Safety communication system based on power distribution wireless private network EPDT-RSMA

Abstract

The invention discloses an EPDT-RSMA secure communication system based on a power distribution wireless private network, which is characterized in that a'perception-communication-calculation-control 'closed loop is constructed between a distributed power distribution terminal and a master station, limited blocks of long and short packets are split into public and private streams by adopting RSMA, and double indexes of security information age outage probability SAOP and weighted average information age W-AOI are introduced; meanwhile, a momentum-diffusion coupling learning algorithm is provided to output the optimal power division factor and the block length in real time, and the real-time performance and the safety of the power grid are ensured.

Description

Technical Field

[0001] This invention relates to the field of power distribution networks, and more particularly to a secure communication system based on the EPDT-RSMA wireless private network for power distribution. Background Technology

[0002] With the large-scale grid connection of distributed renewable energy sources, power distribution systems are exhibiting highly distributed and dynamic characteristics. Numerous intelligent sensors, fault monitoring units, and other distribution terminals need to report operational status data to the master station in real time. The real-time nature of this data directly affects the safe and stable operation of the power grid. For example, critical information such as fault currents needs to be transmitted to the master station within an end-to-end delay of ≤20ms to support rapid protection and control decisions and prevent fault propagation. Timely updates of dynamic information such as load fluctuations and changes in renewable energy output are also prerequisites for precise dispatching. Therefore, the communication link between distribution terminals and the master station must meet stringent low-latency requirements, creating an urgent need for short-packet communication technology.

[0003] Secondly, the current power system primarily relies on the 230MHz frequency band for data transmission. As a dedicated communication band for the power sector, this band features strong diffraction capabilities and low propagation loss, making it suitable for scenarios with wide-area coverage of distribution terminals, and particularly well-suited for signal transmission in complex environments such as suburbs and densely populated urban areas. However, the openness of the 230MHz band exposes it to a high risk of eavesdropping: if encrypted control commands and equipment status information transmitted by distribution terminals are illegally intercepted, it could lead to malicious attacks or leakage of power grid operation information, seriously threatening power system security. Therefore, while ensuring real-time performance, it is necessary to simultaneously improve the anti-eavesdropping capabilities of the communication link.

[0004] Traditional communication methods often employ fixed block lengths and static power allocation strategies, which are ill-suited to the real-time and security requirements of short-packet communication scenarios. On one hand, fixed block lengths cannot adjust transmission efficiency dynamically according to channel changes: too short a block length leads to insufficient channel coding gain, increasing the probability of transmission errors; too long a block length introduces additional latency, exceeding the ≤20ms real-time constraint. On the other hand, static power allocation struggles to balance the broadcast reliability of public information with the anti-eavesdropping requirements of private information, easily resulting in insufficient freshness of critical data or excessive security redundancy.

[0005] Against this backdrop, there is an urgent need for a technical solution that can dynamically optimize short packet communication parameters to achieve coordinated protection of "low latency and high security" in the 230MHz frequency band, supporting the efficient operation of the power system's "sensing-communication-computing-control" closed loop. Summary of the Invention

[0006] This invention aims to address the technical challenges of short-packet transmission in communication between distributed distribution terminals and the master station, particularly in terms of real-time performance, security, and dynamic adaptability. With the digital transformation of power systems, numerous distribution terminals need to report status data with end-to-end delays of ≤20ms. However, the openness of the 230MHz frequency band exposes encrypted control commands and other confidential information to eavesdropping risks. Traditional methods employing fixed block lengths and static power allocation are ill-suited for short-packet scenarios: fixed block lengths either exceed delay constraints due to excessive length or result in insufficient transmission reliability due to excessive shortness; static power allocation fails to balance the broadcast requirements of public information with the anti-eavesdropping requirements of private information and lacks differentiated consideration of the criticality of different data.

[0007] To address this, this invention proposes a secure communication system based on the EPDT-RSMA (Power Distribution Wireless Private Network). Specifically, it is a dynamic optimization method for the age of short packet communication security information in the 230MHz band of the power system based on RSMA. By using RSMA, short packets are split into a common stream (carrying shared information) and a private stream (transmitting encrypted instructions). For the first time, a dual index is introduced: Weighted Average Information Age (W-AOI) and Security Information Age Discontinuation Probability (SAOP). W-AOI is weighted according to data criticality to quantify information freshness; SAOP constrains the risk of exceeding the timeliness limit for eavesdropped information. A dual-objective optimization model is constructed, with minimizing W-AOI as the core and constraining SAOP ≤ 10. -3 Boundary conditions such as rate and block length are considered. A momentum-diffusion coupled learning algorithm is designed, taking environmental conditions such as channel gain, noise power, and eavesdropping coefficient as input. Initial parameters are predicted through a two-layer LSTM, and the algorithm combines a diffusion model with noise addition and denoising to find the optimal value. The algorithm outputs the optimal power allocation and block length within 1000 iterations. A real-time closed-loop control mechanism is established, monitoring indicators every second. If the threshold is exceeded, re-optimization is triggered to dynamically balance latency and security, adapting to dynamic power grid operation scenarios.

[0008] The embodiments of the present invention provide the following technical solutions:

[0009] S1: Establish an RSMA-FBL downlink in the 230MHz band: the base station is equipped with M antennas, and the K power distribution terminals are single antennas. The short packet length is ≤200 symbols, the symbol rate is 19.2 ksym / s, and the bandwidth is 25kHz. This link is adapted to the real-time reporting requirements of the power system, ensuring basic coverage and real-time performance of data transmission.

[0010] S2: Short packet data streams are segmented using Rate Division Multiple Access (RSMA) technology: the common stream carries three types of shared information: network-wide time synchronization code, system status identifier, and spectrum resource allocation instructions; the private stream transmits encrypted control instructions for each terminal. The combined rate of both is limited by channel capacity, and the transmission performance loss caused by finite block length must be considered to ensure that the rate allocation is adapted to the characteristics of short packet communication and to meet the differentiated data transmission needs of terminals.

