Power Grid Safety Production Operation Dispatch Methods and Devices

Abstract

This disclosure provides a method and apparatus for scheduling power grid safety production services. It uses a recurrent neural network model to predict the latency of new data packets sent by target user terminals. Based on the latency prediction results, the priorities of target queues are adjusted to obtain the adjusted priorities of each queue. Then, based on the adjusted priorities, a near-end policy optimization pruning algorithm is used to solve the bandwidth resource allocation model to obtain a bandwidth resource allocation strategy. Finally, based on the bandwidth resource allocation strategy, a scheduling method for power grid safety production services is obtained. Scheduling according to this method can effectively reduce overall transmission latency, improve system throughput and transmission rate, and to some extent solve the problems of network congestion and excessively long scheduling times, meeting the transmission requirements of latency-sensitive services.

Description

Technical Field

[0001] This disclosure relates to the field of smart grid dispatching technology, and in particular to a method and apparatus for dispatching business operations for safe power grid production. Background Technology

[0002] Power grid applications are one of the core application areas of 5G (5th Generation Mobile Communication Technology). The continuous development of wireless networks has also brought technical challenges to 5G technology, such as the unpredictable problems in power grid security production service flows. Power grid security production service flows require extremely low transmission latency and high reliability, but existing technologies often suffer from high transmission latency. This directly affects the quality of data transmission, potentially leading to network congestion and excessively long scheduling times, thus failing to meet the transmission requirements of the service flows. Summary of the Invention

[0003] In view of this, the purpose of this disclosure is to propose a method and apparatus for power grid safety production dispatching, in order to solve or partially solve the above-mentioned technical problems.

[0004] Based on the above objectives, the first aspect of this disclosure proposes a method for power grid safety production dispatching, including:

[0005] Data packets sent by multiple user terminals are acquired, and the multiple data packets are integrated into multiple queues according to the device information of the user terminals, wherein the multiple queues have different priorities;

[0006] When at least one target user terminal among multiple user terminals sends a new data packet, the new data packet is added to the target queue corresponding to the target user terminal, wherein the target queue is at least one of the multiple queues;

[0007] The latency prediction result is obtained by using a recurrent neural network model to predict the latency of the new data packet.

[0008] Based on the results of the latency prediction, the priority of the target queue is adjusted to obtain the adjusted priority of each queue;

[0009] Based on the adjusted priority, the pre-built bandwidth resource allocation model is solved using the near-end strategy optimization pruning algorithm to obtain the bandwidth resource allocation strategy.

[0010] Based on the bandwidth resource allocation strategy, the scheduling method for power grid safety production services is obtained.

[0011] Based on the same inventive concept, a second aspect of this disclosure proposes a power grid safety production business scheduling device, comprising: a queue determination module, a data packet determination module, a delay prediction module, a priority determination module, a bandwidth resource allocation strategy determination module, and the scheduling determination module.

[0012] The queue determination module is configured to acquire data packets sent by multiple user terminals, and integrate the multiple data packets into multiple queues according to the device information of the user terminals, wherein the multiple queues have different priorities.

[0013] The data packet determination module is configured to add the new data packet to the target queue corresponding to the target user terminal when at least one target user terminal among a plurality of user terminals sends a new data packet, wherein the target queue is at least one of the plurality of queues;

[0014] The latency prediction module is configured to use a recurrent neural network model to predict the latency of the new data packet and obtain the latency prediction result.

[0015] The priority determination module is configured to adjust the priority of the target queue based on the result of the delay prediction, so as to obtain the adjusted priority of each queue.

[0016] The bandwidth resource allocation strategy determination module is configured to solve the pre-built bandwidth resource allocation model using a near-end strategy optimization pruning algorithm based on the adjusted priority to obtain the bandwidth resource allocation strategy.

[0017] The scheduling determination module is configured to determine the scheduling method for power grid safe production services based on the bandwidth resource allocation strategy.

[0018] As described above, the power grid safety production service scheduling method and apparatus provided in this disclosure predicts the latency of new data packets sent by target user terminals using a recurrent neural network model. Based on the latency prediction results, the priorities of target queues are adjusted to obtain the adjusted priorities of each queue. Then, based on the adjusted priorities, a near-end policy optimization pruning algorithm is used to solve the bandwidth resource allocation model to obtain a bandwidth resource allocation strategy. Finally, based on the bandwidth resource allocation strategy, the scheduling method for power grid safety production services is obtained. Scheduling according to this method can effectively reduce overall transmission latency, improve system throughput and transmission rate, and to a certain extent solve the problems of network congestion and excessively long scheduling time, meeting the transmission requirements of latency-sensitive services. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the power grid safety production business scheduling method according to an embodiment of the present disclosure;

[0021] Figure 2-A This is an application scenario diagram of an embodiment of this disclosure;

[0022] Figure 2-B This is a schematic diagram illustrating queue priority analysis according to an embodiment of the present disclosure;

[0023] Figure 2-C This is a schematic diagram of the power grid safety production business scheduling framework according to an embodiment of the present disclosure;

[0024] Figure 3 This is a schematic diagram of the structure of the power grid safety production dispatching device according to an embodiment of the present disclosure;

[0025] Figure 4 This is a schematic diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0027] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0028] In related technologies, the power grid safety production business flow often suffers from high transmission latency, which directly affects the service quality of data transmission and may lead to network congestion and excessively long scheduling time, thus failing to meet the transmission requirements of the business flow.

