Electric power information acquisition method based on dual-mode communication

Abstract

The invention relates to the technical field of information acquisition, in particular to an electric power information acquisition method based on dual-mode communication, which comprises the following steps: S1, acquiring multi-dimensional evaluation data of a dual-mode communication channel by an acquisition terminal before an electric power information acquisition task is initiated each time; s2, acquiring task characteristics according to the current network state; according to the invention, through multi-dimensional evaluation and dynamic weight, the success rate of single acquisition is improved, the communication time delay is reduced, only key data is transmitted in a dual-mode concurrent manner, the energy consumption is reduced, the congestion rate of an RF channel is reduced through space-time association and reinforcement learning, the acquisition efficiency in a peak period is improved, and the acquisition efficiency is improved. According to the method, the condition of global degradation caused by local optimization is avoided, through a closed loop of data acquisition-strategy optimization-feedback execution, the method can adapt to long-term changes such as season replacement and equipment aging, and when HPLC interference is increased in winter, the RF priority is automatically improved, manual intervention is not needed, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of information acquisition technology, specifically to a method for acquiring power information based on dual-mode communication. Background Technology

[0002] Dual-mode communication-based power information acquisition is a technical means that combines two different communication technologies to achieve more efficient and reliable power data acquisition and transmission. Dual-mode communication usually refers to the use of high-speed power line carrier communication (such as HPLC) and high-speed wireless communication technology (such as HRF) for data transmission. In this mode, data can be transmitted bidirectionally through dual channels on the power line. The two communication methods can complement each other and be integrated. The appropriate communication method can be selected according to the specific environment to ensure high-speed and stable data transmission.

[0003] Patent publication number CN202411770036.8 describes in its specification that "This invention discloses a method and system for collecting electricity meter information based on dual-mode communication technology. The method includes: both the CCO dual-mode communication module of the concentrator and the STA dual-mode communication module of the electricity meter are equipped with HPLC high-speed power line carrier communication function and HRF high-speed low-power wireless communication function; the concentrator and the electricity meter are respectively subjected to CCO dual-mode communication module time synchronization, STA dual-mode communication module time synchronization, main communication channel selection, and data reception supplementary acquisition processing; finally, the data of all electricity meters are collected. The beneficial effects of this invention are: while accurately collecting electricity meter information, it rationally sets up the means of allocating communication channel resources to the main communication channel, avoiding interference problems caused by the power grid with multiple electricity meters; and it solves the network congestion caused by the large amount of data when collecting data from a large number of electricity meters in a distribution area." The aforementioned technologies address issues such as missing data acquisition and slow dual-channel data verification speeds after acquisition through supplementary acquisition methods. While these technologies improve the accuracy and consistency of acquired data by using precise time synchronization of CCO and STA modules to avoid errors in multi-meter datasets within adjacent 10-minute intervals as in traditional time synchronization methods, they are often based on single indicators, such as RSSI signal strength, or fixed primary / backup strategies for channel selection, ignoring key factors such as instantaneous channel load and historical stability. This leads to increased acquisition latency, data packet loss, or increased network interference, yet HPLC is still prioritized, causing frequent switching. Furthermore, existing technologies only focus on communication optimization for individual terminals, lacking global scheduling. When terminals in the same area independently select optimal channels, they are prone to collectively switching to RF, causing channel congestion. Moreover, the lack of spatiotemporal pattern prediction makes it impossible to anticipate channel quality degradation during peak electricity consumption periods.

[0004] In conclusion, developing a power information acquisition method based on dual-mode communication remains a key issue that urgently needs to be addressed in the field of information acquisition technology. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies, which often rely on single indicators such as signal strength RSSI or fixed primary / backup strategies for channel selection, neglecting key factors such as instantaneous channel load and historical stability. This leads to increased acquisition delays, data packet loss, or frequent switching even when network interference increases. Furthermore, existing technologies only focus on the communication optimization of individual terminals, lacking global scheduling. When terminals in the same area independently select the best option, they are prone to collectively switching to RF, causing channel congestion. Moreover, the lack of spatiotemporal pattern prediction makes it impossible to anticipate the channel quality degradation during peak power consumption periods.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a power information acquisition method based on dual-mode communication, comprising the following steps: S1. Before each power information collection task is initiated, the acquisition terminal obtains multi-dimensional evaluation data of the dual-mode communication channel; S2. Based on the current network status and the characteristics of the data collection task, dynamically adjust the weight coefficients of the multi-dimensional evaluation data and output the weighted evaluation data; S3. Calculate the comprehensive score of the dual-mode communication channel based on the weighted evaluation data, and select a transmission scheme according to the comprehensive score; S4. By aggregating the full-chain data generated by multiple acquisition terminals after performing steps S1-S3 through edge computing nodes, the spatiotemporal correlation of the full-chain data is explored. S5. Construct a reinforcement learning model by collecting input data through a data acquisition terminal, setting a reward function, training the reinforcement learning model, and outputting the execution action. S6. The execution actions output by the reinforcement learning model are sent to the acquisition terminals and edge computing nodes in the corresponding areas.