[0011] S3: Establish a dual-objective optimization model: on the one hand, minimize the weighted average information age W-AOI, assigning different weights to the criticality of three types of data: temperature, voltage, and current; on the other hand, constrain the probability of information age interruption SAOP to not exceed a preset threshold, so as to achieve the coordinated protection of information freshness and transmission security, and adapt to the dual requirements of power system for data real-time and confidentiality.

[0012] S4: Proposes a momentum-diffusion coupled learning algorithm: taking three environmental states—channel gain, noise power, and eavesdropping coefficient—as input, and outputting three decision parameters—common flow power, private flow power, and short packet length. Initial values ​​are predicted using a two-layer Long Short-Term Memory (LSTM) network, noise is added to the diffusion model, and noise is removed along the gradient for optimization. Convergence is achieved after 1000 iterations, efficiently obtaining the optimal solution.

[0013] S5: Construct a real-time closed-loop control mechanism: Monitor W-AOI and SAOP every second; if either indicator exceeds the threshold, algorithm re-optimization is triggered. The updated decision parameters are broadcast to the terminal via RSMA downlink, achieving dynamic balancing end-to-end latency <20ms and SAOP <10-3, adapting to dynamic power system operation scenarios.

[0014] The present invention discloses the following technical effects:

[0015] By deeply integrating RSMA technology with dynamic optimization of security information age, this invention solves the challenge of balancing real-time performance and security in short-packet communication in the 230MHz band of power systems, providing innovative technical support for reliable communication between distributed distribution terminals and master stations. This invention introduces two indicators: Weighted Average Information Age (W-AOI) and Security Information Age Interruption Probability (SAOP). The former measures information freshness based on data key metrics, while the latter constrains the risk of exceeding timeliness limits for eavesdropping information, accurately balancing the transmission needs of different types of data. A momentum-diffusion coupled learning algorithm is designed, using environmental parameters such as channel state and noise power as input. Initial parameters are predicted through a two-layer LSTM, and combined with a diffusion model for efficient optimization, outputting the optimal power allocation and block length within 1000 iterations, rapidly adapting to dynamic changes in the power grid. A real-time closed-loop control mechanism ensures immediate re-optimization when indicators exceed thresholds, dynamically balancing end-to-end latency <20ms and SAOP <10. -3 Ultimately, this enables the system to ensure real-time updates of critical data and resist the risk of eavesdropping in wide-area coverage scenarios of power distribution terminals, stably exerting the closed-loop efficiency of "sensing-communication-computing-control" in the dynamic operation of the power grid. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a secure communication system based on the power distribution wireless private network EPDT-RSMA according to the present invention. Detailed Implementation

[0018] As described in the background section, a key problem urgently needing to be solved in this field is how to construct a mechanism capable of real-time status updates and dynamic optimization based on environmental changes for information transmission between distributed distribution terminals and the master station in the context of short-packet communication in the 230MHz band of power systems. Especially under conditions of dynamic power grid operation such as channel fading, terminal status changes, and security risks such as 230MHz band eavesdropping, traditional single indicators cannot comprehensively measure information freshness and transmission security, and traditional static power allocation and fixed block length strategies cannot meet the needs of flexible system adjustments.

[0019] The core idea of ​​this invention is to construct a closed loop of "sensing-communication-computation-control" in the context of short packet communication in the 230MHz band of power systems. RSMA technology is used to split short packets into public and private flows, and two indices are created: W-AOI (Wireless Automated Opportunity for Information) and the Security Information Age Discontinuation Probability (SAOP). The former measures information freshness based on data criticality, while the latter constrains the risk of exceeding the timeliness limit for eavesdropped information. Through collaborative optimization using a momentum-diffusion coupled learning algorithm, it adapts to dynamic changes in the environment such as channel gain and noise power, while ensuring SAOP < 10. -3 Under the premise of minimizing W-AOI, a short packet communication mechanism for power systems with low latency, high security, high reliability and effective response to dynamic power grid scenarios was constructed.

[0020] See Figure 1 This invention provides a secure communication system based on the power distribution wireless private network EPDT-RSMA, the system comprising:

[0021] S1: Establish an RSMA-FBL downlink in the 230MHz band: the base station is equipped with M antennas, and the K power distribution terminals are single antennas. The short packet length is ≤200 symbols, the symbol rate is 19.2 ksym / s, and the bandwidth is 25kHz. This link is adapted to the real-time reporting requirements of the power system, ensuring basic coverage and real-time performance of data transmission.

[0022] S2: Short packet data streams are segmented using Rate Division Multiple Access (RSMA) technology: the common stream carries three types of shared information: network-wide time synchronization code, system status identifier, and spectrum resource allocation instructions; the private stream transmits encrypted control instructions for each terminal. The combined rate of both is limited by channel capacity, and the transmission performance loss caused by finite block length must be considered to ensure that the rate allocation is adapted to the characteristics of short packet communication and to meet the differentiated data transmission needs of terminals.

[0023] S3: Establish a dual-objective optimization model: on the one hand, minimize the weighted average information age W-AOI, assigning different weights to the criticality of three types of data: temperature, voltage, and current; on the other hand, constrain the probability of information age interruption SAOP to not exceed a preset threshold, so as to achieve the coordinated protection of information freshness and transmission security, and adapt to the dual requirements of power system for data real-time and confidentiality.

[0024] S4: Proposes a momentum-diffusion coupled learning algorithm: taking three environmental states—channel gain, noise power, and eavesdropping coefficient—as input, and outputting three decision parameters—common flow power, private flow power, and short packet length. Initial values ​​are predicted using a two-layer Long Short-Term Memory (LSTM) network, noise is added to the diffusion model, and noise is removed along the gradient for optimization. Convergence is achieved after 1000 iterations, efficiently obtaining the optimal solution.

