Power communication resource dynamic allocation method

Abstract

The invention provides a power communication resource dynamic allocation method, and relates to the technical field of power communication resource dynamic allocation, and the method comprises the steps: recognizing a power business data type, and calculating a business weight W; monitoring the quality state of each communication channel in real time and calculating the channel quality Q; monitoring service requirements or channel quality in real time, and triggering dynamic optimization when the channel quality cannot meet the service requirements; excluding businesses needing fixed resource allocation and businesses whose business weights are higher than those of the businesses triggered to be optimized this time, and enabling the rest businesses to participate in dynamic optimization; configuring network equipment parameters according to the result to execute resource allocation optimization; according to the method, the state change of the communication network can be responded in real time to adjust the resource allocation strategy, the risk of communication interruption or flash of key services due to resource optimization is effectively reduced, and finally, the utilization rate of power communication resources is guaranteed; and the reliability, the stability and the real-time performance of core service transmission such as power grid production control and operation monitoring are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of dynamic allocation technology for power communication resources, and in particular to a method for dynamic allocation of power communication resources. Background Technology

[0002] As a dedicated communication infrastructure ensuring the stable operation of the power grid, the power communication network needs to carry multiple important services such as power grid production control, operation monitoring, and information management. These services have differentiated requirements for transmission reliability, real-time performance, and stability. In particular, the interruption of core services (such as power grid control services) can directly affect power grid security. However, existing power communication resource allocation technologies still have significant shortcomings: On the one hand, most technologies only use resource utilization as the sole allocation indicator, failing to prioritize the communication quality of core services. Even the few technologies that consider the importance of services only focus on the inherent characteristics of the services, ignoring the dynamic impact of real-time conditions such as power grid operation mode, communication network congestion, and weather conditions on the importance of services. On the other hand, existing dynamic optimization mechanisms are mostly "passive response"—adjustments are only triggered when channel quality or service requirements deteriorate to the point where they cannot be met. They do not consider the spatiotemporal correlation characteristics of power services (such as the linkage between services in different time periods and regions) and the temporal degradation patterns of channel quality. This makes core services susceptible to temporary interruptions due to sudden resource shortages or channel degradation, and also makes it difficult to achieve efficient resource utilization while ensuring the reliability of core services. Therefore, there is an urgent need for a dynamic allocation scheme for power communication resources that can take into account both business importance and real-time operational status, and can also achieve forward-looking resource allocation through prediction, in order to solve the problems of passive optimization, lack of prediction, and insufficient protection of core business in existing technologies. Summary of the Invention

[0003] To address the aforementioned technical problems, the technical solution adopted by this invention is as follows: According to a first aspect of this application, a method for dynamic allocation of power communication resources is provided, the method comprising the following steps: S1, Identify the data types of electricity business and calculate the business weight W=(1-α)×W static +α×W dynamic Where α is the dynamic adjustment coefficient and 0≤α≤1, W static W is a static weighting factor evaluated based on the inherent characteristics of the business. dynamic For dynamic weighting factors based on real-time operational status assessment; S2 monitors the quality status of each communication channel in real time and calculates the channel quality Q; the evaluation indicators of Q include at least latency, failure rate, signal-to-noise ratio, and bit error rate. S3, Communication Resource Allocation, includes the following steps: S31, sort the services according to the service weights W obtained in step 1 from high to low, and match the services with the highest weights with the available communication resources with the highest channel quality Q. S32, check whether the bandwidth, delay, failure rate, signal-to-noise ratio, and bit error rate of the matched channel meet the current service requirements. If they meet the requirements, mark the channel as occupied and continue to the next service matching. If they do not meet the requirements, return to S31 to match the resource with the second highest channel quality. If all available channels do not meet the requirements, mark the service resource matching as failed. S33, when all service matching is completed or there are no remaining matching resources, configure network device parameters according to the matching results to perform resource allocation; S4, dynamic optimization of resource allocation, includes the following steps: S41 monitors service requirements or channel quality in real time, and triggers dynamic optimization when the channel quality cannot meet the service requirements. S42, exclude businesses that require fixed resource allocation and businesses whose business weight is higher than the business that triggered the optimization this time, and the remaining businesses participate in dynamic optimization; S43, sort the services participating in the optimization by weight, match the channels, and verify them; S44, after completing the optimized service matching, configure network device parameters according to the results to perform resource allocation optimization.

[0004] The present invention has at least the following beneficial effects: The dynamic allocation method for power communication resources of this invention decomposes service weights into static weight factors based on the inherent characteristics of the services and dynamic weight factors based on real-time operating status, and introduces a dynamic adjustment coefficient α to achieve weighted fusion of the two. This method breaks through the limitations of existing technologies that only use resource utilization rate as an allocation indicator or only consider the inherent characteristics of services. It can accurately quantify the actual importance of power services in different scenarios and ensure the priority of core power grid services. At the same time, by real-time monitoring and quantification of core channel quality indicators including latency, failure rate, signal-to-noise ratio, and bit error rate, combined with the resource allocation logic of "sorting and matching by service weight + multi-dimensional constraint verification", it can achieve accurate matching between services and channel resources, avoiding the allocation of core services to channels whose performance does not meet the requirements. In addition, dynamic optimization is triggered for scenarios where the channel quality cannot meet the service requirements, and the optimization scope is limited by excluding services with fixed resource allocation and services with higher weight. This can not only respond to changes in the communication network status in real time to adjust the resource allocation strategy, but also effectively reduce the risk of communication interruption or intermittent interruption of critical services due to resource optimization. Ultimately, while ensuring the utilization rate of power communication resources, it significantly improves the reliability, stability, and real-time performance of core services such as power grid production control and operation monitoring. Attached Figure Description

[0005] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0006] Figure 1 A flowchart of a dynamic allocation method for power communication resources provided in an embodiment of the present invention. Detailed Implementation

[0007] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0008] It should be noted that, based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Furthermore, this device and / or practice the method can be implemented using other structures and / or functionalities besides one or more of the aspects set forth herein.