[0007] Furthermore, in step S1, the method by which the acquisition terminal obtains multi-dimensional evaluation data of the dual-mode communication channel before each power information acquisition task is initiated is as follows: The dual-mode communication channel includes an HPLC channel and an RF channel. The multi-dimensional evaluation data includes instantaneous channel capacity prediction, historical communication success rate, current channel load rate, and node remaining energy. After preprocessing the multi-dimensional evaluation data, the following calculations are performed based on the Shannon formula. Instantaneous channel capacity at time t ,expression: In the formula, represents Instantaneous channel capacity at time t. Indicates the total bandwidth of the channel. Represents the logarithmic function with base 2. Indicates the first The signal-to-noise ratio correlation term of each subcarrier, express Transmission power at any given time Indicates the first The average amplitude response of each subcarrier Indicates the first The average amplitude response squared of each subcarrier Represents the noise power spectral density. Indicates the first The bandwidth of each subcarrier Indicates the first Signal-to-noise ratio of each subcarrier Representing the The center frequency of each subcarrier; The acquisition terminal estimates the maximum amount of data that the channel can transmit within the next 1 to 4 seconds based on the current channel state information using a linear prediction algorithm, expressed as: In the formula, Indicates the future Channel capacity prediction at time 10:00 To predict the time, Indicates the predicted time forward Historical channel capacity in milliseconds For the first The linear prediction coefficients are solved using the least squares method; The historical communication success rate is the ratio of the number of times the corresponding channel completed the acquisition task within the previous 1 to 3 minutes to the total number of tasks; the current channel load rate is calculated by the number of active nodes in the listening channel or the data packet collision rate.

[0008] Furthermore, in step S2, the method for dynamically adjusting the weight coefficients of the multi-dimensional evaluation data and outputting the weighted evaluation data based on the current network status and the characteristics of the data acquisition task is as follows: The weighting coefficients of the multi-dimensional evaluation data are dynamically adjusted as follows: when the network interference level is greater than or equal to a preset interference threshold, the weighting of historical communication success rate is increased to 30%–50%; when the amount of data collected is greater than or equal to a preset data amount threshold, the weighting of instantaneous channel capacity prediction is increased to 40%–60%. The acquisition terminal obtains the interference power spectral density of the dual-mode channel through the CSI detection module, calculates the total interference power by combining it with the channel bandwidth, and then quantifies the network interference level. The expression is: In the formula, Indicates the total interference power of the channel. For the interference power spectral density, For the first Subcarrier frequencies, Indicates the first The bandwidth of each subcarrier is normalized. Mapping to the [0,1] interval yields the quantized value of the network interference level. The preset interference threshold is , Based on actual power scenario measurements, the HPLC channel... =0.6, RF channel =0.5, used to determine whether the "historical communication success rate weighting" rule is triggered; The data acquisition terminal extracts the amount of data to be acquired in this task from the task instructions. Similarly, normalization is applied to map to the [0,1] interval, and the expression is: In the formula, This is a quantified value of the amount of data collected. The maximum data volume for a single data collection session is set to a preset data volume threshold. , =0.5, corresponding to 50KB for the HPLC channel and 25KB for the RF channel, is used to determine whether the "instantaneous channel capacity weighting" rule is triggered.

[0009] Further, in step S3, the comprehensive score of the dual-mode communication channel is calculated based on the weighted evaluation data, and the method for selecting a transmission scheme based on the comprehensive score is as follows: The comprehensive score for calculating the dual-mode communication channel includes the comprehensive score of the HPLC channel and the comprehensive score of the RF channel, expressed as: In the formula, It is the final comprehensive score of the HPLC channel. The HPLC channel-based weighted evaluation value is obtained from the weighted evaluation data. This is the real-time stability correction value for the HPLC channel. This is the HPLC channel history bias correction amount. It is the final comprehensive score of the RF channel. The RF channel basic weighted evaluation value is obtained from the weighted evaluation data. It is the real-time stability correction value for the RF channel. It is the historical deviation correction amount of the RF channel. At the same time, the comprehensive score is maintained in the range of [0,1] by using truncation processing. The transmission scheme selection includes selecting the channel with the highest comprehensive score as the main communication channel; and, if the difference between the comprehensive scores of the two channels is less than or equal to a preset score threshold, dual-mode concurrent transmission is triggered, and the preset score threshold is dynamically calculated based on the network interference level and the amount of task data. When the dual-mode concurrent transmission is triggered, the acquisition terminal splits the acquired data into critical data segments and non-critical data segments. The critical data segments are transmitted simultaneously through the HPLC and RF channels, while the non-critical data segments are transmitted only through the main communication channel.

[0010] Furthermore, in step S4, the method for aggregating the entire chain of data generated by multiple acquisition terminals after executing steps S1-S3 through edge computing nodes, and mining the spatiotemporal correlation of the entire chain of data, is as follows: The full-chain data includes channel selection results, multi-dimensional evaluation data, and actual communication performance data; The spatiotemporal correlation includes spatial correlation mining and temporal correlation mining; The spatial correlation mining analyzes the channel selection preferences of acquisition terminals within the same transformer area and the same feeder branch, expressed as: In the formula, Discrete coefficients representing channel selection preference This represents the average coefficient used in variance calculations. Indicates the first The percentage of HPLC selection for each terminal. This represents the average HPLC selection percentage across all terminals within the region. Indicates the first Squared HPLC selectivity of each terminal This indicates the average selection percentage of the region using HPLC. This indicates the regional HPLC channel selection preference. Indicates the regional RF channel selection preference. middle Indicates the number of times. Indicates the HPLC channel. Indicates a choice. and correspond, Indicates RF channel, This represents the total number of times the main channel was selected by all acquisition terminals within the target area. This represents the total number of times that all data acquisition terminals within the target area selected RF as the primary communication channel. This indicates the total number of times all acquisition terminals within the target area selected HPLC as the primary communication channel, identifying areas where HPLC interference remained consistently high and RF signals were stable.