[0025] S5: Construct a real-time closed-loop control mechanism: Monitor W-AOI and SAOP every second; if either indicator exceeds the threshold, algorithm re-optimization is triggered. The updated decision parameters are broadcast to the terminal via RSMA downlink, achieving dynamic balancing end-to-end latency <20ms and SAOP <10-3, adapting to dynamic power system operation scenarios.

[0026] The establishment of the RSMA-FBL downlink in the 230MHz band further includes:

[0027] First, a communication link adapted to the power system needs to be designed based on the characteristics of the 230MHz frequency band. As a dedicated communication band for power systems, the 230MHz band features strong diffraction capability and low propagation loss, making it suitable for wide-area coverage scenarios of distribution terminals such as smart sensors and fault monitoring units. This band is less affected by terrain and building obstructions in the distribution network environment, and can meet the real-time data transmission needs of distributed terminals, such as status monitoring and control command interaction. To achieve coordination between "sensing-communication-computing-control," the link design must simultaneously ensure low latency (adapting to real-time control) and high reliability (meeting the data integrity requirements of power services). Therefore, RSMA technology and a finite block length (FBL) transmission mechanism are adopted. The finite block length refers to controlling the short packet length within a range suitable for real-time performance, typically not exceeding several hundred symbols, to reduce transmission latency.

[0028] Secondly, the system model and signal transmission architecture of the link are constructed. The base station is configured with M transmit antennas to achieve spatial diversity and improve the receiving performance of the edge terminals; the K power distribution terminals all adopt a single antenna design.

[0029] Finally, a channel model is established to quantify transmission loss. The channel characteristics of the 230MHz band need to comprehensively consider both large-scale and small-scale fading:

[0030] Large-scale fading: using the path loss model L(d k )=L(d0)·(d k / d0) -α , where d k Let d0 be the distance between the base station and the kth terminal, L(d0) be the reference distance, L(d0) be the path loss at the reference distance, and α be the path loss exponent, which is determined by the distribution network environment. For example, α is smaller in suburban scenarios and larger in dense urban areas.

[0031] Small-scale fading: Due to scattering objects such as utility poles and buildings in the distribution network, the channel exhibits Ricean fading characteristics. The complex channel gain of the k-th terminal is... Where K is the Rice factor (reflecting the proportion of the line-of-sight component), Let I be the deterministic gain of the line-of-sight component, and let I be the identity matrix.

[0032] Considering both large-scale and small-scale fading, the composite channel gain of the k-th terminal is:

[0033]

[0034] The method of segmenting short packet data streams using RSMA technology further includes:

[0035] Firstly, the core value of RSMA technology in short-packet communication in the 230MHz band of power systems lies in its ability to adapt to the differentiated data transmission needs of distribution terminals through a two-layer architecture of "public flow + private flow". In power systems, distribution terminals need to transmit two types of information: one is public information shared across the entire network, such as time synchronization codes, system status identifiers, and spectrum resource allocation instructions. This type of information needs to be received by all terminals, requiring extremely high reliability but with moderate sensitivity to real-time performance. The other type is private information specific to each terminal, such as voltage / current sampling values, equipment status alarms, and encrypted control instructions. This type of information only needs to be received by specific terminals, with stringent requirements for real-time performance (latency <20ms) and security (anti-eavesdropping). RSMA enables non-orthogonal multiplexing of these two types of information, improving spectrum efficiency under finite block length constraints, while simultaneously balancing the conflict between "public information broadcasting" and "private information dedicated transmission" through dynamic allocation of power and rate.

[0036] Secondly, a mathematical model of the RSMA short packet data stream is constructed. The baseband expression for the base station's transmitted signal is:

[0037]

[0038] Among them, P c For common current power, P k The private flow power of the k-th terminal satisfies the power constraint. P max The maximum transmit power specified for the 230MHz band; Shape vectors for the common stream beam. The private stream beamforming vector for the k-th terminal all satisfy the normalization condition ||w||. c || 2 =1、‖w k || 2 =1; s c (t) represents the common flow information symbol, using BPSK modulation because of its strong anti-interference capability and suitability for the high reliability requirements of common information. k (t) represents the private stream information symbol of the k-th terminal. QPSK or 16QAM modulation is used to balance the rate and bit error rate according to the rate requirement.

[0039] Furthermore, the rate expression under finite block length is derived. In short packet communication, the channel coding gain is limited, and the actual achievable rate needs to be corrected by introducing a finite block length. For a common stream, its rate R... c (Unit: bit / symbol) is:

[0040]

[0041] in:

[0042]

[0043] γ c For the k-th terminal to receive the SINR of the common stream, Let P be the channel gain from the base station to the k-th terminal. j Let K be the transmit power of the private stream of the j-th terminal (j = 1, 2, ..., K). The beamforming vector for the j-th terminal's private stream. The first term in the denominator is the co-channel interference of all private streams, and the second term σ... 2 The noise power in the 230MHz band; V(γ) c )=1-(1+γ c ) -2 ε is the common flow channel dispersion coefficient, reflecting the impact of channel randomness on the rate; N is the block length; ε c The decoding error probability of the common stream must be ≤10. -4To ensure high reliability of information synchronization across the entire network, Q -1 (·) is the inverse function of the Gaussian Q-function.

[0044] For a private stream, the rate R of the k-th terminal k for:

[0045]

[0046] in:

[0047]

[0048] γ k For the SINR of the private stream received by the k-th terminal, the denominator includes interference and noise from other private streams; V(γ) k )=1-(1+γ k ) -2 ε represents the dispersion coefficient of the private stream channel; k The decoding error probability of the private stream needs to be ≤10-3 to balance real-time performance and reliability.

[0049] Rate allocation must satisfy the summation constraint: C max This is the upper limit of channel capacity for the 230MHz band.

[0050] Finally, a rate segmentation strategy adapted to power services is designed. Common flow rate R c A minimum threshold should be set based on the minimum requirement for shared information, and remaining rate resources should be allocated to private streams according to terminal priority. Priority is determined by a three-dimensional indicator: "Information Criticality - Freshness - Security".