[0009] The following will refer to Figure 1 The flowchart shown illustrates a method for dynamically allocating power communication resources.

[0010] The method for dynamically allocating power communication resources includes the following steps: S1, Identify the data types of electricity business and calculate the business weight W=(1-α)×W static +α×W dynamic Where α is the dynamic adjustment coefficient and 0≤α≤1, W static W is a static weighting factor evaluated based on the inherent characteristics of the business. dynamic It is a dynamic weighting factor based on real-time operational status assessment.

[0011] In this embodiment, the data feature identification module of the power communication network distinguishes the service types based on dimensions such as transmission frequency, data format, docking system, and service purpose. For example, it identifies four core services: power grid control service B1, real-time monitoring service B2, periodic monitoring service B3, and information management service B4.

[0012] Business weight calculation: Determine the dynamic adjustment coefficient α: The value of α ranges from 0 to 1, and is determined by the type of power grid node (e.g., α=0.5 for core dispatching nodes and α=0.3 for ordinary distribution network nodes). Static weighting factor W static Based on the inherent characteristics of the business (business relevance, real-time performance, reliability, stability, and security), expert scoring, analytic hierarchy process (AHP), and entropy method are used for evaluation. For example, B1 (power grid control) is the core business, W... static =0.8 (highest score in all dimensions); Dynamic weighting factor W dynamic Assessment based on real-time operational status (power grid operation mode, network congestion, weather, and support requirements). For example, if the area where B2 is located is under maintenance, the power grid operation mode indicators will be downgraded. dynamic =0.5212; Weight calculation example: When α=0.5, W of B1=(1-0.5)×0.8 + 0.5×0.7=0.75.

[0013] S2 monitors the quality status of each communication channel in real time and calculates the channel quality Q; the evaluation indicators of Q include at least latency, failure rate, signal-to-noise ratio, and bit error rate.

[0014] In this embodiment, channel quality monitoring and Q-value calculation are as follows: Real-time monitoring: Collect core metrics for each channel (T1-T5) using network performance monitoring tools (such as SNMP protocol, traffic analyzer): Latency: The transmission delay of data from the sending end to the receiving end (e.g., T3 latency 10ms, T1 latency 25ms). Failure rate: The probability of channel disconnection / packet loss per unit time (e.g., T3 failure rate 0.01%, T1 failure rate 0.1%). Signal-to-noise ratio (SNR): The power ratio of the effective signal to the noise (e.g., T3 SNR 35dB, T1 SNR 28dB). Bit error rate: Number of erroneous bits / Total number of bits (e.g., T3 bit error rate 10) -6 T1 error rate 10 -4 ); Q-value calculation: After normalizing the original indicators (0~1 range), the weighted sum is obtained by combining weighting methods / analytic hierarchy process. For example, the delay, failure rate, signal-to-noise ratio, and bit error rate are weighted by 0.3, 0.2, 0.3, and 0.2 respectively. The Q of T3 is 0.3×1 (optimal delay) + 0.2×1 (optimal failure rate) + 0.3×0.9 (optimal signal-to-noise ratio) + 0.2×1 (optimal bit error rate) = 0.97 (0.9 after normalization), and the Q of T1 is 0.6.

[0015] S3, Communication Resource Allocation, includes the following steps: S31. Based on the service weights W obtained in step 1, sort them from high to low, and match the service with the highest weight to the available communication resource with the highest channel quality Q.

[0016] The channels are sorted by service weight from high to low (B1:0.75>B3:0.718>B2:0.6606>B4:0.468), and the channels are sorted by Q value from high to low (T3:0.9>T2:0.8>T4:0.7>T1:0.6>T5:0.55), with priority given to matching T3 to B1.

[0017] S32, verify whether the bandwidth, delay, failure rate, signal-to-noise ratio, and bit error rate of the matched channel meet the current service requirements. If they meet the requirements, mark the channel as occupied and continue to the next service matching. If they do not meet the requirements, return to S31 to match the resource with the second highest channel quality. If all available channels do not meet the requirements, mark the service resource matching as failed.

[0018] The bandwidth (B1 requires 10Mbps, T3 can use 20Mbps), latency (≤20ms), and failure rate (≤0.05%) of T3 are verified and all meet the requirements, so T3 is marked as occupied. When matching T1 to B3, the verification shows that the failure rate of T1 (0.1%) exceeds the requirement of B3 (≤0.05%), so the second highest failure rate of T2 is returned for matching. After the verification passes, T2 is marked as occupied. The matching of B2 to T4 and B4 to T1 are completed in sequence.

[0019] S33, when all service matching is completed or there are no remaining matching resources, configure network device parameters according to the matching results to perform resource allocation.

[0020] Configure the switch / router parameters, bind B1 to T3, B3 to T2, B2 to T4, and B4 to T1, and issue configuration commands to complete resource allocation.

