[0029]

[0030] Among them, e(t) is the model prediction error at the current time t, y(t|t) and y m (t-1) are the actual value and model prediction value at the current moment, E is the average error, and when the set threshold is reached, the prediction model parameters are updated.

[0031] A second aspect of the present invention provides a power data release path planning system based on load balancing, comprising:

[0032] A current transmission status acquisition module is used to obtain the data transmission status of each path at the current moment;

[0033] A transmission state prediction module is used to obtain a predicted value of the data transmission state of each path at the next moment based on the data transmission state of each path at the current moment and the constructed short-term multi-path data transmission state prediction model, and use the predicted value of the data transmission state of each path at the next moment as the input of the short-term multi-path data transmission state prediction model to iteratively obtain a predicted value of the multi-path data transmission state within a specified time period;

[0034] The path planning module is used to perform multi-step predictive control on multi-path selection based on the predicted value of multi-path data transmission state within a specified time period, with the minimum overall data transmission time as the objective function, and path selection as a constraint condition. The MPC rolling optimization is adopted to select each path selection sequence at the first moment in the obtained optimal path path sequence as the path selection transmission operation at the current moment, and the operation at the current moment is applied to the data transmission path selection at the next moment until the selection operation of each path transmission at each moment within the specified time period is obtained.

[0035] A third aspect of the present invention provides a computer-readable storage medium.

[0036] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the above-mentioned method for planning a power data release path based on load balancing.

[0037] A fourth aspect of the present invention provides a computer device.

[0038] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for planning a power data release path based on load balancing as described above are implemented.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention combines LSTM prediction with MPC, and uses LSTM to build a multi-path data transmission state prediction model. Based on the current data transmission selection operation, the data transmission state prediction value of each path in the future period is obtained, so that the data release time in the future period under different selection operations can be predicted according to any data transmission selection operation at the current moment, thereby providing a data basis for the comparison of the advantages and disadvantages of the data transmission selection operation obtained by MPC;

[0041] 2. According to the predicted value of the data transmission status of each path in the future period obtained by the LSTM model, the minimum data release time is set as the objective function, and the MPC is used to determine the optimal data transmission selection operation of each path at the current moment. The data transmission selection operation of each path at the next moment is updated through online rolling optimization. The predicted value of the LSTM model is corrected through feedback from the comparison of actual data and predicted data to ensure that the data transmission selection operation of each path determined by the minimum data release time is met at each moment, thereby realizing the optimization of multi-path transmission strategy for full data release and ensuring the transmission load balance of each path.

[0042] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0044] Figure 1 This is a diagram of the architecture of the power consumption information collection system provided by an embodiment of the present invention.

[0045] Figure 2 It is a diagram of the overall model architecture provided by an embodiment of the present invention;

[0046] Figure 3 LSTM structure diagram provided by an embodiment of the present invention;

[0047] Figure 4 is a schematic diagram of an MPC provided by an embodiment of the present invention;

[0048] Figure 5 is a diagram of an MPC control structure provided by an embodiment of the present invention;

[0049] Figure 6 This is the analysis result of the ablation experiment provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0051] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0052] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0053] The architecture diagram of the electricity consumption information collection system is as follows Figure 1 As shown, the communication front-end 2-4 has the same data publishing path as the communication front-end 1. In order to avoid unclear images due to too many connecting lines, the data publishing path connecting lines of the communication front-end 2-4 are omitted here.

[0054] In the process of synchronous transmission and release of full data, data transmission processing includes four modules: terminal communication, data collection, data processing, and data storage. Among them, the terminal communication module is responsible for communication instruction processing and link maintenance; the data collection module is responsible for data analysis; the data processing module is responsible for file management and data persistence; the data storage module is responsible for storing the full amount of collected data in the marketing business database. However, the data transmission / processing rates of the above four modules are not consistent, so it is necessary to plan the release path of the full amount of collected data based on the load balancing strategy, so as to support the access of tens of millions of devices, shorten the data release cycle, and support the big data analysis of various marketing businesses based on the electricity information collection data.

