Remote intelligent transmission method and system for large-volume fault data of electricity utilization information acquisition terminal

Abstract

The embodiment of the invention discloses a remote intelligent transmission method and system for large-volume fault data of an electricity utilization information acquisition terminal, and the transmission method comprises the steps: carrying out the data transmission based on a current transmission rate, and obtaining the data receiving state information of a transmission channel; further acquiring an adjusted sending rate and a predicted sending rate, and selecting a target sending rate corresponding to the current round; performing data transmission based on the target sending rate corresponding to the current round to obtain corresponding data receiving state information when the transmission channel performs data transmission at the target sending rate of the current round, and obtaining a reward value of the target sending rate of the current round so as to train the current prediction model; and taking the target sending rate as the current sending rate, taking the trained prediction model as the current prediction model, and re-obtaining the adjusted sending rate and the predicted sending rate of the next round. And bandwidth resources are fully excavated and utilized, so that high efficiency and stability of large-volume fault data transmission are realized.

Description

Technical Field

[0001] Multiple embodiments of the present specification relate to the technical field of meter data transmission, and specifically to a method and system for remote intelligent transmission of large-volume fault data from an electricity consumption information collection terminal. Background Art

[0002] The electricity consumption information collection system is a key component of the smart grid, which mainly includes a data platform and an electricity consumption information collection terminal. The data platform is responsible for collecting, processing and analyzing user electricity consumption data collected in real time by the electricity consumption information collection terminal.

[0003] SFTP (SSH File Transfer Protocol) has become one of the preferred protocols for file transfer in an insecure network environment due to its high security and convenient file management functions. The traditional power consumption information collection system architecture is to obtain large-volume fault data from the power consumption information collection terminal through SFTP. However, there are many technical defects in the use of SFTP to transmit large-volume fault data. The communication bandwidth will continue to be occupied during the entire transmission cycle of the large-volume fault data. If the transmission rate of the power consumption information collection terminal exceeds the communication bandwidth, it will cause obstacles to business data interaction and delays in business data reporting. In extreme network environments, business data may even fail to be reported, seriously affecting the completeness of the power consumption data collected by the power consumption information collection system. Summary of the invention

[0004] The embodiments of this specification provide a method and system for remote intelligent transmission of large-volume fault data of an electricity consumption information collection terminal. Through the deep fusion of multi-dimensional information, combined with artificial intelligence and deep reinforcement learning technology, active dynamic management and accurate prediction of transmission channels are achieved. Based on the basic principle of not affecting business transmission, bandwidth resources are fully mined and utilized, thereby achieving high efficiency and stability of large-volume fault data transmission. Its technical solution is as follows:

[0005] In a first aspect, the embodiments of this specification provide a method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal, comprising the following steps:

[0006] S1. Perform data transmission based on the current transmission rate, and obtain data reception state information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception state information includes reception rate, data packet delay, and data packet loss rate;

[0007] S2, based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, the sending rate is adjusted, and based on the current prediction model and the data receiving state information corresponding to the current sending rate, the predicted sending rate is obtained;

[0008] S3, based on all historical adjusted sending rates, all historical predicted sending rates, and the historical average receiving rate, select one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round;

[0009] S4, performing data transmission based on the target transmission rate corresponding to the current round, so as to obtain data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round;

[0010] S5. Based on the data reception state information corresponding to the transmission channel when data is transmitted at the target transmission rate corresponding to the current round, obtain a reward value for transmitting data at the target transmission rate corresponding to the current round under the data reception state information of the current round;

[0011] S6, based on the data receiving state information corresponding to the current round and all the rounds before the current round totaling a preset number, the target sending rate, and the reward value of the target sending rate, the current prediction model is trained to obtain a trained prediction model;

[0012] S7. Use the target sending rate corresponding to the selected current round as the current sending rate and use the trained prediction model as the current prediction model and return to step S2.

[0013] As a preferred solution, the adjusting the sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round includes:

[0014] Obtaining a delay gradient based on the data packet delays corresponding to the data reception state information corresponding to the current transmission rate and the data reception state information corresponding to the previous round of the current round;

[0015] Acquire the network status of the current round based on the delay gradient, where the network status includes under-use, over-use, and normal;

[0016] The current sending rate is adjusted based on the network status of the current round to obtain an adjusted sending rate.

[0017] As a preferred solution, the adjusting the current sending rate based on the network status of the current round to obtain the adjusted sending rate includes:

[0018] When the network status is underutilized, the sum of the current sending rate and the preset increment is used as the adjusted sending rate;

[0019] When the network status is overused, the sending rate is adjusted to half of the current sending rate;

[0020] When the network status is normal, the current sending rate is used as the adjusted sending rate.

[0021] As a preferred solution, the selecting, based on all historical adjusted sending rates, all historical predicted sending rates, and all historical receiving rates, one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round includes:

[0022] Obtaining a first error based on a historical average receiving rate and all historical adjusted sending rates;

[0023] Obtaining a second error based on a historical average receiving rate and all historical predicted sending rates;

[0024] When the absolute difference between the first error and the second error is greater than a preset threshold, the adjusted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round;

[0025] When the absolute difference between the first error and the second error is less than or equal to a preset threshold, the predicted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round.