[0011] Furthermore, in step S4, the method for aggregating the entire chain of data generated by multiple acquisition terminals after executing steps S1-S3 through edge computing nodes, and mining the spatiotemporal correlation of the entire chain of data, is as follows: The time correlation mining involves statistically analyzing channel quality data for each time period from 0:00 to 24:00 daily, and calculating the average channel quality index for each time period over seven consecutive days. The expression is: In the formula, This represents the average channel quality index for a specific time period and a specific channel. This represents the average value. Represents the averaging coefficient. This indicates the total number of data acquisition terminals within the target area. This represents the raw data representing the channel quality of a single sample. It is the number of days. It is a time period. It is the terminal number. It is the channel type, identifying the peak electricity consumption period: 8:00-21:00, and the periodic pattern of increased HPLC noise.

[0012] Further, in step S5, the reinforcement learning model is constructed by collecting input data through a data acquisition terminal, setting a reward function, training the reinforcement learning model, and outputting the execution action. The reinforcement learning model is a deep Q-network model, and the input data includes regional network status, historical communication statistics, and task queue information. The reward function includes regional data collection success rate, average latency, and resource balancing rate, with the actual regional data collection success rate set as follows: The target success rate is =0.985, expression: In the formula, This indicates a reward for successful data collection. This represents the weighting coefficient for the reward based on the success rate of data collection. This indicates the actual data collection success rate in the region. This represents the target value for the success rate of regional data collection. It represents the ratio of the actual success rate to the target success rate.

[0013] Further, in step S5, the reinforcement learning model is constructed by collecting input data through a data acquisition terminal, setting a reward function, training the reinforcement learning model, and outputting the execution action. The execution actions include group-level channel scheduling strategies and task priority arbitration rules; The group-level channel scheduling strategy includes, but is not limited to: specifying a preferred communication pattern for a specific area in the next 6 to 24 minutes, and reserving 22% to 32% of the RF channel bandwidth for peak periods; The task priority arbitration rules include, but are not limited to: fault reporting tasks have higher priority than batch meter reading tasks, and remote fee control commands have higher priority than ordinary electricity data collection tasks.

[0014] Furthermore, in step S6, the method for distributing the execution action output by the reinforcement learning model to the corresponding acquisition terminal and edge computing node is as follows: The group-level channel scheduling strategy described in the execution action is sent to the acquisition terminals in the corresponding area as a constraint condition for weight adjustment in step S2. The sending method adopts a broadcast + unicast confirmation mechanism. The edge computing node first broadcasts the strategy data packet to the target area through the HPLC broadband carrier, and the broadcast period is [missing information]. =10s, broadcast 3 times consecutively. After the acquisition terminal receives the data packet, it verifies the checksum and the area identifier. If a match is confirmed, it returns an ACK confirmation frame. The ACK confirmation frame contains the terminal ID and the policy reception timestamp. The ACK rate of edge computing nodes is expressed as: In the formula, Indicates the ACK feedback rate. This indicates the number of acquisition terminals that successfully returned confirmation frames. This indicates the total number of target data collection terminals for the policy issuance. If the data collection terminal is not confirmed, a unicast retransmission will be used, with the number of retransmissions ≤ 2.

[0015] Furthermore, in step S6, the method for distributing the execution action output by the reinforcement learning model to the corresponding acquisition terminal and edge computing node is as follows: The task priority arbitration rule mentioned in the execution action is applied to the data acquisition task queue management of edge computing nodes, prioritizing the execution of high-priority tasks. The expression is: In the formula, Indicates the first The overall priority score of each task. Indicates the first The weights corresponding to each task type This indicates that the data volume influence coefficient is 0.1. Indicates the first The data size of each task Indicates the maximum amount of data for the task. The influence coefficient for the deadline is 0.3. Indicates the first The deadline for each task. Indicates the current time.

[0016] Beneficial effects Compared with known public technologies, the technical solution provided by this invention has the following beneficial effects: This invention improves the success rate of single data acquisition through multi-dimensional evaluation and dynamic weighting, while reducing communication latency. Dual-mode concurrency transmits only critical data, facilitating energy reduction. Spatiotemporal correlation and reinforcement learning help reduce RF channel congestion and improve acquisition efficiency during peak periods, avoiding local optima leading to global degradation. Through a closed loop of data acquisition, strategy optimization, and feedback execution, this invention can adapt to long-term changes such as seasonal changes and equipment aging. When HPLC interference increases in winter, it automatically increases RF priority without manual intervention, which helps reduce operation and maintenance costs. Attached Figure Description