[0055] In view of the shortcomings of the existing multi-path transmission methods, such as poor adaptability to dynamic network environment, slow transmission path switching and load imbalance, the present invention proposes a multi-path transmission strategy optimization method based on LSTM prediction combined with MPC. First, LSTM is used to construct a multi-path data transmission state prediction model to obtain the data transmission state prediction value of each path at the next moment; then, the prediction value at the next moment is input into the LSTM model to obtain the data transmission prediction amount for a period of time in the future; finally, based on the set minimum release time as the objective function, the rolling optimization and feedback correction of MPC are adopted to obtain the transmission control amount of each path that meets the objective function. The comparative experiments with multiple baseline methods show that the method proposed in the present invention can realize the optimization of multi-path transmission strategy for full data release. Compared with the current mainstream traffic-aware scheduling method, it improves 8.1%, 11.7%, 10.1%, 9.7% and 13.0% in five evaluation indicators respectively, further shortening the release time of full meter reading data and ensuring the stability and security of meter reading data release operation.

[0056] Specifically, there are two improvements:

[0057] (1) A short-term multi-path data transmission status prediction model based on LSTM is proposed, which can use the data transmission status at the current moment to obtain the predicted value of the transmission status of each path at the next moment.

[0058] (2) It is proposed to combine LSTM prediction with MPC. According to the prediction value of the LSTM model, the minimum release time is set as the objective function. The rolling optimization and feedback correction of MPC are used to obtain the transmission control quantity of each path that meets the objective function, thereby realizing the optimization of multi-path transmission strategy for full data release.

[0059] Embodiment 1

[0060] like Figure 2 As shown, this implementation provides a power data release path planning method based on load balancing, comprising the following steps:

[0061] Step 1: Based on the data transmission status of each path at the current moment and the constructed short-term multi-path data transmission status prediction model, obtain the predicted value of the data transmission status of each path at the next moment;

[0062] In this embodiment, in order to predict the transmission delay and packet loss rate of each path at the future moment, this embodiment uses LSTM to build a short-term multi-path data transmission state prediction model. The structure of the LSTM model is as follows: Figure 3 The input of the LSTM model is the bandwidth b of the i-th transmission path at the t-th time. i (t), the transmission delay d of the i-th path at time t i (t) and the packet loss rate l of the ith path at time t i (t), the output is the transmission delay d at time t+1 i (t+1) and packet loss rate l i (t+1);

[0063] LSTM is a special recurrent neural network (RN) structure that solves the gradient vanishing and gradient exploding problems that occur in traditional RNNs when processing long sequences. LSTM contains a cell state and three gating mechanisms: input gate, forget gate, and output gate. These gating mechanisms allow the network to selectively store, update, and output information, allowing LSTM to better handle long-term dependencies of sequences. The formula for the LSTM prediction model is as follows:

[0064] x t =[b i (t),d i (t),l i (t)] (1),

[0065] f t =σ(W f ·[h t-1 ,x t ]+b f ) (2),

[0066] i t =σ(W i ·[h t-1 ,x t ]+b i ) (3),

[0067]

[0068] O t =σ(W O ·[h t-1 ,x t ]+b o ) (6),

[0069] h t =O t tanh(C t ) (7),

[0070] Among them, f t 、i t , o t They are forget gate, input gate and output gate respectively, C t is the cell state, h t-1 is the hidden state (based on C t Get), x t is the input vector at the current time t, which contains the bandwidth, delay and packet loss rate information of each path.

[0071] Step 2: Taking the predicted value of the data transmission state of each path at the next moment as the input of the short-term multi-path data transmission state prediction model, iteratively obtains the predicted value of the multi-path data transmission state within a specified time period;

[0072] Specifically include:

[0073] Step 201: predict the transmission delay and packet loss rate of each path at the next moment based on the bandwidth, transmission delay and packet loss rate data of each path at the current moment and the constructed short-term multi-path data transmission state prediction model;

[0074] Step 202: Calculate the transmission time per unit time of each path at the next moment based on the transmission delay and packet loss rate of each path at the next moment;

[0075] The specific calculation formula is:

[0076] The transmission delay d of each path at time t+1 predicted by the LSTM model i (t+1) and packet loss rate l i (t+1), to calculate the unit time transmission time T of each path i (t+1):

[0077]

[0078] Among them, MSS is the maximum segment length, RTT i (t+1) is the round trip time, which is numerically equal to 2d i (t+1).