[0051] Criticality: Fault alarm information such as short-circuit current is given the highest priority and allocated more data rate;

[0052] Freshness: W-AOI quantization is adopted, and terminals with high real-time requirements, such as load control terminals, are allocated higher rates to reduce AOI;

[0053] Security: SAOP constraints are adopted, and private information such as encrypted control commands must be protected by increasing the private stream power P. k Reduce the probability of being eavesdropped on.

[0054] By dynamically adjusting P c P k R c R k RSMA technology can achieve "wide coverage of public information, high security of private information, and low latency of critical information" with limited block length, providing underlying support for the "sensing-communication-computing-control" closed loop of the 230MHz frequency band of the power system.

[0055] The establishment of the dual-objective optimization model further includes:

[0056] First, we define the weighted average information age W-AOI as an indicator of information freshness. In power systems, the real-time requirements for temperature, voltage, and current data differ. For example, fault current data requires millisecond-level freshness, while ambient temperature data can tolerate second-level delays. Therefore, we introduce a weighting coefficient to quantify criticality. The instantaneous information age of the k-th distribution terminal is Δ. k (t) represents the elapsed time since the terminal last successfully received an update. The system weighted average information age is defined as:

[0057]

[0058] Where, ω k Let be the weight of the k-th terminal, satisfying Fault monitoring terminal takes ω k ∈[0.6,0.8], the environmental monitoring terminal takes ω k ∈[0.2,0.4]; Let K be the expected value; K be the total number of terminals. This formula prioritizes the freshness of key data through weighted allocation.

[0059] Secondly, the probability of information age expiration (SAOP) is defined as a security constraint. SAOP describes the probability that the eavesdropper's information age ΔE(k) is not behind that of the legitimate terminal ΔD(k), i.e., the risk that the eavesdropped information is not outdated. The instantaneous security information age is ΔS(k) = [ΔE(k) - ΔD(k)]. + , [·] + If we denote the value to be non-negative, then SAOP is defined as:

[0060]

[0061] Where, μ th denoted as the target information delay threshold; N is the short packet block length. This formula quantifies the risk of exceeding the timeliness limit for eavesdropped information.

[0062] Furthermore, the mathematical expression for the dual-objective optimization model is constructed. The core objective is to minimize the W-AOI, while simultaneously constraining SAOP to not exceed a preset threshold ε. th =10 -3 The model is as follows:

[0063]

[0064] &0 <N≤200

[0065] Where, θ={P c ,P k Let {P, N} be the optimization decision parameter vector, where Pc For common current power, P k =[P1,P2,...,P K [] represents the private flow power vector, N is the short packet length, and R is the short packet length. c R k These represent the rates of the public flow and the private flow, respectively; C max The upper limit of the channel capacity in the 230MHz band is determined by the bandwidth of 25kHz and the symbol rate of 19.2ksym / s; the block length constraint N≤200 is adapted to the low latency requirements of short packet communication.

[0066] Finally, the cooperative mechanism of the model is reflected in: power allocation through RSMA technology, P c With P k To balance the reliability of public stream broadcasts and the security of private streams, the block length N is adjusted to achieve dual protection of "low W-AOI for high-priority data + low SAOP for all data".

[0067] The proposed momentum-diffusion coupling learning algorithm further includes:

[0068] First, the overall algorithm framework design. The momentum-diffusion coupled learning algorithm aims to achieve dynamic optimization of decision parameters in short packet communication in the 230MHz band of power systems through a three-order process of "prediction-optimization-iteration". Its core input is the real-time environmental state vector s={h,σ 2 ,ξ}, where h=[h1,h2,...,h K ] represents the channel gain vector of each terminal, σ 2 Let ξ be the noise power in the 230MHz band, and ξ be the eavesdropping coefficient, reflecting the ratio of the eavesdropper's channel quality to that of the legitimate terminal, where ξ ∈ [0,1]. The output is the optimization decision vector θ = {P} c ,P k ,N}, where P c For common current power, P k =[P1,P2,...,P K [] represents the private flow power vector, and N is the short packet length. The algorithm uses a two-layer LSTM to predict initial parameters, and combines noise perturbation of the diffusion model with gradient descent to quickly converge to the optimal solution within 1000 iterations.

[0069] Secondly, the two-layer LSTM initial prediction module. This module adopts a two-level architecture of "feature extraction-parameter mapping": the first layer is an LSTM with an input dimension of 3×K and a hidden layer dimension of 64. It performs spatiotemporal feature fusion on the environmental state and outputs a high-dimensional feature vector. The second LSTM layer, with an input dimension of 64 and an output dimension of 2K+2, maps the features to the initial decision parameters θ. init ={P c,init ,Pk,init N init}. Among them, P c,init This indicates the initial transmit power of the public stream.

[0070] P k,init =[P 1,init ,P 2,init ,…,P K,init ]

[0071] Where P k,init The initial transmit power corresponding to the k-th terminal's private stream, N init The initial block length of the short data packet is predicted using the following formula:

[0072] θ init =LSTM2(LSTM1(s))

[0073] In LSTM1, the activation function is ReLU, used to capture the nonlinear relationship between channel gain and noise; the output layer of LSTM2 uses the Sigmoid function to constrain the power parameter to [0, P]. max ], where P max The maximum transmit power for the 230MHz band, with an initial block length N. init The linear mapping constraint is applied within the symbol range of [50, 200].

[0074] Furthermore, noise perturbation and gradient optimization in the diffusion model. The diffusion model achieves fine-grained parameter optimization through "stepwise noise addition-gradient denoising":

[0075] Noise addition: for initial parameters θ init Gaussian noise is added according to the iteration step size t (t=1,2,...,1000), and the parameters after perturbation are:

[0076]

[0077] Where, α t This is the noise attenuation coefficient. Let θ be a Gaussian noise vector. t The decision parameter vector for the t-th iteration ensures that the parameters are fully explored within the search space.