[0026] As a preferred solution, the step S1 further includes a step of selecting a preset threshold, and the selection of the preset threshold includes:

[0027] Using a plurality of different thresholds as preset thresholds to perform data transmission based on steps S1-S7 respectively;

[0028] For any of multiple thresholds as the target threshold:

[0029] Based on the target threshold, the target sending rate corresponding to each round in the test round, the data receiving status information corresponding to the last round in the test round and the historical average receiving rate under the target threshold are obtained to obtain the score value of the target threshold;

[0030] Repeat the above process until the score values ​​corresponding to all thresholds are obtained;

[0031] The threshold corresponding to the maximum score value is selected as the preset threshold.

[0032] As a preferred solution, the reward value r for performing data transmission at the target transmission rate corresponding to the current round under the data reception status information of the current round is t , specifically:

[0033]

[0034] Among them, c t ,d t , lt are the receiving rate, packet delay, and packet loss rate in the data receiving status information corresponding to the target transmission rate of the transmission channel in the current round, respectively, linear-to-log (c t ) means to convert c t Function to convert from linear scale to logarithmic scale.

[0035] As a preferred solution, the predicted sending rate is obtained based on the current prediction model and the data receiving state information corresponding to the current sending rate, specifically:

[0036] The data receiving state information corresponding to the current sending rate is used as the model input of the current prediction model, and the predicted sending rate is obtained based on the maximized advantage function.

[0037] As a preferred solution, the training of the current prediction model to obtain a trained prediction model based on the data reception state information corresponding to the current round and all rounds of a preset number before the current round, the target sending rate, and the reward value of the target sending rate includes:

[0038] Based on the value function in the advantage function of the current prediction model, obtain the expected return of the current round;

[0039] The value function is updated based on the data receiving state information corresponding to the current round and all the rounds before the current round, a preset number of rounds, the target sending rate and the reward value of the target sending rate, the expected return of the current round and the minimized loss function, and the value error is calculated;

[0040] The value function is updated based on the value error to train the current prediction model to obtain a trained prediction model.

[0041] As a preferred solution, the value function in the maximization advantage function in the current prediction model is used to obtain the expected return of the prediction output of the current prediction model, and the expected return is calculated by a k-step time difference method, where k = a preset number.

[0042] In a second aspect, the embodiments of this specification provide a remote intelligent transmission system for large-volume fault data of an electricity consumption information collection terminal, including:

[0043] Data collection module, hybrid prediction module, adaptive switching module and learning module;

[0044] The data collection module performs data transmission based on the current transmission rate, and obtains data reception status information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception status information includes reception rate, data packet delay, and data packet loss rate;

[0045] The hybrid prediction module obtains an adjusted sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, and obtains a predicted sending rate based on the current prediction model and the data receiving state information corresponding to the current sending rate;

[0046] The adaptive switching module selects one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round based on all historical adjusted sending rates, all historical predicted sending rates and the historical average receiving rate;

[0047] The data collection module performs data transmission based on the target transmission rate corresponding to the current round to obtain the corresponding data receiving state information when the transmission channel performs data transmission at the target transmission rate corresponding to the current round.

[0048] The learning module obtains a reward value for performing data transmission at the target transmission rate corresponding to the current round under the data reception state information of the current round based on the data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round; trains the current prediction model based on the data reception state information, the target transmission rate, and the reward value of the target transmission rate corresponding to all rounds including the current round and a total of a preset number of rounds before the current round to obtain a trained prediction model;

[0049] The hybrid prediction module uses the target sending rate corresponding to the selected current round as the current sending rate and uses the trained prediction model as the current prediction model to re-acquire the adjusted sending rate and the predicted sending rate.

[0050] In the third aspect, an embodiment of this specification provides an electronic device, including a processor and a memory; the processor is connected to the memory; the memory is used to store executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the steps described in the first aspect of the above embodiment.

[0051] In a fourth aspect, an embodiment of this specification provides a computer storage medium, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the steps described in the first aspect of the above embodiment.

[0052] The beneficial effects brought by the technical solutions provided by some embodiments of this specification include at least:

[0053] On the one hand, the bandwidth of the transmission channel is estimated based on the data reception status information to obtain the adjusted sending rate. On the other hand, the prediction model is trained and the bandwidth prediction is performed to output the predicted sending rate. The more accurate one of the two is selected as the new sending rate for data transmission, realizing active dynamic management and accurate prediction of the transmission channel. Adjusting the sending rate provides a fast and stable initial prediction, while the predicted sending rate provides a more accurate prediction when there is sufficient data, thereby fully tapping and utilizing bandwidth resources without affecting the basic principle of business transmission, thereby achieving high efficiency and stability of large-volume fault data transmission.