[0079] Step 203: input the transmission delay and packet loss rate prediction values ​​of each path at the next moment into the short-term multi-path data transmission state prediction model, and repeat the iteration to obtain the transmission time per unit moment of each path within the future specified time period.

[0080] Step 3: Taking the minimum overall data transmission time as the objective function, based on the predicted amount of multi-path data transmission in the future, adopt MPC rolling optimization and feedback correction to obtain the data transmission selection operation of each path at the next moment that meets the objective function. Repeat this process iteratively to achieve multi-path transmission strategy optimization for full data release.

[0081] The principle diagram of model predictive control is as follows: Figure 4 As shown, assuming that the current time is t, in the prediction sequence, the prediction length is p, and in the control sequence, the control length is m.

[0082] The reference trajectory is the expected data transmission volume; the predicted output is the data transmission volume predicted at time t for the next p moments based on the executed control volume; Figure 4 It can be seen that the predicted output cannot reach the data transmission volume of the reference trajectory at time t. The data transmission volume of each transmission path is adjusted by the predicted control amount, so that the predicted output and the reference trajectory can coincide or have the minimum difference at the future time p.

[0083] Among them, from time t to time t+m is a control time domain, at this time, n paths can choose to transmit or not transmit. There are 2 n There are two paths for transmission operation control, so from time t+m to time t+p, (pm) moments form 2 n ×(pm) path selection transmission operations, the vertical time axis is represented by a predicted control quantity time series set composed of pm path selection transmission operations, a set of path selection transmission operation control sets that can obtain the maximum cumulative data transmission volume from time t to time t+p is selected as the optimal path data transmission selection operation, and each path selection transmission operation in the first control time domain of the optimal path data transmission selection operation is implemented, and the above process is repeated according to the implementation result;

[0084] Among them, the predicted control amount is the path data transmission prediction amount output by the LSTM model from time t+1 to time t+m, obtained according to the path selection transmission operation control made at time t;

[0085] The sampling time refers to a moment; the control time domain is multiple sampling times. This is because the short-term multipath data transmission state prediction model can output the prediction value of m moments in the future at one time, so multiple moments constitute a control time domain.

[0086] Among them, rolling optimization, that is, the path data transmission selection operation is repeatedly performed online. At each sampling moment, MPC determines the optimal path data transmission selection operation at the current moment based on the multi-path data transmission prediction amount for a period of time in the future obtained by the LSTM model. This step only involves a short period of time from this moment to the next moment, and at the same time, this future period will move forward at the same time; at the same time, according to the data release time of the current moment obtained after the actual path data transmission selection operation, the LSTM model prediction error is calculated, and then the prediction error is used to correct the prediction value of the LSTM model, thereby realizing feedback correction. Model predictive control can be divided into three steps: prediction model, rolling optimization and feedback correction. The prediction model is used to estimate the future behavior of the system. The present invention uses LSTM to construct a short-term multi-path data transmission state prediction model. The rolling optimization is continuously updated online to ensure the optimality of the control, and the feedback correction can compensate or adjust the errors that may occur in the future, thereby forming a closed-loop control system. The model predictive control algorithm in this article transforms the multi-path selection problem into a problem of solving the optimal value of the objective function. The control block diagram is shown in Figure 5 .

[0087] The multi-path selection model predictive control takes the minimum overall transmission time as the objective function and optimizes the path selection. The MPC objective function is as follows:

[0088]

[0089] Among them, T total (k) is the total transmission time predicted in the kth step, α is the weight coefficient of different paths, and P is the total number of predicted steps;

[0090] The path selection constraints are:

[0091]

[0092] Among them, x i (k) indicates whether the i-th path is selected in the k-th step, 1 indicates selection, and 0 indicates no selection. The path selection constraint is controlled by the predicted value of the transmission time per unit moment of the path at the next moment, and n indicates the total number of paths.