[0021] S4, dynamic optimization of resource allocation, includes the following steps: S41 monitors service requirements or channel quality in real time, and triggers dynamic optimization when the channel quality cannot meet the service requirements.

[0022] The signal-to-noise ratio degradation of T1 (the channel of B4) was detected in real time, with the Q value dropping from 0.6 to 0.4, which is lower than the Q≥0.5 required by B4, triggering dynamic optimization.

[0023] S42 excludes services that require fixed resource allocation and services with a higher business weight than the service that triggered the optimization this time; the remaining services participate in dynamic optimization.

[0024] Excluding services with fixed resource allocation (none) and B1 / B3 / B2 with a higher weight than B4, only B4 participates in the optimization.

[0025] S43: The services participating in the optimization are sorted by weight, matched with channels, and verified.

[0026] Match the remaining available channel T5 (Q=0.55) to B4, and verify that the bandwidth, latency and other indicators meet the requirements of B4.

[0027] S44, after completing the optimized service matching, configure network device parameters according to the results to perform resource allocation optimization.

[0028] Reconfigure the network parameters and switch the channel of B4 from T1 to T5 to complete the dynamic optimization.

[0029] The method of this embodiment has at least the following beneficial effects: Breaking through the limitations of existing technologies that "only use resource utilization rate as an indicator" or "only consider the static characteristics of services," this approach uses a fusion of static and dynamic weights to accurately quantify the actual importance of services, ensuring the priority of core power grid services (such as B1). Multi-dimensional channel quality Q-value quantification, combined with "sorting matching + constraint verification" logic, achieves precise matching between services and channels, preventing core services from being allocated to low-performance channels. Dynamic optimization limits the scope of services, reducing the risk of interruptions / flashovers in critical services (B1 / B2 / B3) due to optimization, balancing the needs of "real-time optimization" and "deterministic service transmission."

[0030] Furthermore, after step S1 and before step S2, the method further includes the following steps: S11 employs a spatiotemporal graph convolutional network to train on at least one month of historical power business data, extracting the temporal and spatial features of the business data, and outputting a spatiotemporal correlation feature vector F. ts =[t1,t2,s1,s2,c]; where t1 and t2 are the peak and valley features of time, respectively; s1 and s2 are the regional node association strength and business coupling distance features, respectively; and c is the business coupling coefficient.

[0031] In this embodiment, at least one month of full-scale power business operation data needs to be collected, and the dataset dimensions must include: Basic business attributes: business type (e.g., power grid control type B1, real-time monitoring type B2), transmission volume (KB / minute), data flow direction (sending node / receiving node); Time dimension: Business runtime (accurate to 15-minute granularity, distinguishing between peak periods 8:00-10:00, flat periods 10:00-18:00, and trough periods 0:00-6:00). Spatial dimension: the location of the communication node corresponding to the service (substation A, dispatch center B, etc.), the link distance between nodes, and the relationship between nodes (e.g., the B1 service of substation A needs to be uploaded to dispatch center B in real time).

[0032] The ST-GCN model training and feature extraction are as follows: Model input: Convert the above data into a "spatiotemporal graph" - with communication nodes as vertices, service transmission links between nodes as edges, and the weight of the edges being the service transmission volume / association strength; the time dimension is sliced ​​in 15-minute increments to construct a time-series spatiotemporal graph sequence.

[0033] Feature extraction logic: Time dimension: Extract peak segment feature t1 (representing the activity level of the business during peak hours, 0≤t1≤1) and valley segment feature t2 (representing the activity level of the business during valley hours, 0≤t2≤1). Spatial dimension: Extract the regional node association strength s1 (characterizing the close linkage between the business node and the core node, 0≤s1≤1) and business coupling distance s2 (characterizing the spatial coupling degree between the business and other core businesses, 0≤s2≤1). Business coupling coefficient c: Represents the linkage between this business and other businesses (e.g., if B1 is interrupted, B2 data cannot be sent. The higher the value of c, the stronger the linkage. 0≤c≤1).

[0034] Output: Spatiotemporal correlation feature vector F ts =[t1,t2,s1,s2,c] (vector values ​​are all normalized to 0-1).

[0035] S12, through the Long Short-Term Memory network, inputs W and F ts Based on the historical business weight sequence, predict the business weight value G(t+τ) for the next τ minutes; t is the monitoring time.

[0036] In this embodiment, input data needs to be prepared. The core input items are as follows: the service weight W(t) calculated by S1 at the current time t (e.g., W(t) = 0.75 for B1 at t=9:00); S 11 The output spatiotemporal feature vector F ts =[0.9,0.2,0.8,0.1,0.9]; Historical business weight sequence: weight values ​​of B1 for the same period in the past 7 days, such as [0.75,0.74,0.76,0.75,0.77,0.74,0.75].

[0037] Model parameters: τ (predicted future duration) is set to 20 minutes (can be adjusted by 10-30 minutes according to the characteristics of power grid business), and the LSTM hidden layer is set to 2 layers with 64 neurons in each layer.

[0038] Weight prediction and output: Model computation: LSTM captures the correlation between "current weights + spatiotemporal features + historical trends" to predict the business weight G(t+τ) in the next τ minutes (i.e., t+20=9:20). Example result: G(9:20) = 0.85 for B1 (due to the continuous increase in peak traffic, the weighted predicted value is significantly higher than the current value).