[0017] Figure 1 This is a flowchart of a power information acquisition method based on dual-mode communication according to the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1: like Figure 1As shown, this invention provides a power information acquisition method based on dual-mode communication, comprising the following steps: S1. Before each power information collection task is initiated, the acquisition terminal obtains multi-dimensional evaluation data of the dual-mode communication channel; The dual-mode communication channel includes an HPLC channel and an RF channel. Multi-dimensional evaluation data includes instantaneous channel capacity prediction, historical communication success rate, current channel load rate, and node remaining energy. After preprocessing the multi-dimensional evaluation data, calculations are performed based on Shannon's formula. Instantaneous channel capacity at time t ,expression: In the formula, represents Instantaneous channel capacity at time t. Indicates the total bandwidth of the channel. Represents the logarithmic function with base 2. Indicates the first The signal-to-noise ratio correlation term of each subcarrier, express Transmission power at any given time Indicates the first The average amplitude response of each subcarrier Indicates the first The average amplitude response squared of each subcarrier Represents the noise power spectral density. Indicates the first The bandwidth of each subcarrier Indicates the first Signal-to-noise ratio of each subcarrier Representing the The center frequency of each subcarrier; It should be noted that the acquisition terminal uses a linear prediction algorithm based on the current channel state information to estimate the maximum amount of data that the channel can transmit within the next 1 to 4 seconds. The expression is: In the formula, Indicates the future Channel capacity prediction at time 10:00 To predict the time, Indicates the predicted time forward Historical channel capacity in milliseconds For the first The linear prediction coefficients are solved using the least squares method; In addition, the historical communication success rate is the ratio of the number of times the corresponding channel completed the acquisition task in the previous 1 to 3 minutes to the total number of tasks; the current channel load rate is calculated by the number of active nodes in the listening channel or the data packet collision rate. As examples, in this embodiment, the data acquisition terminal adopts an HPLC+RF dual-mode module, with a built-in CSI detection chip (model AD9361) and a power detection module (model MAX17048), supporting channel listening function; the edge computing node is deployed next to the transformer in the distribution area, equipped with an edge computing module, CPU model RK3568, supporting data storage and reinforcement learning model operation; the HPLC adopts the GB / T39841-2021 standard, and the RF adopts the LoRaWAN protocol; In this embodiment, the multi-dimensional evaluation data covers real-time, historical, and energy consumption dimensions. Combined with Shannon's formula and linear prediction algorithm, the channel capacity evaluation takes into account both the current state and short-term change trends, which helps to avoid misjudging the channel capability due to instantaneous interference. It also has a wide range of hardware modules and protocol standards to adapt to. The preprocessed evaluation data provides accurate input for S2 dynamic weight adjustment, avoiding subsequent decisions from relying on a single indicator.

[0021] S2. Based on the current network status and the characteristics of the data collection task, dynamically adjust the weight coefficients of the multi-dimensional evaluation data and output the weighted evaluation data; It is important to note that dynamically adjusting the weighting coefficients of multi-dimensional evaluation data includes increasing the weighting of historical communication success rate to 30%–50% when the network interference level is ≥ a preset interference threshold; and increasing the weighting of instantaneous channel capacity prediction to 40%–60% when the amount of data collected is ≥ a preset data volume threshold. As examples, in this embodiment, the acquisition terminal obtains the interference power spectral density of the dual-mode channel through the CSI detection module, calculates the total interference power by combining it with the channel bandwidth, and then quantifies the network interference level. The expression is: In the formula, Indicates the total interference power of the channel. For the interference power spectral density, For the first Subcarrier frequencies, Indicates the first The bandwidth of each subcarrier is normalized. Mapping to the [0,1] interval yields the quantized value of the network interference level. The preset interference threshold is , Based on actual power scenario measurements, the HPLC channel... =0.6, RF channel =0.5, used to determine whether the "historical communication success rate weighting" rule is triggered; In addition, the acquisition terminal extracts the amount of data to be acquired in this task from the task instructions. Similarly, normalization is applied to map to the [0,1] interval, and the expression is: In the formula, This is a quantified value of the amount of data collected. The maximum data volume for a single data collection session is set to a preset data volume threshold. , =0.5, corresponding to 50KB for the HPLC channel and 25KB for the RF channel, used to determine whether the "instantaneous channel capacity weighting" rule is triggered; In this embodiment, dynamic weight adjustment facilitates improved accuracy in assessing the appropriate scenario. It avoids misjudgment by relying on historical success rates during periods of high interference and focuses on capacity to prevent transmission timeouts during periods of large data volumes. This helps improve the success rate of data collection in complex scenarios. The dual-mode concurrency mechanism ensures reliable transmission of critical data when channel differences are small, achieving a 100% success rate for critical data transmission. At the same time, it helps save channel resources for non-critical data and reduces delays in electricity billing caused by missed power information collection.

[0022] S3. Calculate the comprehensive score of the dual-mode communication channel based on the weighted evaluation data, and select the transmission scheme according to the comprehensive score; As examples, in this embodiment, the comprehensive score of the dual-mode communication channel includes the comprehensive score of the HPLC channel and the comprehensive score of the RF channel, expressed as: In the formula, It is the final comprehensive score of the HPLC channel. The HPLC channel-based weighted evaluation value is obtained from the weighted evaluation data. This is the real-time stability correction value for the HPLC channel. This is the HPLC channel history bias correction amount. It is the final comprehensive score of the RF channel. The basic weighted evaluation value of the RF channel is obtained from the weighted evaluation data. It is the real-time stability correction value for the RF channel. It is the historical deviation correction amount of the RF channel. At the same time, the comprehensive score is maintained in the range of [0,1] by using truncation processing. In addition, the transmission scheme selection includes selecting the channel with the highest comprehensive score as the main communication channel; and if the difference between the comprehensive scores of the two channels is less than or equal to a preset score threshold, dual-mode concurrent transmission is triggered, and the preset score threshold is dynamically calculated based on the network interference level and the amount of task data. It should be noted that when dual-mode concurrent transmission is triggered, the acquisition terminal splits the acquired data into critical data segments and non-critical data segments. The critical data segments are transmitted simultaneously through the HPLC and RF channels, while the non-critical data segments are transmitted only through the main communication channel. In this embodiment, the comprehensive scoring integrates a basic weighted value, stability, and historical deviation correction in three dimensions to avoid misjudgments caused by a single indicator, thus improving the success rate of data collection tasks in commercial complexes. The dynamic threshold, combined with interference and data volume adjustment, helps avoid the rigidity problems caused by fixed thresholds. For example, the threshold is lowered when interference is high during peak electricity consumption, making it easier to trigger concurrent transmission to ensure critical data transmission. During peak periods, the transmission latency of critical data such as peak load is shortened to 150ms. The dual-mode concurrent transmission strategy of separating critical and non-critical data helps to reduce RF channel resource waste while ensuring the reliability of core data, which is conducive to reducing RF channel load rate and improving the overall utilization of communication resources and the continuity of power data collection services.