[0054] The prediction model is gradually trained during the data transmission process in the transmission channel, which ultimately enables the model to maintain a high prediction accuracy even in the face of complex network conditions, thereby fully utilizing bandwidth resources and achieving high efficiency and stability in the transmission of large-volume fault data.

[0055] Through deep reinforcement learning, bandwidth prediction is made intelligent and automated. The prediction model can continuously learn and optimize during operation to adapt to the ever-changing network environment, thereby continuously providing high-quality bandwidth prediction services. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a flowchart of a method for remote intelligent transmission of large-volume fault data of an electricity consumption information collection terminal provided in an embodiment of this specification;

[0058] Figure 2 is a schematic diagram of the prediction model architecture provided by the embodiments of this specification;

[0059] Figure 3 It is a structural diagram of a remote intelligent transmission system for large-volume fault data of an electricity consumption information collection terminal provided in an embodiment of this specification;

[0060] Figure 4 It is a structural schematic diagram of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of this specification will be described clearly and completely below in conjunction with the drawings in the embodiments of this specification.

[0062] The terms "first", "second", "third", etc. in the description and claims of this specification and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0063] The following description provides examples and does not limit the scope, applicability or examples set forth in the claims. Changes may be made to the functions and arrangements of the elements described without departing from the scope of the present specification. Various processes or components may be appropriately omitted, substituted or added to each example. For example, the described method may be performed in an order different from the order described, and various steps may be added, omitted or combined. In addition, features described with respect to some examples may be combined in other examples.

[0064] In the architecture of the traditional power consumption information collection system, there are many technical defects in using SFTP to obtain large-volume fault data from the terminal; when the SFTP method is used to transmit large-volume fault data, the communication bandwidth will be continuously occupied during the entire transmission cycle, resulting in obstacles to normal business data interaction, delays in reporting of important business data, and even data failures in extreme cases, which seriously affects the power consumption information collection system's demand for completeness of power consumption data. Therefore, it is necessary to predict the real-time bandwidth of the transmission channel so as to adjust the transmission rate of the power consumption information collection terminal to be less than the real-time bandwidth of the transmission channel, so as to achieve the stability of large-volume fault data transmission. At the same time, the more accurate the prediction of the real-time bandwidth of the transmission channel, the more likely it is that the transmission rate of the power consumption information collection terminal can be adjusted to be as close to the real-time bandwidth of the transmission channel as possible, thereby making full use of the bandwidth resources of the transmission channel and achieving high efficiency of large-volume fault data transmission.

[0065] Reference Figure 1 As shown, Figure 1 A flowchart of a method for remote intelligent transmission of large-volume fault data of an electricity consumption information collection terminal provided in one embodiment of this specification may at least include the following steps:

[0066] S1. Perform data transmission based on the current transmission rate, and obtain data reception state information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception state information includes reception rate, data packet delay, and data packet loss rate;

[0067] S2, based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, the sending rate is adjusted, and based on the current prediction model and the data receiving state information corresponding to the current sending rate, the predicted sending rate is obtained;

[0068] S3, based on all historical adjusted sending rates, all historical predicted sending rates, and the historical average receiving rate, select one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round;

[0069] S4, performing data transmission based on the target transmission rate corresponding to the current round, so as to obtain data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round;

[0070] S5. Based on the data reception state information corresponding to the transmission channel when data is transmitted at the target transmission rate corresponding to the current round, obtain a reward value for transmitting data at the target transmission rate corresponding to the current round under the data reception state information of the current round;

[0071] S6, based on the data receiving state information corresponding to the current round and all the rounds before the current round totaling a preset number, the target sending rate, and the reward value of the target sending rate, the current prediction model is trained to obtain a trained prediction model;

[0072] S7. Use the target sending rate corresponding to the selected current round as the current sending rate and use the trained prediction model as the current prediction model and return to step S2.

[0073] Illustratively, statistical data is continuously collected during data transmission, and preprocessing operations are performed on the collected statistical data, including feature selection, outlier processing, time window division, etc. The above preprocessing operations are commonly used methods in data processing and will not be repeated here.

[0074] Explanatory, each time window corresponds to a round, the current sending rate is the sending rate of the transmission channel in the current time window, and when the current time window ends, the data receiving status information reflecting the network status in the current time window can be obtained. The data receiving status information includes the receiving rate, data packet delay, and data packet loss rate. Each time window corresponds to a sending rate and data receiving status information.

[0075] After obtaining the data reception status information, on the one hand, the bandwidth of the transmission channel is estimated to obtain the adjusted sending rate; on the other hand, the predicted sending rate is output through processing by the prediction model, and the more accurate one of the two is selected as the target sending rate for the current time window, that is, the sending rate of the transmission channel in the next time window, to achieve active dynamic management and accurate prediction of the transmission channel.

[0076] The data reception status information corresponding to data transmission at the target sending rate of the current time window in the next time window of the current time window is used as the basis for evaluating the accuracy of the target sending rate selected for the current time window, and the reward value of the target sending rate selected for the current time window is calculated.