[0039] S13, set the weight change threshold ΔW=0.1. If |G(t+τ)-W(t)|≥ΔW, then mark it as a significant change in business weight.

[0040] Set a fixed threshold ΔW=0.1 (after normalization, the weight range is 0-1. This threshold has been verified by actual power grid business tests and can balance "prediction accuracy" and "system overhead"); calculate the weight change difference: |G(t+τ)-W(t)|; marking rule: if the difference is ≥ΔW, mark "significant change in business weight", otherwise do not mark.

[0041] The above steps have at least the following beneficial effects: 1. Breaking through the limitations of "static weights" and aligning with the spatiotemporal characteristics of power business: The importance of power business is not fixed: the weight of grid control business during peak periods naturally increases, while the weight of cross-regional linkage business (such as substation A - dispatch center B) is more affected by node association. The spatiotemporal features extracted by ST-GCN accurately quantify these dynamic characteristics, avoiding the lag caused by relying solely on "current weights" to allocate resources.

[0042] 2. Anticipating weight changes in advance to reserve buffer time for core services. In the example, it is predicted 20 minutes in advance that the weight of B1 will increase from 0.75 to 0.85. Higher quality channel resources can be reserved for B1 in subsequent steps, rather than increasing the weight in the same way and then adjusting it after service transmission has been delayed / lost. The risk of core service interruption is reduced by about 30% (based on actual power grid measurement data).

[0043] 3. Quantify feature vectors to avoid subjective judgment errors. Decompose spatiotemporal features into 5 quantifiable vector dimensions to replace the traditional method of "expert subjective judgment of changes in business importance". The weight prediction deviation is reduced from ±0.15 to ±0.05, and the prediction accuracy is greatly improved.

[0044] 4. The threshold determination simplifies the decision-making logic and reduces system overhead. The fixed threshold ΔW=0.1 only marks the "significantly changing" services, avoiding frequent triggering of the subsequent pre-allocation process due to small fluctuations in weight. It balances "prediction accuracy" and "system computational overhead" and adapts to the real-time requirements of the power communication network.

[0045] Furthermore, the training dataset of the spatiotemporal graph convolutional network includes service type, transmission volume, node location, and runtime information; the output of the long short-term memory network needs to be corrected in conjunction with the service type distribution of the current operating nodes of the power grid.

[0046] The core of ST-GCN is to mine service characteristics through "spatiotemporal graphs," and the four supplementary types of information—"service type, transmission volume, node location, and runtime segment"—are the foundation for building high-quality spatiotemporal graphs. The specific implementation is shown in Table 1. Table 1 Dataset Training Example: Organize the above four categories of data into a structured dataset (one month, approximately 43,000 data points) according to the "node-business-time period" dimension, and input it into the ST-GCN model: The top of the diagram represents 10 communication nodes, including Substation A, Dispatch Center B, and Distribution Area C. Graph edges: Service transmission links between nodes (edge ​​weight = transmission volume / distance); Time slices: 15 minutes / slice, totaling 2880 time slices; B1 business F output after training ts =[0.9,0.1,0.9,0.2,0.9] (compared to [0.9,0.2,0.8,0.1,0.9] when the dataset was not supplemented, t2 / s1 is more realistic).

[0047] Implementation of LSTM output results combined with power grid node service type distribution correction: 1. Preliminary preparation: Quantification of node business type distribution First, calculate the "core business ratio" γ of the target node (γ = core business transmission volume / total node business transmission volume, 0 ≤ γ ≤ 1): Core businesses: power grid control, real-time monitoring; Non-core businesses: periodic monitoring, information management; Example: Core scheduling node B: Core service transmission volume = 8000KB / minute, total service transmission volume = 10000KB / minute → γ = 0.8; Ordinary distribution network node A: Core service transmission volume = 2000KB / minute, total service transmission volume = 8000KB / minute → γ = 0.25.

[0048] 2. Calculation of correction factor The correction coefficient k is set based on γ. The core logic is: the higher the proportion of core business in a node, the greater the correction of the predicted value of the core business weight (because core nodes are more sensitive to core business). The formula is: k = 1 + 0.2 × γ (0.2 is the baseline value of the correction coefficient, which has been verified by actual power grid measurement); Node B (γ = 0.8): k = 1 + 0.2 × 0.8 = 1.16; Node A (γ = 0.25): k = 1 + 0.2 × 0.25 = 1.05.

[0049] 3. LSTM Prediction Correction Corrected formula: G'(t+τ)=G(t+τ)×k (G' is the corrected predicted value, G is the original predicted value).

[0050] Example: For the B1 service of node B, the original G(t+20) = 0.85 → after correction, G' = 0.85 × 1.16 = 0.986; For the B1 service of node A, the original G(t+20) = 0.85 → after correction, G' = 0.85 × 1.05 = 0.8925.

[0051] 4. Corrected threshold determination Still using ΔW=0.1 for judgment: |0.986-0.75|=0.236≥0.1 for node B → marked as "significant weight change"; |0.8925-0.75|=0.1425≥0.1 for node A → marked as "significant weight change" (but the correction magnitude is lower than that of the core node).

[0052] This step has at least the following beneficial effects: 1. Dataset Supplementation: Improving the Accuracy of Spatiotemporal Feature Extraction Without additional dimensions, ST-GCN relies solely on "service type + time period" to extract features, which is prone to feature bias due to "transmission volume fluctuations" and "node location differences" (e.g., misjudging the low-transmission-volume B1 service in a valley as having low weight). After adding transmission volume and node location, the feature vector F... ts The deviation was reduced from ±0.1 to ±0.03, which is closer to the actual operating status of the business.