[0078] Gradient denoising: Noise removal is achieved by calculating the gradient using the Lagrangian function of a dual-objective optimization model. The Lagrangian function is:

[0079]

[0080] Where λ is the penalty coefficient. The gradient is calculated as follows:

[0081]

[0082] in This is an indicator function; it takes the value 1 if the condition is met, and 0 otherwise. The momentum term accelerates gradient descent.

[0083]

[0084] θ t =θ t-1 -v t

[0085] Where γ is the momentum factor, η is the learning rate, and v t This is the momentum update vector at the t-th iteration, to avoid the optimization getting trapped in local optima.

[0086] Finally, the iterative convergence mechanism. The algorithm is set to an upper limit of 1000 iterations, and the optimization effect is judged by the following metrics in each iteration:

[0087] Decrease in objective function: when Early convergence;

[0088] Constraint satisfaction: Check SAOP≤ε th With power constraint P c +∑P k ≤P max .

[0089] The final output optimized parameters need to be converted into physical layer signal parameters through the RSMA modulation module: power allocation factor α c =P c / P max α k =P k / P max The block length N is directly used for short packet frame structure generation. The algorithm accelerates convergence through momentum terms and enhances search diversity through a diffusion model. It can still maintain a fast response within 1000 iterations in dynamic distribution network scenarios, meeting the real-time requirement of "control commands <20ms" for power systems.

[0090] The aforementioned construction of a real-time closed-loop control mechanism further includes:

[0091] First, the design of the real-time monitoring module is the foundation of closed-loop control. This module needs to collect operating data from K distribution terminals across the entire network every second, calculate W-AOI and SAOP, and compare them with preset thresholds. The real-time calculation of W-AOI uses the sliding window method, as shown in the following formula:

[0092]

[0093] Where t is the current time; ω k Δ is the weight coefficient for the k-th terminal; k(t) represents the instantaneous information age of the k-th terminal, defined as the elapsed time since the last successful reception of an update, calculated as follows:

[0094] Δ k (t)=tt′ k

[0095] Where t′ k This represents the moment when the k-th terminal most recently successfully received an update. The real-time calculation of SAOP is based on random network calculus theory, deriving its instantaneous value using real-time channel parameters.

[0096]

[0097] Among them, ΔS(k)=[ΔE(k)-ΔD(k)] + Let ΔE(k) be the age of the instantaneous security information, ΔD(k) be the age of the eavesdropper's information, and μ be the age of the legitimate terminal's information. th t is the target information lag threshold; N(t) is the current short packet length.

[0098] Secondly, the threshold triggering mechanism needs to balance real-time performance and stability. When W-AOI(t) > 0.02 seconds or SAOP(t) > 10 is detected... -3 Upon failure, the re-optimization process is immediately triggered. To avoid frequent triggering leading to system oscillations, a hysteresis threshold is set: algorithm re-optimization is only initiated when the metric exceeds the threshold for three consecutive sampling periods. Simultaneously, a channel quality factor γ is introduced. avg (t) Auxiliary decision-making, the formula is:

[0099]

[0100] Among them, P k (t) represents the private current power of the k-th terminal; h k (t) represents the channel gain from the base station to the k-th terminal; h k,j (t) represents the interference channel gain of the j-th terminal to the k-th terminal; σ 2 Noise power. When γ avg When (t) < -5dB, preventative optimization is triggered even if the metric does not exceed the threshold.

[0101] Furthermore, the re-optimization process employs a fast iterative version of the momentum-diffusion coupled learning algorithm, introducing a dynamic learning rate η(t) during the iteration process:

[0102]

[0103] Where η0 is the initial learning rate; T ηThis is a decay constant to ensure convergence stability in the later stages of the iteration. Simultaneously, to meet the constraint of end-to-end latency <20ms, the algorithm needs to embed latency checks during iterations:

[0104]

[0105] Among them, R sym =19.2ksym / s is the symbol rate for the 230MHz band; T proc For base station latency processing; T prop =d k / c represents the propagation delay, d k Let c be the distance from the base station to the k-th terminal, and c = 3 × 10. 8 m / s is the speed of light.

[0106] Finally, parameter updates and broadcast mechanisms ensure network-wide synchronization. The optimized decision parameters are broadcast to all terminals via RSMA downlink, employing a two-layer frame structure of "common stream + private stream": the common stream carries network-wide parameters, using BPSK modulation and repetitive coding to ensure reliability; the private stream allocates a dedicated time slot to each terminal to transmit personalized power parameters P. k (t), QPSK modulation is used to improve the data rate. To avoid parameter update conflicts, a timestamp mechanism is introduced:

[0107] Timestamp = t opt +Δt sync

[0108] Among them, t opt The time when the algorithm completes optimization; Δt sync To ensure synchronization and buffering time, all terminals enable the new parameters simultaneously. After broadcasting, the real-time monitoring module immediately enters the next round of sampling, forming a closed loop of "monitoring-decision-execution-feedback" to dynamically balance information freshness and security. Through this mechanism, the system can continuously meet W-AOI < 0.02 seconds and SAOP < 10 under scenarios such as distribution network load fluctuations and channel abrupt changes. -3 The constraints ensure reliable communication in the 230MHz band of the power system.

[0109] Based on the above system, it can be seen that this application solves the problem of balancing real-time performance and security in short-packet communication in the 230MHz band of power systems by deeply integrating RSMA technology with dynamic optimization of security information age, providing innovative technical support for reliable communication between distributed distribution terminals and master stations. This invention introduces two indicators: Weighted Average Information Age (W-AOI) and Security Information Age Interruption Probability (SAOP). The former measures information freshness based on data key metrics, while the latter constrains the risk of exceeding the timeliness limit for eavesdropping information, accurately balancing the transmission needs of different types of data. A momentum-diffusion coupled learning algorithm is designed, using environmental parameters such as channel state and noise power as input. Initial parameters are predicted through a two-layer LSTM, and combined with a diffusion model for efficient optimization, outputting the optimal power allocation and block length within 1000 iterations, quickly adapting to dynamic changes in the power grid. A real-time closed-loop control mechanism ensures immediate re-optimization when indicators exceed thresholds, dynamically balancing end-to-end latency <20ms and SAOP <10. -3 Ultimately, this enables the system to ensure real-time updates of critical data and resist the risk of eavesdropping in wide-area coverage scenarios of power distribution terminals, stably exerting the closed-loop efficiency of "sensing-communication-computing-control" in the dynamic operation of the power grid.