[0093] The data volume of each path transmission at the next moment is iteratively input into the short-term multi-path data transmission state prediction model, and M path selection sequences for the next P steps are output. The objective function value within the P steps is calculated to obtain the optimal path selection sequence for the next moment, and the first step transmission path selection of the optimal path selection sequence is used as the multi-path data transmission selection scheme for the next moment.

[0094] Each time the current state is updated, the prediction model uses the latest state as input for calculation, and this cycle repeats. However, over time, the output of the offline LSTM may gradually deviate from the actual situation, and the model needs to be retrained.

[0095] To address this problem, this embodiment designs a feedback correction mechanism in a control system.

[0096] e(t)=|y(t|t)-y m (t-1)| (11),

[0097]

[0098] Among them, e(t) is the model prediction error at the current time t, y(t|t) and y m (t-1) are the actual value and model prediction value at the current moment, E is the average error, and when the set threshold is reached, the prediction model parameters are updated.

[0099] To sum up, the specific execution steps of the MPC algorithm are as follows:

[0100] 1) System initialization: collect the bandwidth, delay and packet loss rate data of each path and determine the initial state of the system.

[0101] 2) Input the data information collected at the current moment into the LSTM model, predict the delay and packet loss rate at the next moment, and then calculate the transmission time per unit moment of each path.

[0102] 3) Repeat step 2) to calculate the transmission time per unit time of each path in the future prediction time domain N at the current time t.

[0103] 4) Taking the minimum overall transmission time and the minimum packet loss rate as the objective function, the multi-path selection is subject to P-step predictive control. Each step has two path selection modes, namely transmission or non-transmission, for a total of 2 PP Group operation sequence.

[0104] 5) Apply each path selection operation at the first moment in the operation sequence that satisfies the minimum overall transmission time to the data transmission route selection operation at the next moment, and return to 1) to repeat the above steps.

[0105] Experimental Section

[0106] The experimental background is that a city company's electricity consumption information collection system transmits full meter reading data to the marketing business application system. In order to fully support the release of full meter reading electricity consumption data of tens of millions of users every day, it is necessary to continuously optimize the execution efficiency of full data release and ensure the timeliness of meter reading data release. The sample data set comes from the full data transmitted from the company's electricity consumption information collection system to the marketing business application system in 2023. Each data transmission lasts about 1 hour, and the monitoring time interval of various indicators is 5 minutes. The specific experimental environment configuration is as follows:

[0107] Currently, the number of access connections to communication front-end machines can reach 250,000, and the maximum average communication traffic is 8.8Mb / s.

[0108] It takes about 10ms for the collection front-end to process a heartbeat data frame and about 20ms for a historical data frame. Therefore, the collection front-end can process a maximum of 3200 heartbeat frames or 1600 historical data frames per second.

[0109] The data processing front-end can process an average of approximately 40,000 data items per second, with a maximum IO read and write speed of 17w / s for the distributed database.

[0110] The experimental data scale configuration is shown in Table 1.

[0111] Table 1 Experimental data scale configuration

[0112]

[0113] Data collection situation: Each link sends a heartbeat every 5 minutes, and each electric energy meter sends the voltage, current, power, power factor and electric energy indicator curve at one point every 5 minutes, and daily frozen data is sent once a day.

[0114] This experiment selected five evaluation indicators to evaluate the model performance: average communication traffic, number of heartbeat frames processed per second, amount of data processed per second, IO read and write, and transmission time. The model in this article is written in Python program, and the model is constructed with the help of environment libraries such as numpy and Pytorch. The experiment is divided into comparative experiments and ablation experiments. Through these indicators and comparisons, the optimization effect of the proposed model in the multi-path transmission scenario for full data release can be comprehensively evaluated.