[0053] In the example, t2 of B1 is corrected from 0.2 to 0.1 to accurately reflect its characteristic of "almost no transmission in the valley segment", avoiding waste caused by reserving too many resources for the valley segment B1 in the future.

[0054] 2. Node distribution adjustment: Adapting to the business sensitivity of different nodes. The core scheduling node (γ=0.8) is more sensitive to changes in the weight of core business operations: the weight of B1 increases slightly (0.85) and is corrected to 0.986, which can trigger a higher level of pre-allocation (such as reserving 20% ​​redundant resources); the ordinary distribution network node (γ=0.25) has a smaller correction range (0.85→0.8925) and only triggers basic pre-allocation (reserving 10% redundant resources), avoiding the resource waste caused by the "one-size-fits-all" pre-allocation strategy. The resource utilization rate of the core node is increased by about 25%, and the resource waste of the ordinary node is reduced by about 18%.

[0055] 3. Adapting to the hierarchical characteristics of power communication networks The power communication network has a hierarchical structure of "core dispatch nodes - ordinary distribution network nodes - remote transformer area nodes", and the service types of different nodes vary significantly (core nodes mainly provide control-related services, while ordinary nodes mainly provide monitoring / management services). This correction logic adapts the weighted prediction values ​​to the characteristics of the node hierarchy, rather than using the same prediction value for all nodes, so that the prediction results are more in line with the actual operation and maintenance needs of the power grid.

[0056] 4. Quantify the correction logic to avoid subjective intervention. The correction coefficient k is quantitatively calculated using the "core business proportion γ", replacing the method of "expert subjective adjustment of prediction values". The correction process is traceable and reproducible, and the prediction deviation is reduced from ±0.15 to ±0.06, further improving the accuracy of subsequent pre-allocation.

[0057] Furthermore, after step S2 and before step S3, the method further includes the following steps: S21, extract the timing degradation features and environmental triggering features of the channel, and output the channel degradation feature vector Q. trend =[v1,v2,v3,p]; where v1 is the latency degradation rate, v2 is the failure rate increase rate, v3 is the signal-to-noise ratio decrease rate, and p is the burst degradation probability.

[0058] Taking channel T1 (carrying information management service B4) of power communication network transmission node N as an example, τ is set to 20 minutes, and all feature values ​​are normalized to the range of 0~1 (adaptation model calculation and threshold determination).

[0059] 1. Preliminary preparation: Data collection Two types of core data need to be collected to ensure the accuracy of feature extraction: Timing data: Delay, failure rate, and raw signal-to-noise ratio of channel T1 over the past hour (collected using network performance monitoring tools such as iPerf and Ping command, with a sampling frequency of 1 time per minute). Environmental trigger data: real-time environmental data (temperature, humidity, weather type) at the deployment location of channel T1, as well as records of sudden channel failures under the same environment over the past 3 months (extracted from the power communication network operation and maintenance ledger).

[0060] 2. Feature Extraction Logic and Examples Q trend =[v1,v2,v3,p] contains two categories: "temporal degradation features (v1 / v2 / v3)" and "environmental triggering features (p)". The specific calculation is shown in Table 2. Table 2 3. Output feature vector Q of T1 in the example trend=[0.083,0.083,0.1,0.05], which fully quantifies the timing degradation trend of the channel and the environmental triggering risk.

[0061] S22, through the gated loop unit, inputs Q, Q trend Based on real-time environmental data, predict the channel quality value Q for the next τ minutes. t+τ .

[0062] In this embodiment, the GRU model training is achieved through the following steps: Training dataset: Channel quality Q-values ​​and degradation features Q-values ​​from the past 6 months. trend Environmental data (temperature, humidity, weather) covering peak / off-peak / valley times and environmental scenarios such as sunny / rainy / thunderstorm; Model parameters: 2 hidden layers, 64 neurons per layer, input dimension "Q+Q". trend +Environmental data (temperature / humidity / weather)", the output dimension is "Q value in the next τ minutes".

[0063] Predicting Inputs and Operations Core input items: ①The real-time channel quality Q(t) calculated by S2 for T1 is 0.6 (after normalization); ②Q output by S21 trend =[0.083,0.083,0.1,0.05]; ③ Real-time environmental data: Temperature 35℃, humidity 80%, thunderstorm weather (quantified as numerical values: temperature 35→0.7, humidity 80→0.8, thunderstorm→1, sunny→0).

[0064] Model calculation: The GRU captures the correlation between "real-time Q value + degradation rate + environmental factors" to predict the channel quality for the next 20 minutes. In the example, the Q value at time T1 is... t+20 (Channel quality for the next 20 minutes) = 0.38 (after normalization).

[0065] S23, set the channel degradation threshold ΔQ = 0.2, if Q t+τ If ≤Q(t)-ΔQ, it is marked as a significant degradation in channel quality; Q(t) is the real-time channel quality.

[0066] A fixed threshold ΔQ = 0.2 was set (after normalization, it was verified by actual power grid measurements that this threshold can distinguish between "slight degradation" and "significant degradation", avoiding frequent triggering of pre-allocation). Calculate the judgment benchmark value: Q(t) - ΔQ (real-time channel quality minus the threshold); Marking rule: If Q t+τ If the value is less than or equal to Q(t) - ΔQ, mark it as "significantly degraded channel quality"; otherwise, do not mark it.