[0023] S4. By aggregating the full-chain data generated by multiple acquisition terminals after executing steps S1-S3 through edge computing nodes, and mining the spatiotemporal correlation of the full-chain data; In addition, the full-chain data includes channel selection results, multi-dimensional evaluation data, and actual communication performance data; As some examples, in this embodiment, spatiotemporal correlation includes spatial correlation mining and temporal correlation mining; It should be noted that spatial correlation mining analyzes the channel selection preferences of acquisition terminals within the same transformer area and the same feeder branch. The expression is: In the formula, Discrete coefficients representing channel selection preference This represents the average coefficient used in variance calculations. Indicates the first The percentage of HPLC selection for each terminal. This represents the average HPLC selection percentage across all terminals within the region. Indicates the first Squared HPLC selectivity of each terminal This indicates the average selection percentage of the region using HPLC. This indicates the regional HPLC channel selection preference. Indicates the regional RF channel selection preference. middle Indicates the number of times. Indicates the HPLC channel. Indicates a choice. and correspond, Indicates RF channel, This represents the total number of times the main channel was selected by all acquisition terminals within the target area. This represents the total number of times that all data acquisition terminals within the target area selected RF as the primary communication channel. This indicates the total number of times all acquisition terminals within the target area selected HPLC as the primary communication channel, identifying areas with persistently high HPLC interference and stable RF signals. Among them, time correlation mining involves statistically analyzing channel quality data for each time period within 0-24 hours each day, and calculating the average channel quality index for each time period over 7 consecutive days, expressed as: In the formula, This represents the average channel quality index for a specific time period and a specific channel. This represents the average value. Represents the averaging coefficient. This indicates the total number of data acquisition terminals within the target area. This represents the raw data representing the channel quality of a single sample. It is the number of days. It is a time period. It is the terminal number. It is the channel type, identifying peak electricity consumption periods: 8:00-21:00, and the periodic pattern of increased HPLC noise; In this embodiment, spatial correlation mining accurately locates regions with differences in channel adaptability, which helps avoid acquisition failures caused by blindly using HPLC and improves the acquisition success rate. Temporal correlation mining clarifies channel quality patterns during peak hours, providing data support for adjusting HPLC weights and reserving RF bandwidth between 8:00 and 21:00, thus reducing transmission latency during peak hours. Through full-chain data aggregation and correlation analysis, high-quality training data is provided for the S5 reinforcement learning model, avoiding decision bias caused by the model relying on single-dimensional data. At the same time, it provides maintenance personnel with a clear profile of regional channel problems, reducing on-site troubleshooting time and improving the work efficiency of staff.

[0024] S5. Construct a reinforcement learning model by collecting input data through a data acquisition terminal, setting a reward function, training the reinforcement learning model, and outputting the execution action. It should be noted that the reinforcement learning model is a deep Q-network model, and the input data includes the regional network status, historical communication statistics, and task queue information; The reward function includes regional data collection success rate, average latency, and resource balancing rate, with the actual regional data collection success rate set at [value missing]. The target success rate is =0.985, expression: In the formula, This indicates a reward for successful data collection. This represents the weighting coefficient for the reward based on the success rate of data collection. This indicates the actual data collection success rate in the region. This represents the target value for the success rate of regional data collection. This represents the ratio of the actual success rate to the target success rate. In addition, the actions performed include group-level channel scheduling strategies and task priority arbitration rules; Among them, group-level channel scheduling strategies include, but are not limited to: specifying a preferred communication mode for a specific area in the next 6 to 24 minutes, and reserving 22% to 32% of RF channel bandwidth for peak periods; As some examples, in this embodiment, the task priority arbitration rules include, but are not limited to: fault reporting tasks have higher priority than batch meter reading tasks, and remote fee control instructions have higher priority than ordinary electricity data collection tasks. In this embodiment, the input data covers the entire dimensions of network, history, and task. Combined with the deep learning capabilities of the DQN model, the output scheduling strategy is more in line with the complex and dynamic scenarios of the park. Compared with traditional fixed strategies, it is easier to improve the success rate of regional data collection and get closer to the target value. Through multi-dimensional weighting of the reward function, it avoids the model from overly pursuing a single indicator, which facilitates the stabilization of average communication latency, improves resource balancing rate, reduces RF channel congestion, and the output actions directly guide subsequent scheduling, shortens the response time of fault reporting tasks, avoids the risk of power outage due to fault delays, and reduces the workload of maintenance personnel in manually adjusting strategies, thereby improving the automation and intelligence level of the system.