[0115] In the comparative experiment, the LSTM-MPC model proposed in the present invention is compared with the following models:

[0116] Round Robin is a fair transmission path allocation algorithm that allocates resources to multiple tasks or paths in a fixed order. In essence, round robin selects each available path for data transmission in order, without considering the status or load of the current transmission path.

[0117] 1) Shortest Path First (SPF) is a data transmission path selection strategy that selects the path with the lowest transmission delay for data transmission without considering other factors such as bandwidth or path elasticity.

[0118] 2) Traffic-Aware Scheduling is a dynamic scheduling strategy that monitors and analyzes the traffic load of each transmission path in real time and intelligently selects the most suitable path for data transmission to optimize network performance, reduce latency, and avoid path overload.

[0119] Table 2 Experimental results

[0120]

[0121] The experimental results are shown in Table 2. The model of the present invention performs best in all evaluation indicators. Polling scheduling and shortest path priority perform poorly compared to flow-aware scheduling. This is because they do not take into account the dynamically changing multi-path transmission environment, which leads to some paths being overloaded while other paths being idle. They cannot effectively balance the multi-path transmission load and balance resource utilization. In contrast, flow-aware scheduling can make more intelligent path selections based on the current network conditions. The model constructed by the present invention predicts the optimal transmission performance that can be achieved by multiple paths under the state of all adjustable transmission amounts in multiple steps in the future, thereby selecting the transmission amount of each currently controllable path, and adjusting the transmission control input of each path with iterative optimization prediction feedback, thereby foreseeing and avoiding potential network congestion in advance, while flow-aware scheduling lacks this ability. Compared with flow-aware scheduling, the method of the present invention improves the five indicators by 8.1%, 11.7%, 10.1%, 9.7% and 13.0% respectively.

[0122] The present invention uses MPC to optimize the multi-path transmission control of the full data release, which can take into account future states and constraints. This method makes up for the shortcomings of LSTM in adapting to dynamic network environments and actively avoiding congestion. In order to verify the role of the MPC module in the model, the following ablation experiment is conducted. The proposed multi-path selection strategy based on transmission state prediction control is compared with the multi-path selection strategy without MPC. In order to make a fair comparison, when conducting the ablation experiment, the parameters of other parts are adjusted to the best, and the results are selected from the same training period. The experimental results are shown in Figure 6 As shown in the figure, the model in this paper is superior to the multi-path selection strategy that only uses LSTM without combining MPC in all indicators. This proves that MPC can indeed significantly improve the model performance. It shows that the model in this paper has the ability to improve the efficiency of full power data transmission, can better cope with complex network environments, and achieve more optimized path selection.

[0123] Embodiment 2

[0124] This implementation provides a power data release path planning system based on load balancing, including:

[0125] A current transmission status acquisition module is used to obtain the data transmission status of each path at the current moment;

[0126] A transmission state prediction module is used to obtain a predicted value of the data transmission state of each path at the next moment based on the data transmission state of each path at the current moment and the constructed short-term multi-path data transmission state prediction model, and use the predicted value of the data transmission state of each path at the next moment as the input of the short-term multi-path data transmission state prediction model to iteratively obtain a predicted value of the multi-path data transmission state within a specified time period;

[0127] The path planning module is used to perform multi-step predictive control on multi-path selection based on the predicted value of multi-path data transmission state within a specified time period, with the minimum overall data transmission time as the objective function, and path selection as a constraint condition. The MPC rolling optimization is adopted to select each path selection sequence at the first moment in the obtained optimal path path sequence as the path selection transmission operation at the current moment, and the operation at the current moment is applied to the data transmission path selection at the next moment until the selection operation of each path transmission at each moment within the specified time period is obtained.

[0128] Embodiment 3

[0129] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the power data release path planning method based on load balancing as described above are implemented.

[0130] Embodiment 4

[0131] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the load balancing-based power data release path planning method as described above are implemented.