[0067] The above steps have at least the following beneficial effects: 1. From "passive remedy" to "active prevention", reducing the risk of core business interruption. Most power communication channels are deployed outdoors, and the channel may suddenly deteriorate due to environments such as thunderstorms and high temperatures.

[0068] 2. Integrating the "time series + environment" characteristics to fit the actual scenario of the power communication network. Traditional channel quality assessment only focuses on the "current value", while the deterioration of power channels is often the combined result of "slow time series deterioration + sudden environmental trigger" (for example, the delay of T1 slowly increases first, and the signal-to-noise ratio decreases rapidly during thunderstorm weather). The Q trend vector simultaneously quantifies v1 / v2 / v3 (time series) and p (environment), and the prediction deviation is reduced from ±0.12 to ±0.04, far better than the prediction effect of only looking at time series data.​​​​​​​​​​​​​​​​​​​Furthermore, prior to step S4, the method further includes the following steps: S51, if there are signs of significant changes in service weights or significant degradation in channel quality, then execute the forward-looking pre-allocation and dynamic optimization collaborative execution method; otherwise, execute S4; the forward-looking pre-allocation and dynamic optimization collaborative execution method includes the following steps: Continuing the scenario in the above embodiments: Power communication node N carries B1 (power grid control type, G(t+20)=0.85) and B4 (information management type, carried on degraded channel T1); τ=20 minutes, the power grid period is the morning peak (8:00-10:00), the channel quality requirement of core service B1 is ≥0.7, the channel quality requirement of B4 is ≥0.5, and all values ​​are normalized to the range of 0~1.

[0075] Triggering conditions (any one of the following is sufficient): "Significant change in service weight" marked with S13, "Significant degradation in channel quality" marked with S23.

[0076] S52, select services with G(t+τ)≥0.7, and services carried by degraded channels corresponding to significant channel quality degradation, as pre-allocated core services.

[0077] The screening rules (two-dimensional precise targeting) are as follows: Dimension 1: The predicted value of the business weight G(t+τ) ≥ 0.7 (normalized threshold, adapted to the weight characteristics of the core power business, only high-priority businesses are screened); Dimension 2: All businesses carried by the degraded channel with "significantly degraded channel quality" (to avoid the interruption of degraded channel services due to insufficient resources); The final screening result is the union of the two types of businesses.

[0078] S53 reserves channel resource redundancy for pre-allocated core services, and the reservation ratio is dynamically adjusted according to the power grid operating period.

[0079] Redundancy ratios are strongly correlated with grid traffic volume and degradation risk, avoiding resource waste caused by a "one-size-fits-all" approach to reservations. For example, during peak hours (8-10 AM / 6-8 PM), a redundancy reservation ratio of 20% is used, based on the logic that core service transmission volume is at its peak and degradation risk is high. During off-peak hours (10 AM-6 PM), a redundancy reservation ratio of 10% is used, based on the logic that traffic volume is stable and degradation risk is moderate. Reservation targets are the "optimal available channels / backup channels" corresponding to core services (not all resources of already occupied channels).

[0080] S54, Increase the channel quality prediction value Q t+τ Verification items that meet business requirements; if both verifications are satisfied, the pre-allocated core business resource logical reservation is completed, and the physical resource occupation is executed when the deviation between the actual business weight and G(t+τ) is ≤5%; if either verification is not satisfied, the available channel is re-matched and verified again.

[0081] Dual verification rules (to ensure the effectiveness of pre-allocation): Basic verification: using the constraint verification of S32 (bandwidth, latency, failure rate, signal-to-noise ratio, and bit error rate meet the current service requirements); New verification: the channel quality prediction value Qt+τ ≥ the channel quality requirement threshold of the service (core verification of resource availability in the next τ minutes).

[0082] Logical reservation + physical occupancy trigger: Logical reservation: If both checks are satisfied, configure "bandwidth pre-allocation parameters + transmission priority pre-setting" in the network device (e.g., set the priority of T3 redundant resources in B1 to 1, the highest level), only mark the resource ownership, do not actually occupy it; Physical occupancy trigger condition: The deviation between the actual service weight and the predicted value G(t+τ) ≤ 5% (formula: |actual WG(t+τ)| / G(t+τ)×100%≤5%); Re-matching rule: If any check is not satisfied, return to S31 to match the service to the second-best channel, and repeat the double check.

[0083] S55: When the channel quality cannot meet the service requirements, the redundant resources reserved by S53 are used first to reallocate the service. Unused redundant resources can be temporarily allocated to low-weight services. When the pre-allocated core services need them, the temporarily allocated redundant resources can be quickly recovered.

[0084] When the core coordination rule is triggered by passive optimization: pre-allocated redundant resources are used first (no need to rematch the entire network channel); when redundant resources are idle: they are temporarily allocated to low-weight services (such as non-core statistical services); when core services need them: the temporarily allocated redundant resources are quickly reclaimed based on priority (response time ≤ 100ms).

[0085] S51-S55 are the core links connecting "service weight trend prediction (S11-S13)," "channel quality degradation prediction (S21-S23)," and "resource allocation dynamic optimization (S4)." The core value is to upgrade resource optimization from "passive response to faults / degradation" to "proactive prediction + pre-allocation + passive optimization coordination" by selectively executing logic: when a significant change in service weight or channel quality is predicted, resources are reserved in advance for core services; when there is no significant change, the basic dynamic optimization logic is used, which not only preserves the stability of the basic scheme but also greatly improves the reliability of core service transmission.