[0110] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A secure communication system based on the power distribution wireless private network EPDT-RSMA, characterized in that, Includes the following steps: S1: Construct an RSMA-FBL downlink in the 230MHz band: The base station is equipped with M antennas, the K power distribution terminals are single antennas, the short packet length is ≤200 symbols, the symbol rate is 19.2ksym / s, and the bandwidth is 25kHz. S2: The short packet data stream is segmented using Rate Division Multiple Access (RSMA) technology: the common stream carries three types of shared information: the network-wide time synchronization code, the system status identifier, and the spectrum resource allocation instruction. The private stream transmits the encrypted control instructions of each terminal. The sum of the rates of the two is limited by the channel capacity. The transmission performance loss caused by the finite block length must be considered to ensure that the rate allocation is adapted to the characteristics of short packet communication and to meet the differentiated data transmission needs of the terminals. S3: Establish a dual-objective optimization model: on the one hand, minimize the weighted average information age W-AOI, and assign different weights to the criticality of the three types of data: temperature, voltage, and current; on the other hand, constrain the probability of information age interruption SAOP to not exceed a preset threshold, so as to achieve the coordinated guarantee of information freshness and transmission security, and adapt to the dual requirements of power system for data real-time and confidentiality. S4: Propose a momentum-diffusion coupled learning algorithm: take three environmental states, namely channel gain, noise power and eavesdropping coefficient, as inputs, and output three decision parameters, namely public flow power, private flow power and short packet block length. Predict the initial value through a two-layer long short-term memory network LSTM, add noise to the diffusion model and denoise along the gradient to find the best solution. Iterate 1000 times to converge and efficiently obtain the optimal solution. S5: Construct a real-time closed-loop control mechanism: Monitor W-AOI and SAOP every second. If any indicator exceeds the threshold, the algorithm will be re-optimized. The updated decision parameters will be broadcast to the terminal via RSMA downlink. The dynamic balance end-to-end delay is <20ms and SAOP is <10-3, which can adapt to the dynamic operation scenario of the power system.

2. The EPDT-RSMA secure communication system based on a power distribution wireless private network according to claim 1, wherein establishing an RSMA-FBL downlink in the 230MHz frequency band further includes: First, a communication link adapted to the power system is designed based on the characteristics of the 230MHz frequency band. RSMA technology and finite block length FBL transmission mechanism are adopted. The finite block length refers to the short packet length being controlled within the range of adaptability to real-time requirements, usually not exceeding several hundred symbols, in order to reduce transmission delay. Secondly, the system model and signal transmission architecture of the link are constructed: the base station is configured with M transmit antennas to achieve spatial diversity and improve the receiving performance of the edge terminal; all K power distribution terminals adopt a single antenna design. Finally, a channel model is established to quantify transmission loss. The channel characteristics of the 230MHz band need to comprehensively consider both large-scale and small-scale fading. Large-scale fading: using the path loss model L(d k )=L(d0)·(d k / d0) -α , where d k Let d0 be the distance between the base station and the kth terminal, L(d0) be the reference distance, L(d0) be the path loss at the reference distance, and α be the path loss exponent, which is determined by the distribution network environment. For example, α is smaller in suburban scenarios and larger in dense urban areas. Small-scale fading: Due to scattering objects such as utility poles and buildings in the distribution network, the channel exhibits Ricean fading characteristics. The complex channel gain of the k-th terminal is... Where K is the Rice factor (reflecting the proportion of the line-of-sight component), Let I be the deterministic gain of the line-of-sight component, and let I be the identity matrix. Considering both large-scale and small-scale fading, the composite channel gain of the k-th terminal is:

3. The EPDT-RSMA secure communication system based on a power distribution wireless private network according to claim 1, wherein the segmentation of short packet data streams using RSMA technology further includes: First, the core value of RSMA technology in short packet communication in the 230MHz band of power systems lies in its ability to adapt to the differentiated data transmission needs of distribution terminals through a two-layer architecture of "public flow + private flow". In power systems, distribution terminals need to transmit two types of information: one is public information shared by the entire network, which needs to be received by all terminals and has extremely high reliability requirements but moderate sensitivity to real-time performance; the other is private information exclusive to each terminal, which only needs to be received by specific terminals and has stringent requirements for real-time performance and security. RSMA can achieve non-orthogonal multiplexing of the two types of information, improve spectrum efficiency under the constraint of finite block length, and balance the contradiction between "public information broadcasting" and "private information dedicated transmission" through dynamic allocation of power and rate. Secondly, a mathematical model of the RSMA short packet data stream is constructed, and the baseband expression of the base station transmitted signal is as follows: Among them, P c For common current power, P k The private flow power of the k-th terminal satisfies the power constraint. P max The maximum transmit power specified for the 230MHz band; Shape vectors for the common stream beam. The private stream beamforming vector for the k-th terminal all satisfy the normalization condition ||w||. c || 2 =1、‖W k || 2 =1; s c (t) represents the common flow information symbol, using BPSK modulation because of its strong anti-interference capability and suitability for the high reliability requirements of common information. k (t) represents the private stream information symbol of the kth terminal. QPSK or 16QAM modulation is used to balance the rate and bit error rate according to the rate requirement. Furthermore, deriving the rate expression under finite block length is crucial. In short packet communication, the channel coding gain is limited, and the actual achievable rate requires a finite block length correction. For a common stream, its rate R... c (Unit: bit / symbol) is: in: γ c For the k-th terminal to receive the SINR of the common stream, Let P be the channel gain from the base station to the k-th terminal. j Let K be the transmit power of the private stream of the j-th terminal (j = 1, 2, ..., K). The beamforming vector for the j-th terminal's private stream, with the first term in the denominator being the co-channel interference of all private streams, and the second term being σ. 2 The noise power in the 230MHz band; V(γ) c )=1-(1+γ c ) -2 ε is the common flow channel dispersion coefficient, reflecting the impact of channel randomness on the rate; N is the block length; ε c The decoding error probability of the common stream must be ≤10. -4 To ensure high reliability of information synchronization across the entire network, Q -1 (·) is the inverse function of the Gaussian Q-function; For a private stream, the rate R of the k-th terminal k for: in: γ k For the SINR of the private stream received by the k-th terminal, the denominator includes interference and noise from other private streams; V(γ) k )=1-(1+γ k ) -2 ε represents the dispersion coefficient of the private stream channel; k The decoding error probability of the private stream needs to be ≤10-3, balancing real-time performance and reliability; Rate allocation must satisfy the summation constraint: C max This is the upper limit of channel capacity for the 230MHz band; Finally, a rate segmentation strategy adapted to power services is designed, with a common flow rate R. c A minimum threshold needs to be set based on the minimum requirements for shared information. Remaining rate resources should be allocated to private streams according to terminal priority, which is determined by a three-dimensional index of "information criticality - freshness - security". Criticality: Fault alarm information such as short-circuit current is given the highest priority and allocated more data rate; Freshness: W-AOI quantization is adopted, and terminals with high real-time requirements, such as load control terminals, are allocated higher rates to reduce AOI; Security: SAOP constraints are adopted, and private information such as encrypted control commands must be protected by increasing the private stream power P. k Reduce the probability of being eavesdropped on; By dynamically adjusting P c P k R c R k RSMA technology can achieve "wide coverage of public information, high security of private information, and low latency of critical information" under limited block length, providing underlying support for the "sensing-communication-computing-control" closed loop of the 230MHz band of the power system.

4. The EPDT-RSMA secure communication system based on a power distribution wireless private network according to claim 1, wherein establishing the dual-objective optimization model further includes: First, the weighted average information age W-AOI is defined as an indicator of information freshness. In the power system, the real-time requirements of temperature, voltage, and current data differ. Therefore, a weighting coefficient is introduced to quantify criticality. The instantaneous information age of the k-th distribution terminal is Δ. k (t) represents the elapsed time since the terminal last successfully received an update. The system weighted average information age is defined as: Where, ω k Let be the weight of the k-th terminal, satisfying Fault monitoring terminal takes ω k ∈[0.6,0.8], the environmental monitoring terminal takes ω k ∈[0.2,0.4]; The expected value is K; K is the total number of terminals. This formula ensures the priority and freshness of key data through weight allocation. Secondly, the probability of information age expiration (SAOP) is defined as a security constraint. SAOP describes the probability that the eavesdropper's information age ΔE(k) is not behind that of the legitimate terminal ΔD(k), i.e., the risk that the eavesdropped information is not outdated. The instantaneous information age is ΔS(k) = [ΔE(k) - ΔD(k)]. + , [·] + If we denote the value to be non-negative, then SAOP is defined as: Where, μ th The formula quantifies the risk of exceeding the timeliness limit of eavesdropping information, where N is the target information lag threshold and N is the short packet length. Furthermore, the mathematical expression for the dual-objective optimization model is constructed, with minimizing W-AOI as the core objective, while constraining SAOP to not exceed a preset threshold ε. th =10 -3 The model is as follows: &0<N≤200 Where, θ={P c ,P k Let {P, N} be the optimization decision parameter vector, where P c For common current power, P k =[P1,P2,...,P K [] represents the private flow power vector, N is the short packet length, and R is the short packet length. c R k These represent the rates of the public flow and the private flow, respectively; C max The upper limit of the channel capacity in the 230MHz band is determined by the bandwidth of 25kHz and the symbol rate of 19.2ksym / s; the block length constraint N≤200 is adapted to the low latency requirements of short packet communication. Finally, the cooperative mechanism of the model is reflected in: power allocation through RSMA technology, P c With P k To balance the reliability of public stream broadcasts and the security of private streams, the block length N is adjusted to achieve dual protection of "high-priority data with low W-AOI + all data with low SAOP".