[0132] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0133] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0136] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power data release path planning method based on load balancing, characterized in that: The steps include: Get the data transmission status of each path at the current moment; Based on the data transmission status of each path at the current moment and the constructed short-term multi-path data transmission status prediction model, the data transmission status prediction value of each path at the next moment is obtained, and the data transmission status prediction value of each path at the next moment is used as the input of the short-term multi-path data transmission status prediction model, and the multi-path data transmission status prediction value within the specified time period is iteratively obtained; Taking the minimum overall data transmission time as the objective function and path selection as the constraint condition, the MPC rolling optimization is adopted based on the predicted value of the multi-path data transmission state within the specified time period to perform multi-step predictive control on the multi-path selection. The path selection sequence of the first moment in the obtained optimal path selection path sequence is used as the path selection transmission operation of each moment, and the operation of the current moment is applied to the data transmission path selection at the next moment, until the selection operation of each path transmission at each moment in the specified time period is obtained.

2. The method for planning a power data release path based on load balancing according to claim 1, characterized in that: The calculation process of the multipath data transmission state prediction value for a specified time period includes: According to the bandwidth, transmission delay and packet loss rate data of each path at the current moment and the constructed short-term multi-path data transmission state prediction model, the transmission delay and packet loss rate of each path at the next moment are predicted; The transmission time per unit time of each path at the next moment is calculated based on the transmission delay and packet loss rate of each path at the next moment; The transmission delay and packet loss rate prediction values ​​of each path at the next moment are input into the short-term multipath data transmission state prediction model, and the unit moment transmission time of each path within the specified time period is obtained by repeated iteration.

3. The method for planning a power data release path based on load balancing according to claim 2, characterized in that: The calculation formula of the transmission time per unit time of each path at the next moment is: Among them, MSS is the maximum segment length, RTT i (t+1) is the round trip time, which is numerically equal to 2d i (t+1); d i (t+1) and l i (t+1) is the transmission delay and packet loss rate of each path at time t+1.

4. The method for planning a power data release path based on load balancing according to claim 1, characterized in that: The objective function is: The constraints are: x i (k)∈{0,1}, Among them, T total (k) is the total transmission time predicted in the kth step, α is the weight coefficient of different paths, P is the total number of predicted steps, x i (k) indicates whether the i-th path is selected in the k-th step, 1 indicates selection, 0 indicates no selection, and n indicates the total number of paths.

5. The method for planning a power data release path based on load balancing according to claim 1, characterized in that: The short-term multipath data transmission status prediction model adopts the LSTM model.

6. The method for planning a power data release path based on load balancing according to claim 1, characterized in that: After obtaining the actual path data transmission selection operation, the current data release time is obtained, the prediction error of the short-term multi-path data transmission state prediction model is calculated, and then the prediction error is used to correct the prediction value of the model, thereby realizing feedback correction.

7. The method for planning a power data release path based on load balancing according to claim 6, characterized in that: The calculation formula for the prediction error is: e(t)=|y(t|t)-y m (t-1)|, Among them, e(t) is the model prediction error at the current time t, y(t|t) and y m (t-1) are the actual value and model prediction value at the current moment, E is the average error, and when the set threshold is reached, the prediction model parameters are updated.

8. The power data release path planning system based on load balancing is characterized by: include: A current transmission status acquisition module is used to obtain the data transmission status of each path at the current moment; A transmission state prediction module is used to obtain a predicted value of the data transmission state of each path at the next moment based on the data transmission state of each path at the current moment and the constructed short-term multi-path data transmission state prediction model, and use the predicted value of the data transmission state of each path at the next moment as the input of the short-term multi-path data transmission state prediction model to iteratively obtain a predicted value of the multi-path data transmission state within a specified time period; The path planning module is used to perform multi-step predictive control on multi-path selection based on the predicted value of multi-path data transmission state within a specified time period, with the minimum overall data transmission time as the objective function, and path selection as a constraint condition. The MPC rolling optimization is adopted to select each path selection sequence at the first moment in the obtained optimal path path sequence as the path selection transmission operation at the current moment, and the operation at the current moment is applied to the data transmission path selection at the next moment until the selection operation of each path transmission at each moment within the specified time period is obtained.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the power data release path planning method based on load balancing as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the power data release path planning method based on load balancing are implemented as described in any one of claims 1-7.

Images