[0077] The data receiving status information, target sending rate and reward value of the target sending rate in the current time window are taken as a set of sample data, and the sample data corresponding to a preset number of time windows before the current time window and including the current time window are taken as the basis for optimizing the current prediction model.

[0078] The target sending rate corresponding to the current time window is used as the current sending rate and the trained prediction model is used as the current prediction model, and the above steps are repeated with the next time window of the current time window as the new current time window. In this way, the prediction model is continuously trained to gradually make the sending rate of the power consumption information collection terminal approach the available bandwidth of the transmission channel that changes in real time, and finally achieve the effect of making full use of the bandwidth resources of the transmission channel and realize the stability and efficiency of large-volume fault data transmission.

[0079] Exemplarily, at the beginning of data transmission, due to the small amount of sample data, the prediction model cannot make accurate predictions. Therefore, data transmission is performed at a preset starting transmission rate, and the adjusted transmission rate is used as the target transmission rate of each time window for continuous data transmission. In each time window, the adjusted transmission rate and the predicted transmission rate of the time window are compared and selected, and the deviations of the two from all historical receiving rates estimated using network tools are calculated respectively, and the training of the prediction model is judged according to the deviations of the two. When the training of the prediction model is considered to be relatively mature, the predicted transmission rate is used as the target transmission rate of each subsequent time window for continuous data transmission. When the network state of a certain time window changes significantly, it is judged that the prediction model has not been trained in place under the changed network state according to the deviation of the two, so that the adjusted transmission rate is selected again as the target transmission rate of each subsequent time window for continuous data transmission. After repeated selection and replacement, the prediction model can be trained to have accurate prediction capabilities for various complex network conditions.

[0080] In one embodiment of the present specification, the adjusting the sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round includes:

[0081] Obtaining a delay gradient based on the data packet delays corresponding to the data reception state information corresponding to the current transmission rate and the data reception state information corresponding to the previous round of the current round;

[0082] Acquire the network status of the current round based on the delay gradient, where the network status includes under-use, over-use, and normal;

[0083] The current sending rate is adjusted based on the network status of the current round to obtain an adjusted sending rate.

[0084] Explanation: the sending rate is adjusted based on the estimated data packet delay in the data receiving status information corresponding to the two time windows before and after, and the delay gradient (slope factor) of the two data packet delays is calculated to evaluate the degree of network congestion. According to the change of the delay gradient, the network status is judged as "under-utilized", "over-utilized" or "normal".

[0085] Specifically, when the network is congested, data packets will queue in the intermediate network devices and wait for forwarding, which will cause the delay gradient to grow. When the network traffic drops, the network devices quickly consume (forward) the data packets in their send queues, and the subsequent data packets have a shorter queue time. At this time, the delay gradient decreases or becomes negative. Through this change, we can determine whether the current network is in a congested or non-congested state. If the delay gradient continues to grow, it may mean that the network is experiencing congestion (overuse); if the delay gradient is negative, it may mean that the bandwidth is not fully utilized (underuse); if the delay gradient decreases, it may mean that the network is in good condition (normal). In this way, the data sending rate can be dynamically adjusted to adapt to the current network conditions.

[0086] For example, the bandwidth can be estimated through feedback information based on the congestion control mechanism of SFTP. Bandwidth prediction is mainly achieved through loss controller and delay controller. The loss controller is located at the sending end and adjusts the sending rate based on the packet loss rate; the delay controller is located at the receiving end and judges the network status based on the delay gradient of the packet arrival, and adjusts the sending rate accordingly.

[0087] In one embodiment of the present specification: adjusting the current sending rate based on the network status of the current round to obtain the adjusted sending rate includes:

[0088] When the network status is underutilized, the sum of the current sending rate and the preset increment is used as the adjusted sending rate;

[0089] When the network status is overused, the sending rate is adjusted to half of the current sending rate;

[0090] When the network status is normal, the current sending rate is used as the adjusted sending rate.

[0091] Explanatory, the Additive Increase Multiplicative Decrease (AIMD) algorithm is used to estimate the currently available bandwidth. If the network status is "underutilized", the send rate is adjusted to increase appropriately based on the current receive rate; if the network status is "overutilized", the send rate is adjusted to decrease exponentially based on the current receive rate. This avoids the situation where the network is overutilized and causes data reporting to lag or fail to report. The adjustment of the send rate is achieved through simple logic and a small amount of historical data. Adjusting the send rate provides fast and relatively stable adjustments at the beginning of data transmission, providing a data basis for the prediction of subsequent prediction models. And it only relies on some hard-coded configuration parameters. Although this may lead to inaccurate adjustments under certain network conditions, it can provide reasonable bandwidth estimates in most cases and ensure the robustness of the system.

[0092] In one embodiment of the present specification: selecting one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round based on all historical adjusted sending rates, all historical predicted sending rates, and all historical receiving rates includes:

[0093] Obtaining a first error based on all historical receiving rates and all historical adjusted sending rates;

[0094] Obtain a second error based on all historical receiving rates and all historical predicted sending rates;

[0095] When the absolute difference between the first error and the second error is greater than a preset threshold, the adjusted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round;

[0096] When the absolute difference between the first error and the second error is less than or equal to a preset threshold, the predicted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round.