[0025] S6. Send the execution actions output by the reinforcement learning model to the corresponding data acquisition terminals and edge computing nodes; In this process, the group-level channel scheduling strategy is distributed to the corresponding acquisition terminals as a constraint for weight adjustment in step S2. The distribution method adopts a broadcast + unicast confirmation mechanism. The edge computing nodes first broadcast the strategy data packet to the target area via the HPLC broadband carrier, with a broadcast period of [missing information]. =10s, broadcast 3 times consecutively. After the acquisition terminal receives the data packet, it verifies the checksum and area identifier. If a match is confirmed, it returns an ACK confirmation frame. The ACK confirmation frame contains the terminal ID and the policy reception timestamp. The ACK rate of edge computing nodes is expressed as: In the formula, Indicates the ACK feedback rate. This indicates the number of acquisition terminals that successfully returned confirmation frames. This indicates the total number of target data collection terminals for the policy issuance. If the data collection terminal is not confirmed, a unicast retransmission will be used, with the number of retransmissions ≤ 2. Furthermore, the task priority arbitration rule in the execution action is applied to the management of the data acquisition task queue of the edge computing node, prioritizing the execution of high-priority tasks, as expressed in the following expression: In the formula, Indicates the first The overall priority score of each task. Indicates the first The weights corresponding to each task type This indicates that the data volume influence coefficient is 0.1. Indicates the first The data size of each task Indicates the maximum amount of data for the task. The influence coefficient for the deadline is 0.3. Indicates the first The deadline for each task. Indicates the current time; In this embodiment, by using broadcast + unicast confirmation to distribute group-level policies, it is easy to ensure that the policies are efficient and fully covered, avoid channel selection confusion caused by terminal missed policies, and reduce acquisition failures caused by HPLC interference; task priority arbitration makes it easy to quickly execute critical tasks such as fault reporting, and avoid the expansion of faults caused by queuing delays.

[0026] Example 2: like Figure 1 As shown, this invention provides a power information acquisition method based on dual-mode communication, comprising the following steps: Before each acquisition, the acquisition terminal obtains the instantaneous channel capacity prediction values ​​of the HPLC and RF channels through the built-in CSI detection module to predict the maximum transmission volume within 1 to 5 seconds; and uses the local storage unit to statistically analyze the historical communication success rate of the previous 1 to 3 minutes. The channel listening module is used to obtain the number of currently active nodes and calculate the channel load rate. The data acquisition terminal adjusts the weights of each dimension according to preset rules: when the network interference level is ≥-80dBm, the weight of historical communication success rate is increased from 20% to 40%, and the weight of instantaneous channel capacity is increased from 30% to 35%. When the data volume of the acquisition task is ≥10KB, such as in batch log transmission, the instantaneous channel capacity weight is increased from 30% to 50%, and the channel load rate weight is increased from 25% to 20%. When a task is of high priority, such as fault reporting, the weight of historical communication success rate is increased from 20% to 35%, and the weight of channel load rate is increased from 25% to 30%. Calculate the combined score of the two channels and select the channel with the higher score as the primary channel; Under a scoring system with a maximum score of 100 points, if the score difference is ≤5 points, dual-mode concurrent transmission is triggered—critical data, such as fault codes, is transmitted simultaneously through dual channels, while non-critical data, such as ordinary electricity consumption data, is transmitted only through the main channel to reduce energy consumption. Every 5 minutes, the edge computing node receives the "channel selection result, multi-dimensional evaluation data, actual communication effect: success / failure, latency" uploaded by the collection terminal in the jurisdiction, and performs spatial correlation and temporal correlation analysis. Spatial correlation: Statistically analyze the channel selection ratio of terminals within the same distribution area. If the HPLC selectivity of a certain distribution area is ≤30% for three consecutive periods, the area is marked as an "HPLC interference sensitive area". Time correlation: Statistical analysis of channel quality for each time period of the day; if the average bit error rate of HPLC from 9:30 to 11:30 is ≥ This period is marked as the "HPLC quality trough period"; Reinforcement learning model construction and training: Input data includes: region markers such as whether it is an HPLC interference sensitive area, current time period such as whether it is a low quality period, average channel load rate of the region, queue of tasks to be executed, and queue of tasks to be executed with priority; Execution actions: Group-level strategies such as "Prioritize RF in HPLC interference-sensitive areas for the next 10 minutes" and "Reserve 25% RF bandwidth during periods of low quality"; Task arbitration such as "Fault reporting tasks have higher priority than batch meter reading". Reward function: +2 points for every 1% increase in regional acquisition success rate, +1 point for every 10ms decrease in average latency, +3 points for achieving the resource balance rate target, and a difference of ≤20% between the load rates of each channel is considered to meet the target. Model training: The DQN algorithm is used. The model is trained offline using three months of historical data, and the model parameters are updated online every hour to ensure strategy optimization. Edge computing nodes distribute group-level policies to the acquisition terminals in the corresponding areas. For example, they distribute a constraint of "increase RF weight by 10%" to the terminal in the "HPLC interference sensitive area" to guide it to perform the weight adjustment in step S2. At the same time, they sort the task queue according to the task arbitration rules and prioritize the execution of high-priority tasks to avoid channel resource contention. In this embodiment, multi-dimensional evaluation and dynamic weighting facilitate the improvement of the success rate of single acquisition, while reducing communication latency. Dual-mode concurrency transmits only key data, which helps reduce energy consumption. Spatiotemporal correlation and reinforcement learning help reduce the congestion rate of the RF channel and improve the acquisition efficiency during peak hours, avoiding the situation where local optima lead to global degradation. Through the closed loop of data acquisition-strategy optimization-feedback execution, this invention can adapt to long-term changes such as seasonal changes and equipment aging. When HPLC interference increases in winter, it automatically increases RF priority without manual intervention, which helps reduce operation and maintenance costs.