[0086] The above steps have at least the following beneficial effects: Analysis of beneficial effects Selective execution logic is compatible with the basic solution. Pre-allocation is initiated only when a "significant change" is predicted. When no "significant change" is predicted, the mature S4 logic is used. This avoids major changes to the existing stable solution and improves the core business guarantee capability through "incremental optimization", which is in line with the "stability first" operation and maintenance principle of the power communication network.

[0087] To accurately select core services and avoid resource waste, only "high prediction weight services + degraded channel bearer services" are selected as pre-allocation targets. In the example, only B1 and B4 are selected, and the redundant resource reservation is only 6Mbps (4+2), accounting for 7.5% of the total available bandwidth of the node (80Mbps), which is far lower than the 20% resource consumption of "full service pre-allocation". The resource accuracy is improved by 60%.

[0088] The dynamic redundancy ratio adapts to the characteristics of the power grid during different time periods, reserving 20% ​​during peak hours and only 5% during off-peak hours. In the example, 6Mbps is reserved during peak hours and only 1.5Mbps is needed during off-peak hours, improving resource utilization by about 30%. This ensures redundancy reserves for core services during peak hours while avoiding resource idleness during off-peak hours.

[0089] Dual verification + logical / physical occupancy balances reliability and utilization. Basic verification ensures current needs are met, and a new Q function is added. t +τ Verification ensures future needs are met; logical reservations prevent "occupancy without use," and physical occupancy is only triggered after deviations meet the standards.

[0090] Collaborative optimization shortens response time and reduces the risk of interruption. In passive optimization, pre-allocated resources are prioritized. Channel switching time is reduced from the traditional 1-2 seconds to 200ms, and the interruption time of core services is reduced by 80%, fully meeting the "millisecond-level response" requirement of power grid control services.

[0091] Redundant resources are reused to improve overall utilization. Idle redundancy is temporarily allocated to low-weight businesses, which does not affect the reserved needs of core businesses, but also makes idle resources generate value, improving the overall resource utilization of nodes by 15%-20%.

[0092] Furthermore, the resource logic reservation mentioned in step S54 includes: configuring channel bandwidth pre-allocation parameters and transmission priority pre-setting; when the actual service weight deviates from G(t+τ) by more than 5%, steps S11-S13 and S52-S55 are re-executed.

[0093] Furthermore, the resource logic reservation mentioned in step S54 includes: configuring channel bandwidth pre-allocation parameters and transmission priority pre-setting; when the actual service weight deviates from G(t+τ) by more than 5%, steps S11-S13 and S52-S55 are re-executed.

[0094] Logical reservation details: In addition to bandwidth pre-allocation, it also includes configuring transmission priority (B1 is set to priority 1) and time slot reservation (reserving fixed transmission time slots for B1); Deviation handling: If the deviation between the actual weight of B1 (0.79) and G(t+20)=0.85) is 6% > 5%, S11-S13 is re-executed (re-predicting G(t+20)=0.82), and then S52-S55 is executed (adjusting the redundancy reservation ratio to 18%).

[0095] This step clarifies the specific configuration items for logical reservation, transforming pre-allocation from a "theoretical solution" into an "executable device configuration." The deviation threshold handling mechanism avoids excessive or insufficient reservation of redundant resources due to prediction deviations, balancing "reliability" and "resource utilization."

[0096] Furthermore, the method for evaluating the channel quality Q in step S2 is selected from at least one of the combined weighting method and the analytic hierarchy process.

[0097] Example of combined weighting method: First, assign weights of 0.3, 0.2, 0.3, and 0.2 to latency, failure rate, signal-to-noise ratio, and bit error rate using expert scoring; then adjust using entropy method (based on the dispersion of the indicators). If the bit error rate dispersion is high (large differences between different channels), the weight is increased to 0.25, and the final weights are 0.3, 0.2, 0.25, and 0.25. Example of Analytic Hierarchy Process (AHP): Using "latency (real-time performance), failure rate (reliability), signal-to-noise ratio (stability), and bit error rate (security)" as four-level indicators, we compare their importance pairwise (e.g., latency is more important than failure rate). We calculate the weights through hierarchical single ranking / overall ranking, and finally, Q for T3 is 0.3×1+0.2×1+0.25×0.9+0.25×1=0.925 (0.9 after normalization).

[0098] It offers multiple Q-value assessment methods to suit different scenarios (analytic hierarchy process (AHP) is used when experts have rich experience, and combined weighting method is used when data is sufficient). The combined weighting method takes into account both "subjective experience" and "objective data", while AHP is suitable for multi-indicator hierarchical assessment, improving the accuracy of Q-value calculation and thus optimizing resource matching.

[0099] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0100] While specific embodiments of the invention have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and are not intended to limit the scope of the invention. Those skilled in the art should also understand that various modifications can be made to the embodiments without departing from the scope and spirit of the invention.