[0097] For example, according to: Among them, F i represents the estimated value, A i Represents the true value, and n represents the number of samples. The range of sMAPE is 0% to 200%, and the smaller the value, the smaller the estimation error. When sMAPE is equal to 0%, it means that the estimated value and the true value are exactly the same; when sMAPE is equal to 200%, it means that the estimated value and the true value are very different.

[0098] The receiving rate of each time window is used as the actual bandwidth capacity of the transmission channel during the duration of each time window for calculation. In fact, the receiving rate is less than or equal to the bandwidth capacity. Of course, network tools can also be used to estimate the actual bandwidth capacity, and the estimation results in each time window are used as the actual bandwidth capacity of each time window for calculation. The symmetric mean absolute percentage error of the adjusted sending rate is calculated to obtain sMAPE1, that is, the first error, and the symmetric mean absolute percentage error of the predicted sending rate is calculated to obtain sMAPE2, that is, the second error. When the absolute difference between the first error and the second error is less than or equal to the preset threshold, it means that the prediction accuracy of the prediction model under the current network state has met the requirements, and the adjusted sending rate can be used as the target sending rate instead, and when the network state does not change much, it can be confirmed that the prediction accuracy will become higher and higher.

[0099] In one embodiment of the present specification: before step S1, a step of selecting a preset threshold is also included, and the selection of the preset threshold includes:

[0100] Using a plurality of different thresholds as preset thresholds to perform data transmission based on steps S1-S7 respectively;

[0101] For any of multiple thresholds as the target threshold:

[0102] Based on the target threshold, the target sending rate corresponding to each round in the test round, the data receiving status information corresponding to the last round in the test round and the historical average receiving rate under the target threshold are obtained to obtain the score value of the target threshold;

[0103] Repeat the above process until the score values ​​corresponding to all thresholds are obtained;

[0104] The threshold corresponding to the maximum score value is selected as the preset threshold.

[0105] Illustratively, the selection of the preset threshold will directly affect the selection result of the target transmission rate, thereby affecting the effect of utilizing the bandwidth resources of the transmission channel. In advance, multiple different thresholds are used as preset thresholds for simulation tests, such as 5%, 10%, 20%, 25%, 30%, 40%, and 50%. With 5% as the preset threshold, the adjusted transmission rate and the predicted transmission rate are selected in each time window to obtain the target transmission rates corresponding to each of the multiple time windows. The number of samples n is selected, and the sMAPE is calculated based on all target transmission rates in the n groups. At the same time, the data reception status information of the nth time window is considered to score the preset threshold. Among them, the bandwidth prediction accuracy is obtained by subtracting sMAPE from 1 and multiplying it by 100, which is used to measure the accuracy of the prediction. The network score is obtained by the data reception status information of the nth time window. The network score combines the delay score (ds), the receiving rate score (cs), and the packet loss score (ls). In the specific calculation formula, the delay score and the packet loss score are calculated by fixed formulas, while the receiving rate score is calculated based on the ratio of the actual receiving rate to the theoretical maximum receiving rate. Finally, the score corresponding to the threshold of 5% is obtained. Similarly, simulation tests are performed with other thresholds as preset thresholds to obtain their corresponding score values.

[0106] The test results are shown in the table below:

[0107] Threshold (%) 5 10 20 25 30 40 50 sMAPE 0.75 0.68 0.27 0.18 0.13 0.43 0.54 Rating value 55 59 70 78 93 61 60

[0108] Preferably, the preset threshold is 30%.

[0109] In multiple embodiments of this specification: the reward value r for performing data transmission at the target transmission rate corresponding to the current round under the data reception status information of the current round is t , specifically:

[0110]

[0111] Among them, c t ,d t , l t are the receiving rate, packet delay, and packet loss rate in the data receiving status information corresponding to the target transmission rate of the transmission channel in the current round, respectively, linear-to-log (c t ) means to convert c t Function to convert from linear scale to logarithmic scale.

[0112] Illustratively, linear-to-log(c t ) This transformation is used to emphasize low rate values ​​because the logarithmic function changes more dramatically at lower values. Ensure that the penalty does not exceed 1, even if the delay d tWhen the delay is between 0 and 1000, the penalty is proportional to the delay; after 1000, the penalty remains at 1. Combining these parts, the reward function r t Interpretation: Higher receiving rate c t (on a logarithmic scale) will increase the reward. Higher delay d t The reward will be reduced, but the penalty has an upper limit. t The reward will be reduced.

[0113] In multiple embodiments of this specification: the predicted sending rate is obtained based on the current prediction model and the data receiving state information corresponding to the current sending rate, specifically:

[0114] The data receiving state information corresponding to the current sending rate is used as the state input of the current prediction model, and the predicted sending rate is obtained based on the maximized advantage function.