[0027] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power information collection method based on dual-mode communication, characterized by, The method comprises the following steps: S1, before each power information collection task is initiated, the terminal acquires multi-dimensional evaluation data of the dual-mode communication channel; S2, according to the current network state and the collection task characteristics, the weight coefficient of the multi-dimensional evaluation data is dynamically adjusted, and weighted evaluation data is output; S3, the comprehensive score of the dual-mode communication channel is calculated based on the weighted evaluation data, and the transmission scheme is selected according to the comprehensive score; S4, the edge computing node collects the whole chain data generated after the collection terminal executes steps S1-S3, and mines the space-time correlation of the whole chain data; S5, a reinforcement learning model is constructed, input data is collected by the collection terminal, a reward function is set, the reinforcement learning model is trained, and the execution action is output; S6, the execution action output by the reinforcement learning model is sent to the collection terminal and the edge computing node in the corresponding area.

2. The power information collection method based on dual-mode communication according to claim 1, characterized in that, In step S1, the method for the collection terminal to acquire multi-dimensional evaluation data of the dual-mode communication channel before each power information collection task is initiated is: The dual-mode communication channel includes an HPLC channel and an RF channel, the multi-dimensional evaluation data includes an instantaneous channel capacity prediction value, a historical communication success rate, a current channel load rate, and a node residual energy, and the instantaneous channel capacity prediction value is calculated based on a Shannon formula after preprocessing the multi-dimensional evaluation data instantaneous channel capacity at the moment , expression: wherein denotes the instantaneous channel capacity at time instant denotes the total bandwidth of the channel, denotes the logarithm function with base 2, denotes the signal-to-noise ratio related term of the -th subcarrier, denotes the transmit power at time instant denotes the average amplitude response of the -th subcarrier, denotes the average amplitude response squared of the -th subcarrier, denotes the noise power spectral density, denotes the bandwidth of the -th subcarrier, denotes the signal-to-noise ratio of the -th subcarrier, denotes the center frequency of the -th subcarrier; The collection terminal estimates the maximum data amount that can be transmitted in the future 1-4 seconds based on the current channel state information using a linear prediction algorithm, and the expression is: In the formula, Indicates the future Channel capacity prediction at time 10:00 To predict the time, Indicates the predicted time forward Historical channel capacity in milliseconds For the first The linear prediction coefficients are solved using the least squares method; The historical communication success rate is the ratio of the number of collection tasks completed to the total number of tasks within 1-3 minutes before the collection terminal is counted; and the current channel load rate is calculated by listening to the number of active nodes or the data packet collision rate in the channel.

3. The power information collection method based on dual-mode communication according to claim 2, characterized in that, In step S2, the method for dynamically adjusting the weight coefficient of the multi-dimensional evaluation data according to the current network state and the collection task characteristics, and outputting the weighted evaluation data is: The dynamic adjustment of the weight coefficient of the multi-dimensional evaluation data includes increasing the weight proportion of the historical communication success rate to 30%-50% when the network interference level is greater than or equal to a preset interference threshold; and increasing the weight proportion of the instantaneous channel capacity prediction value to 40%-60% when the collection task data amount is greater than or equal to a preset data amount threshold; The collection terminal acquires the total interference power by detecting the interference power spectral density of the dual-mode channel through the CSI detection module combined with the channel bandwidth, and then quantifies the network interference level, and the expression is: In the formula, represents the total channel interference power, is the interference power spectral density, is the first subcarrier frequency, represents the bandwidth of the first subcarrier, and the normalized processing is used to map to the interval [0, 1] to obtain the network interference level quantization value , and the preset interference threshold is , According to the actual measurement of the power scene, the HPLC channel =0.6, and the RF channel =0.5, which are used to determine whether to trigger the "historical communication success rate weight promotion" rule. The collection terminal extracts the data volume of the current collection from the task instruction , also normalized to the interval [0, 1], expression: In the formula, is the acquisition task data volume quantization value, is the maximum data volume of a single acquisition, and the preset data volume threshold is , = 0.5, corresponding to 50 KB of the HPLC channel and 25 KB of the RF channel, for judging whether to trigger the "instantaneous channel capacity weight promotion" rule.

4. The power information collection method based on dual-mode communication according to claim 3, characterized in that, In step S3, the method for calculating the comprehensive score of the dual-mode communication channel based on the weighted evaluation data and selecting the transmission scheme according to the comprehensive score is: The calculation of the comprehensive score of the dual-mode communication channel includes the comprehensive score of the HPLC channel and the comprehensive score of the RF channel, and the expression is: wherein, is the HPLC channel final comprehensive score, is the HPLC channel base weighted evaluation value obtained from the weighted evaluation data, is the HPLC channel real-time stability correction amount, is the HPLC channel historical deviation correction amount, is the RF channel final comprehensive score, is the RF channel base weighted evaluation value obtained from the weighted evaluation data, is the RF channel real-time stability correction amount, is the RF channel historical deviation correction amount, and meanwhile, the comprehensive score is maintained in the interval [0, 1] by using truncation processing. The selection of the transmission scheme includes selecting the channel with the highest comprehensive score as the main communication channel; and when the difference between the comprehensive scores of the two channels is less than or equal to a preset score threshold, triggering dual-mode concurrent transmission, and calculating the dynamic preset score threshold based on the network interference level and the task data amount; When the dual-mode concurrent transmission is triggered, the collection terminal splits the collection data into a key data segment and a non-key data segment, the key data segment is transmitted through the HPLC and RF channels at the same time, and the non-key data segment is transmitted through the main communication channel.