5. The EPDT-RSMA secure communication system based on a power distribution wireless private network according to claim 1, wherein the proposed momentum-diffusion coupling learning algorithm further includes: First, the overall algorithm framework is designed. The momentum-diffusion coupled learning algorithm aims to achieve dynamic optimization of decision parameters in short packet communication in the 230MHz band of power systems through a three-order process of "prediction-optimization-iteration". Its core input is the real-time environmental state vector s={h,σ 2 ,ξ}, where h=[h1,h2,...,h K ] represents the channel gain vector of each terminal, σ 2 Let ξ be the noise power in the 230MHz band, and ξ be the eavesdropping coefficient, reflecting the ratio of the eavesdropper's channel quality to that of the legitimate terminal, where ξ∈[0,1]. The output is the optimization decision vector θ={P c ,P k ,N}, where P c For common current power, P k =[P1,P2,...,P K ] is the private flow power vector, N is the short packet length. The algorithm uses a two-layer LSTM to predict the initial parameters and combines the noise perturbation of the diffusion model with gradient descent to quickly converge to the optimal solution within 1000 iterations. Secondly, the two-layer LSTM initial prediction module adopts a two-level architecture of "feature extraction-parameter mapping": the first layer LSTM has an input dimension of 3×K and a hidden layer dimension of 64, which performs spatiotemporal feature fusion on the environmental state and outputs a high-dimensional feature vector. The second LSTM layer, with an input dimension of 64 and an output dimension of 2K+2, maps the features to the initial decision parameters θ. init ={P c,init ,P k,init N init }, where P c,init This indicates the initial transmit power of the public stream. P k,init =[P 1,init ,P 2,init ,…,P K,init ] Where P k,init The initial transmit power corresponding to the k-th terminal's private stream, N init The initial block length of the short data packet is predicted using the following formula: θ init =LSTM2(LSTM1(s)) In LSTM1, the activation function is ReLU, used to capture the nonlinear relationship between channel gain and noise; the output layer of LSTM2 uses the Sigmoid function to constrain the power parameter to [0, P]. max ], where P max The maximum transmit power for the 230MHz band, with an initial block length N. init Constrained by linear mapping within the sign range of [50, 200]; Furthermore, regarding noise perturbation and gradient optimization in the diffusion model, the diffusion model achieves refined parameter optimization through "stepwise noise addition-gradient denoising": Noise addition: for initial parameters θ init Gaussian noise is added according to the iteration step size t (t=1,2,...,1000), and the parameters after perturbation are: Where, α t This is the noise attenuation coefficient. Let θ be a Gaussian noise vector. t The decision parameter vector for the t-th iteration ensures that the parameters are fully explored within the search space. Gradient denoising: The gradient is calculated using the Lagrangian function of a dual-objective optimization model to remove noise. The Lagrangian function is: Where λ is the penalty coefficient, and the gradient is calculated as follows: in This is an indicator function that takes the value 1 when the condition is met and 0 otherwise, accelerating gradient descent through the momentum term. i t =θ t-1 -v t Where γ is the momentum factor, η is the learning rate, and c t This is the momentum update vector at the t-th iteration, to avoid the optimization getting trapped in local optima; Finally, the iterative convergence mechanism sets an upper limit of 1000 iterations for the algorithm, and the optimization effect is judged by the following indicators for each iteration: Decrease in objective function: when Early convergence; Constraint satisfaction: Check SAOP≤ε th With power constraint P c +∑P k ≤P max ; The final output optimized parameters need to be converted into physical layer signal parameters through the RSMA modulation module: power allocation factor α c =P c / P max α k =P k / P max The block length N is directly used for short packet structure generation. The algorithm accelerates convergence through momentum terms and enhances search diversity through diffusion models. It can still maintain a fast response within 1000 iterations in dynamic distribution network scenarios, meeting the real-time requirement of "control commands <20ms" for power systems.

6. The EPDT-RSMA secure communication system based on a power distribution wireless private network according to claim 1, wherein the construction of a real-time closed-loop control mechanism further includes: First, the design of the real-time monitoring module is the foundation of closed-loop control. This module needs to collect operating data from K distribution terminals across the entire network every second, calculate W-AOI and SAOP, and compare them with preset thresholds. The real-time calculation of W-AOI uses the sliding window method, and the formula is as follows: Where t is the current time; ω k Δ is the weight coefficient for the k-th terminal; k (t) represents the instantaneous information age of the k-th terminal, defined as the elapsed time since the last successful reception of an update, calculated as follows: Δ k (t)=t-t′ k Where t′ k The real-time calculation of SAOP is based on the theory of random network calculus, using real-time channel parameters to derive its instantaneous value, which is the moment when the k-th terminal most recently successfully received an update. Among them, ΔS(k)=[ΔE(k)-ΔD(k)] + Let ΔE(k) be the age of the instantaneous security information, ΔD(k) be the age of the eavesdropper's information, and μ be the age of the legitimate terminal's information. th The target information lag threshold is N(t); N(t) is the current short packet length. Secondly, the threshold triggering mechanism needs to balance real-time performance and stability. When W-AOI(t) > 0.02 seconds or SAOP(t) > 10 is detected... -3 Upon failure, the re-optimization process is immediately triggered. To avoid frequent triggering leading to system oscillation, a hysteresis threshold is set: algorithm re-optimization is only initiated when the indicator exceeds the threshold for three consecutive sampling periods. Simultaneously, a channel quality factor γ is introduced. avg (t) Auxiliary decision-making, the formula is: Among them, P k (t) represents the private current power of the k-th terminal; h k (t) represents the channel gain from the base station to the k-th terminal; h k,j (t) represents the interference channel gain of the j-th terminal to the k-th terminal; σ 2 Noise power, when γ avg When (t) < -5dB, preventative optimization is triggered even if the index does not exceed the threshold. Furthermore, the re-optimization process employs a fast iterative version of the momentum-diffusion coupled learning algorithm, introducing a dynamic learning rate η(t) during the iteration process: Where η0 is the initial learning rate; T η To ensure convergence stability in the later stages of iteration, and to meet the constraint of end-to-end latency <20ms, the algorithm needs to embed latency verification during iteration. Among them, R sym =19.2ksym / s is the symbol rate for the 230MHz band; T proc For base station latency processing; T prop =d k / c represents the propagation delay, d k Let c be the distance from the base station to the k-th terminal, and c = 3 × 10. 8 m / s is the speed of light; Finally, the parameter update and broadcast mechanism ensures network-wide synchronization. The optimized decision parameters are broadcast to all terminals via RSMA downlink, employing a two-layer frame structure of "common stream + private stream": the common stream carries network-wide parameters and uses BPSK modulation and repetitive coding to ensure reliability; the private stream allocates a dedicated time slot for each terminal to transmit personalized power parameters P. k (t), QPSK modulation is used to improve the data rate. To avoid parameter update conflicts, a timestamp mechanism is introduced: Timestamp=t opt +Δt sync Among them, t opt The time when the algorithm completes optimization; Δt sync To ensure synchronization and buffer time, all terminals enable the new parameters simultaneously. After broadcasting, the real-time monitoring module immediately enters the next round of sampling, forming a closed loop of "monitoring-decision-execution-feedback". This dynamically balances information freshness and security. Through this mechanism, the system can continuously meet W-AOI < 0.02 seconds and SAOP < 10 under scenarios such as distribution network load fluctuations and channel abrupt changes. -3 The constraints ensure reliable communication in the 230MHz band of the power system.

Images