[0115] Explanatory, in reinforcement learning, based on the input state, an action probability distribution is generated according to the model's strategy, and an action output (i.e., predicted sending rate) is selected from it. Figure 2 The figure is a schematic diagram of the prediction model architecture provided by the embodiment of this specification. The strategy update method selected in this embodiment is the Actor-Critic method. In this method, the Actor model usually uses a strategy network (which can be a neural network) that learns how to assign probabilities to each possible action based on the current state input, updates the probabilities based on the advantage function provided by the Critic model, and selects the action output with the updated probability based on the new state input; the Critic model is responsible for evaluating the value Q of the action selected by the Actor model θ (s t ,a t ) or estimate the value V of the state under the current strategy θ (s t ).

[0116] Figure 2The Observed State in represents the state input of the prediction module, including the receiving rate, packet delay, packet loss rate of the current round, and the target sending rate of a preset number of historical rounds; RTC GYM Env (ns-3) represents the simulation environment, a real-time communication (RTC) scenario built on the ns-3 network simulation platform, which can simulate different network topologies, traffic patterns, and failure scenarios, and can generate a large amount of data for training and verifying models, which is used as a data source in the prediction model test. The main process of the prediction model architecture is the entire cycle of receiving state input, processing input through a neural network, predicting action output, executing actions, receiving reward values ​​for actions, and adjusting the policy network parameters according to the reward values. The action output is a quantized series of discrete bandwidth intervals (such as 0.01-2Mbps, 2-4Mbps, etc.).

[0117] In each time window, the Actor model selects an action and executes it based on the data reception status information corresponding to the current time window, and then the Critic model receives the data status information and the corresponding reward value r corresponding to the next time window. t To update the value estimate V θ (s t ), this value estimate reflects the expected cumulative reward that can be obtained in the future by following the current strategy based on the current state input.

[0118] The advantage function is:

[0119] A θ (s t ,a t )=Q θ (s t ,a t )-V θ (s t )

[0120] Q θ (s t ,a t ) represents the input state s t Take action a t The value of V θ (s t ) represents the input state s t The estimated value of the following.

[0121] Maximizing the advantage function is to select the a corresponding to the largest advantage function t As action output.

[0122] In multiple embodiments of the present specification: the training of the current prediction model to obtain the trained prediction model based on the data reception state information corresponding to the current round and all rounds of a preset number before the current round, the target sending rate, and the reward value of the target sending rate includes:

[0123] Based on the value function in the advantage function of the current prediction model, obtain the expected return of the current round;

[0124] The value function is updated based on the data receiving state information corresponding to the current round and all the rounds before the current round, a preset number of rounds, the target sending rate and the reward value of the target sending rate, the expected return of the current round and the minimized loss function, and the value error is calculated;

[0125] The value function is updated based on the value error to train the current prediction model to obtain a trained prediction model.

[0126] Explanatory, according to the state input of each time window, that is, the data receiving state information and its corresponding reward value r t , for the same state input s t Continuously update their corresponding value estimates V θ (s t ). For example, the reward value r of the current time window t The positive or negative error between the sum of the value estimates of the next time window and the value estimate of the current time window indicates whether the action selected by the Actor model is good or bad relative to the expected return of the value function. This error can be used to update the policy network of the Actor model so that in similar future states, the Actor model is more inclined to choose actions that can obtain higher value.

[0127] The value error corresponds to the loss function:

[0128]

[0129] y t is an estimate based on real returns, is the value function network for the current state s t Minimizing the loss function is to update the value estimate so that the loss function is as small as possible.

[0130] In one embodiment of the present specification: in the expected return of the predicted output of the current prediction model obtained based on the value function in the maximized advantage function in the current prediction model, the expected return is calculated by a k-step time difference method, where k = a preset number.

[0131] Explanatory, the value error is calculated based on the data reception state information, target sending rate and reward value of the target sending rate corresponding to each preset number of rounds. The cumulative return within the preset number of steps starting from the current state is considered to update the value estimate, which helps to reduce the randomness in the single-step return. The updated value estimate V θ (s t ) is more stable, but it will affect the learning efficiency of the prediction model to a certain extent. The larger the preset number, the greater the value estimate V θ (s t ) is more stable and the learning speed is faster. Balance the learning speed and stability by adjusting the size of the preset number.

[0132] Estimation y based on real returns t for:

[0133]

[0134] Where k = the preset number, γ is the discount factor, R t+1 is the reward value corresponding to step t+1, V(S t+k ) is the state S of step t+k t+k expected return.

[0135] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims 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, multitasking and parallel processing are also possible or may be advantageous.

[0136] See next Figure 3 , Figure 3 A schematic diagram of the structure of a remote intelligent transmission system for large-volume fault data of an electricity consumption information collection terminal provided in an embodiment of this specification is shown.