5. The power information collection method based on dual-mode communication according to claim 4, characterized in that, In step S4, the method for the edge computing node to collect the whole chain data generated after the collection terminal executes steps S1-S3, and mine the space-time correlation of the whole chain data is: The full-chain data includes channel selection results, multi-dimensional evaluation data, and actual communication effect data. The spatio-temporal correlation includes spatial correlation mining and temporal correlation mining. The spatial correlation mining is to analyze the channel selection preferences of the collection terminals in the same area and the same feeder branch, and the expression is: In the formula, The discrete coefficients represent the channel selection preference. This represents the average coefficient used in variance calculations. Indicates the first The percentage of HPLC selection for each terminal. This represents the average HPLC selection percentage across all terminals within the region. Indicates the first Squared HPLC selectivity of each terminal This indicates the average selection percentage of the region using HPLC. This indicates the regional HPLC channel selection preference. Indicates the regional RF channel selection preference. middle Indicates the number of times. Indicates the HPLC channel. Indicates a choice. and correspond, Indicates RF channel, This represents the total number of times the main channel was selected by all acquisition terminals within the target area. This represents the total number of times that all data acquisition terminals within the target area selected RF as the primary communication channel. This indicates the total number of times all acquisition terminals within the target area selected HPLC as the primary communication channel, identifying areas where HPLC interference remained consistently high and RF signals were stable.

6. The power information collection method based on dual-mode communication according to claim 5, wherein, In step S4, the method for the edge computing node to converge the full-chain data generated after the collection terminals perform steps S1-S3 and mine the spatio-temporal correlation of the full-chain data is: The temporal correlation mining is to count the channel quality data in each time period within 0-24 hours per day, and count the average channel quality index in each time period within 7 consecutive days, and the expression is: In the formula, represents the average channel quality index of a specific period, a specific channel, represents the average value, represents the average coefficient, represents the total number of collection terminals in the target station area, represents the channel quality raw data of a single sample, is the number of days, is the period, is the terminal number, is the channel type, identifying the electricity peak period: 8:00-21:00, the periodic law of HPLC noise increase.

7. The power information collection method based on dual-mode communication according to claim 6, wherein, In step S5, the method for constructing a reinforcement learning model, collecting input data through the collection terminal, setting a reward function, training the reinforcement learning model, and outputting an execution action is: The reinforcement learning model is a deep Q network model, and the input data includes regional network status, historical communication statistics data, and task queue information. The reward function includes a regional collection success rate, an average time delay, and a resource balance rate, and the actual regional collection success rate is set as , the target success rate is =0.985, and the expression is: In the formula, represents the collection success rate reward, represents the weight coefficient of the collection success rate reward, represents the actual collection success rate of the area, represents the target value of the collection success rate of the area, represents the ratio of the actual success rate to the target success rate.

8. The power information collection method based on dual-mode communication according to claim 7, characterized in that, In step S5, the method for constructing a reinforcement learning model, collecting input data through the collection terminal, setting a reward function, training the reinforcement learning model, and outputting an execution action is: The execution action includes a group-level channel scheduling strategy and a task priority arbitration rule. The group-level channel scheduling strategy includes but is not limited to: specifying a preferential communication mode within 6-24 minutes in the future for a specific area, and reserving 22%-32% of the RF channel bandwidth for peak hours. The task priority arbitration rule includes but is not limited to: the priority of a fault reporting task is higher than that of a batch meter reading task, and the priority of a remote fee control instruction is higher than that of an ordinary electricity data collection task.

9. The power information collection method based on dual-mode communication according to claim 8, wherein, In step S6, the method for the reinforcement learning model to output the execution action to the collection terminal and the edge computing node of the corresponding area is: The group-level channel scheduling strategy in the performing action is issued to the collection terminal of the corresponding area, as a constraint condition for the weight adjustment in step S2, and the issuing mode adopts a broadcast + unicast confirmation mechanism. The edge computing node first broadcasts a strategy data packet to the target area through an HPLC wideband carrier, the broadcast period is = 10s, and the broadcast is performed 3 times continuously. After receiving the data packet, the collection terminal verifies the checksum and the area identifier, and returns an ACK confirmation frame if the verification is matched. The ACK confirmation frame contains a terminal ID and a strategy receiving timestamp . The edge computing node calculates the ACK rate, and the expression is: In the formula, ACK feedback rate, The number of collection terminals that successfully return the confirmation frame, The total number of target collection terminals under the policy, and If not, unicast retransmission is used for the collection terminal that has not been confirmed, and the number of retransmissions is ≤2.

10. The power information collection method based on dual-mode communication according to claim 8, wherein, In step S6, the method for the reinforcement learning model to output the execution action to the collection terminal and the edge computing node of the corresponding area is: The task priority arbitration rule in the execution action is applied to the collection task queue management of the edge computing node, and high-priority tasks are preferentially executed, and the expression is: In the formula, represents the comprehensive priority score of the th task, represents the weight corresponding to the th task type, represents that the data amount influence coefficient is 0.1, represents the data amount of the th task, represents the maximum task data amount, represents that the deadline influence coefficient is 0.3, represents the deadline of the th task, represents the current time.

Images