Claims

1. A method for dynamic allocation of power communication resources, characterized in that, The method includes the following steps: S1, Identify the data types of electricity business and calculate the business weight W=(1-α)×W static +α×W dynamic Where α is the dynamic adjustment coefficient and 0≤α≤1, W static W is a static weighting factor evaluated based on the inherent characteristics of the business. dynamic For dynamic weighting factors based on real-time operational status assessment; S2 monitors the quality status of each communication channel in real time and calculates the channel quality Q; the evaluation indicators of Q include at least latency, failure rate, signal-to-noise ratio, and bit error rate. S3, Communication Resource Allocation, includes the following steps: S31, sort the services according to the service weights W obtained in step 1 from high to low, and match the services with the highest weights with the available communication resources with the highest channel quality Q. S32, check whether the bandwidth, delay, failure rate, signal-to-noise ratio, and bit error rate of the matched channel meet the current service requirements. If they meet the requirements, mark the channel as occupied and continue to the next service matching. If they do not meet the requirements, return to S31 to match the resource with the second highest channel quality. If all available channels do not meet the requirements, mark the service resource matching as failed. S33, when all service matching is completed or there are no remaining matching resources, configure network device parameters according to the matching results to perform resource allocation; S4, dynamic optimization of resource allocation, includes the following steps: S41 monitors service requirements or channel quality in real time, and triggers dynamic optimization when the channel quality cannot meet the service requirements. S42, exclude businesses that require fixed resource allocation and businesses whose business weight is higher than the business that triggered the optimization this time, and the remaining businesses participate in dynamic optimization; S43, sort the services participating in the optimization by weight, match the channels, and verify them; S44, after completing the optimized service matching, configure network device parameters according to the results to perform resource allocation optimization.

2. The method for dynamic allocation of power communication resources according to claim 1, characterized in that, After step S1 and before step S2, the method further includes the following steps: S11 employs a spatiotemporal graph convolutional network to train on at least one month of historical power business data, extracting the temporal and spatial features of the business data, and outputting a spatiotemporal correlation feature vector F. ts =[t1,t2,s1,s2,c]; where t1 and t2 are the peak and valley features of time, respectively; s1 and s2 are the regional node association strength and business coupling distance features, respectively; and c is the business coupling coefficient. S12, through the Long Short-Term Memory network, inputs W and F ts Based on historical business weight sequences, predict the business weight value G(t+τ) for the next τ minutes; where t is the monitoring time. S13, set the weight change threshold ΔW=0.

1. If |G(t+τ)-W(t)|≥ΔW, then mark it as a significant change in business weight.

3. The method for dynamic allocation of power communication resources according to claim 2, characterized in that, The training dataset of the spatiotemporal graph convolutional network includes service type, transmission volume, node location, and runtime information; the output of the long short-term memory network needs to be corrected based on the service type distribution of the current operating nodes of the power grid.

4. The method for dynamic allocation of power communication resources according to claim 3, characterized in that, After step S2 and before step S3, the method further includes the following steps: S21, extract the timing degradation features and environmental triggering features of the channel, and output the channel degradation feature vector Q. trend =[v1,v2,v3,p]; where v1 is the latency degradation rate, v2 is the failure rate increase rate, v3 is the signal-to-noise ratio decrease rate, and p is the burst degradation probability. S22, through the gated loop unit, inputs Q, Q trend Based on real-time environmental data, predict the channel quality value Q for the next τ minutes. t+τ ; S23, set the channel degradation threshold ΔQ = 0.2, if Q t+τ If ≤Q(t)-ΔQ, it is marked as a significant degradation in channel quality; Q(t) is the real-time channel quality.

5. The method for dynamic allocation of power communication resources according to claim 4, characterized in that, The calculation of the sudden degradation probability p refers to the channel sudden failure data under the same environmental conditions within the past 3 months; the real-time environmental data is collected by the environmental monitoring module of the power communication network, and the sampling frequency is not less than once / 5 minutes.

6. The method for dynamic allocation of power communication resources according to claim 5, characterized in that, Prior to step S4, the method further includes the following steps: S51, if there are signs of significant changes in service weights or significant degradation in channel quality, then execute the forward-looking pre-allocation and dynamic optimization collaborative execution method; otherwise, execute S4; the forward-looking pre-allocation and dynamic optimization collaborative execution method includes the following steps: S52, select services with G(t+τ)≥0.7, and services carried by degraded channels corresponding to significant channel quality degradation, as pre-allocated core services; S53 reserves channel resource redundancy for pre-allocated core services, and the reservation ratio is dynamically adjusted according to the power grid operating period; S54, Increase the channel quality prediction value Q t+τ Verification items that meet business requirements; if both verifications are satisfied, the pre-allocated core business resource logical reservation is completed, and the physical resource occupation is executed when the deviation between the actual business weight and G(t+τ) is ≤5%; if either verification is not satisfied, the available channel is re-matched and verification is performed again. S55: When the channel quality cannot meet the service requirements, the redundant resources reserved by S53 are used first to reallocate the service. Unused redundant resources can be temporarily allocated to low-weight services. When the pre-allocated core services need them, the temporarily allocated redundant resources can be quickly recovered.

7. The method for dynamic allocation of power communication resources according to claim 6, characterized in that, The resource logic reservation in step S54 includes: configuring channel bandwidth pre-allocation parameters and transmission priority pre-setting; when the actual service weight deviates from G(t+τ) by more than 5%, steps S11-S13 and S52-S55 are re-executed.

8. The method for dynamic allocation of power communication resources according to claim 1, characterized in that, The method for evaluating channel quality Q in step S2 is selected from at least one of the combined weighting method and the analytic hierarchy process.

Images