[0137] like Figure 3 As shown, the transmission system 300 includes a data collection module 301, a hybrid prediction module 302, an adaptive switching module 303 and a learning module 304;

[0138] The data collection module 301 performs data transmission based on the current transmission rate, and obtains data reception state information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception state information includes reception rate, data packet delay, and data packet loss rate;

[0139] The hybrid prediction module 302 obtains an adjusted sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, and obtains a predicted sending rate based on the current prediction model and the data receiving state information corresponding to the current sending rate;

[0140] The adaptive switching module 303 selects one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round based on all historical adjusted sending rates, all historical predicted sending rates and the historical average receiving rate;

[0141] The data collection module 301 performs data transmission based on the target transmission rate corresponding to the current round to obtain the corresponding data receiving state information when the transmission channel performs data transmission at the target transmission rate corresponding to the current round.

[0142] The learning module 304 obtains a reward value for performing data transmission at the target transmission rate corresponding to the current round under the data reception state information of the current round based on the data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round; trains the current prediction model based on the data reception state information corresponding to each of all rounds including the current round and a total of a preset number of rounds before the current round, the target transmission rate, and the reward value of the target transmission rate to obtain a trained prediction model;

[0143] The hybrid prediction module 302 uses the target sending rate corresponding to the selected current round as the current sending rate and uses the trained prediction model as the current prediction model to re-acquire the adjusted sending rate and the predicted sending rate.

[0144] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the transmission system embodiment, since it is basically similar to the remote intelligent transmission method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the remote intelligent transmission method embodiment.

[0145] See also Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this specification is shown.

[0146] like Figure 4 As shown, the electronic device 400 may include: at least one processor 401 , at least one network interface 404 , a user interface 403 , a memory 405 , and at least one communication bus 402 .

[0147] The communication bus 402 may be used to realize the connection and communication among the above-mentioned components.

[0148] The user interface 403 may include buttons, and the optional user interface may also include a standard wired interface or a wireless interface.

[0149] The network interface 404 may include, but is not limited to, a Bluetooth module, an NFC module, a Wi-Fi module, etc.

[0150] Among them, the processor 401 may include one or more processing cores. The processor 401 uses various interfaces and lines to connect the various parts of the entire electronic device 400, and executes various functions and processes data of the electronic device 400 by running or executing instructions, programs, code sets or instruction sets stored in the memory 405, and calling data stored in the memory 405. Optionally, the processor 401 can be implemented in at least one hardware form of DSP, FPGA, and PLC. The processor 401 can integrate one or a combination of CPU, GPU, modem, etc. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 401, and it can be implemented separately through a chip.

[0151] Among them, the memory 405 may include RAM or ROM. Optionally, the memory 405 includes a non-transient computer-readable medium. The memory 405 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 405 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above various method embodiments, etc. The memory 405 may also be at least one storage device located away from the aforementioned processor 401. The memory 405 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a remote intelligent transmission application. The processor 401 can be used to call the remote intelligent transmission application stored in the memory 405 and execute the steps of the remote intelligent transmission method mentioned in the aforementioned embodiment.

[0152] The embodiments of this specification also provide a computer-readable storage medium, which stores instructions, and when the instructions are executed on a computer or a processor, the computer or the processor executes one or more steps in the above-mentioned remote intelligent transmission method embodiment. If the component modules of the above-mentioned electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0153] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of this specification is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (Digital Subscriber Line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, data center, etc. that contains one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0154] Those 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. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes. In the absence of conflict, the technical features in this embodiment and the implementation scheme can be combined arbitrarily.

[0155] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Without departing from the design spirit of this specification, various modifications and improvements made to the technical solutions of this specification by ordinary technicians in this field should fall within the scope of protection determined by the claims of this specification.

Claims

1. A remote intelligent transmission method for large volume fault data of an electricity consumption information collection terminal, characterized in that: The following steps are involved: S1. Perform data transmission based on the current transmission rate, and obtain data reception state information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception state information includes reception rate, data packet delay, and data packet loss rate; S2, based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, the sending rate is adjusted, and based on the current prediction model and the data receiving state information corresponding to the current sending rate, the predicted sending rate is obtained; S3, based on all historical adjusted sending rates, all historical predicted sending rates, and all historical receiving rates, select one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round; S4, performing data transmission based on the target transmission rate corresponding to the current round, so as to obtain data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round; S5. Based on the data reception state information corresponding to the transmission channel when data is transmitted at the target transmission rate corresponding to the current round, obtain a reward value for transmitting data at the target transmission rate corresponding to the current round under the data reception state information of the current round; S6, based on the data receiving state information corresponding to the current round and all the rounds before the current round totaling a preset number, the target sending rate, and the reward value of the target sending rate, the current prediction model is trained to obtain a trained prediction model; S7. Use the target sending rate corresponding to the selected current round as the current sending rate and use the trained prediction model as the current prediction model and return to step S2.

2. According to claim 1, a method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal is characterized in that: The adjusting the sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round includes: Obtaining a delay gradient based on the data packet delays corresponding to the data reception state information corresponding to the current transmission rate and the data reception state information corresponding to the previous round of the current round; Acquire the network status of the current round based on the delay gradient, where the network status includes under-use, over-use, and normal; The current sending rate is adjusted based on the network status of the current round to obtain an adjusted sending rate.

3. A method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 2, characterized in that: The adjusting the current sending rate based on the network status of the current round to obtain the adjusted sending rate includes: When the network status is underutilized, the sum of the current sending rate and the preset increment is used as the adjusted sending rate; When the network status is overused, the sending rate is adjusted to half of the current sending rate; When the network status is normal, the current sending rate is used as the adjusted sending rate.

4. The method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 1 is characterized in that: The selecting, based on all historical adjusted sending rates, all historical predicted sending rates, and the historical average receiving rate, one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round includes: Obtaining a first error based on a historical average receiving rate and all historical adjusted sending rates; Obtaining a second error based on a historical average receiving rate and all historical predicted sending rates; When the absolute difference between the first error and the second error is greater than a preset threshold, the adjusted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round; When the absolute difference between the first error and the second error is less than or equal to a preset threshold, the predicted sending rate corresponding to the current round is used as the target sending rate corresponding to the current round.

5. A method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 4, characterized in that: Before step S1, a step of selecting a preset threshold is also included, and the selection of the preset threshold includes: Using a plurality of different thresholds as preset thresholds to perform data transmission based on steps S1-S7 respectively; For any of multiple thresholds as the target threshold: Based on the target threshold, the target sending rate corresponding to each round in the test round, the data receiving status information corresponding to the last round in the test round and the historical average receiving rate under the target threshold are obtained to obtain the score value of the target threshold; Repeat the above process until the score values ​​corresponding to all thresholds are obtained; The threshold corresponding to the maximum score value is selected as the preset threshold.

6. The method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 1 is characterized in that: The reward value r for transmitting data at the target transmission rate corresponding to the current round under the data reception status information of the current round is t , specifically: Among them, c t d t , l t are the receiving rate, packet delay, and packet loss rate in the data receiving status information corresponding to the target transmission rate of the transmission channel in the current round, respectively, linear-to-log (c t ) means to convert c t Function to convert from linear scale to logarithmic scale.

7. The method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 1 is characterized in that: The predicted sending rate is obtained based on the current prediction model and the data receiving state information corresponding to the current sending rate, specifically: The data receiving state information corresponding to the current sending rate is used as the model input of the current prediction model, and the predicted sending rate is obtained based on the maximized advantage function.

8. A method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 7, characterized in that: The training of the current prediction model to obtain a trained prediction model based on the data receiving state information corresponding to the current round and all rounds of a preset number before the current round, the target sending rate, and the reward value of the target sending rate includes: Based on the value function in the advantage function of the current prediction model, obtain the expected return of the current round; The value function is updated based on the data receiving state information corresponding to the current round and all the rounds before the current round, a preset number of rounds, the target sending rate and the reward value of the target sending rate, the expected return of the current round and the minimized loss function, and the value error is calculated; The value function is updated based on the value error to train the current prediction model to obtain a trained prediction model.

9. A method for remote intelligent transmission of large volume fault data of an electricity consumption information collection terminal according to claim 8, characterized in that: The value function in the maximized advantage function in the current prediction model is based on obtaining the expected return of the prediction output of the current prediction model, and the expected return is calculated by the k-step time difference method, where k=preset number.

10. A transmission system using the remote intelligent transmission method for large-volume fault data of an electricity consumption information collection terminal as claimed in any one of claims 1 to 9, characterized in that: It includes a data collection module, a hybrid prediction module, an adaptive switching module and a learning module; The data collection module performs data transmission based on the current transmission rate, and obtains data reception status information corresponding to the transmission channel when performing data transmission at the current transmission rate, wherein the data reception status information includes reception rate, data packet delay, and data packet loss rate; The hybrid prediction module obtains an adjusted sending rate based on the current sending rate, the data receiving state information corresponding to the current sending rate, and the data receiving state information corresponding to the previous round of the current round, and obtains a predicted sending rate based on the current prediction model and the data receiving state information corresponding to the current sending rate; The adaptive switching module selects one of the adjusted sending rate corresponding to the current round and the predicted sending rate corresponding to the current round as the target sending rate corresponding to the current round based on all historical adjusted sending rates, all historical predicted sending rates and the historical average receiving rate; The data collection module performs data transmission based on the target transmission rate corresponding to the current round to obtain the corresponding data receiving state information when the transmission channel performs data transmission at the target transmission rate corresponding to the current round. The learning module obtains a reward value for performing data transmission at the target transmission rate corresponding to the current round under the data reception state information of the current round based on the data reception state information corresponding to the transmission channel when performing data transmission at the target transmission rate corresponding to the current round; and trains the current prediction model based on the data reception state information, the target transmission rate, and the reward value of the target transmission rate corresponding to all rounds including the current round and a total of a preset number of rounds before the current round to obtain a trained prediction model; The hybrid prediction module uses the target sending rate corresponding to the selected current round as the current sending rate and uses the trained prediction model as the current prediction model to re-acquire the adjusted sending rate and the predicted sending rate.

Images