[0029] This embodiment proposes a power grid safety production business scheduling method, such as... Figure 1 As shown, it includes:

[0030] Step 101: Obtain data packets sent by multiple user terminals, and integrate the multiple data packets into multiple queues according to the device information of the user terminals. The multiple queues have different priorities.

[0031] In this step, such as Figure 2-A As shown, based on the 5G Time-Sensitive Network (TSN) converged architecture, this scenario involves 5G uplink data transmission from a single base station and N user terminals. In this scenario, only the generation and scheduling of power grid safety production service flows are considered. Multiple user terminals can randomly send data packets of variable length indefinitely. Data packets generated by different user terminals will be assigned to different queues during queue consolidation, each queue having a different priority, which can be dynamically configured. The device information of the user terminal represents the device's identity identifier, which can be represented by a device ID (Identity Document).

[0032] Step 102: When at least one target user terminal among the multiple user terminals sends a new data packet, the new data packet is added to the target queue corresponding to the target user terminal, wherein the target queue is at least one of the multiple queues.

[0033] In this step, the target user terminal refers to any user terminal among multiple user terminals that sends a new data packet. When the target user terminal sends a new data packet, the data packet is added to the target queue corresponding to the target user terminal.

[0034] Step 103: Use a recurrent neural network model to predict the latency of the new data packet and obtain the latency prediction result.

[0035] In this step, the recurrent neural network model (simpleRNN) can be an RNN (Recurrent Neural Networks) model or a multiple linear regression model.

[0036] In some embodiments, step 103 includes:

[0037] Step 1031: Train the recurrent neural network model, modifying the parameters of the recurrent neural network model in each training session, until the accuracy of the output of the recurrent neural network model reaches a preset threshold, and then stop training the recurrent neural network model.

[0038] Step 1032: Use the trained recurrent neural network model to predict the latency of the new data packet and obtain the latency prediction result.

[0039] In some embodiments, the recurrent neural network model uses a neural network structure with multidimensional input, multi-step prediction, and a single output. Its model design is shown below:

[0040] The input matrix of the model is X = {x1, x2}, where x1 is the feature vector of the current queue length L, and x2 is the feature vector of the current queue priority. The label matrix is ​​Y = {D1, D2, ..., D...} m}, where m is the size of the test dataset, and D i (i < m) represents the actual delay of data packet transmission.

[0041] The neural network structure of an RNN model consists of an input layer, a single intermediate layer (hidden layer), and an output layer, with fully connected layers between each other. The connection between two recurrent neurons is shown in the following formula:

[0042] h (t) =f(uh (t-1) +wX (t) +b) where u represents state-state weights, w represents state-input weights, b represents the bias of the neural network layer, and h (t) X represents the neuron state in layer t. (t) Let f(x) represent the data input to the RNN model, and let f(x) represent the activation function of the t-th layer.

[0043] Output data y (t) The calculation method is shown in the following formula:

[0044] y (t) =g(vh) (t) +c)

[0045] Where v represents the weight, c represents the bias, and h (t) Let y represent the state vector of the neuron in layer t. (t) This represents the final result to be output, and g(x) represents the activation function.

[0046] During the training of the RNN model, the overfitting problem of the time delay prediction results was taken into account, and the dropout algorithm was used to further optimize the recurrent neural network model.

[0047] After training the RNN model using X and Y derived from the test set data, and ensuring that the weights of each layer reach appropriate values, the trained recurrent neural network model is used to predict latency when a new data packet arrives. Assume that a data packet generated by user terminal i enters queue Q. i Queue Q i The queue length at time t is L. i The queue priority is Pr i Then, the latency Dp of data packets in queue i can be predicted based on the RNN model. i for:

[0048] Dp i =RNN(X×{L i Pr i})

[0049] To address unexpected business situations, ensure overall scheduling performance, and meet the Quality of Service (QoS) requirements of data streams, the priority of queues is set to be dynamically variable. This allows the priority Pr of each queue to be dynamically adjusted based on latency prediction results, meaning that higher priority values ​​receive scheduling and resource allocation earlier.

[0050] Using recurrent neural network models, a machine learning approach, for latency prediction enables dynamic resource allocation, thereby improving system throughput and reducing overall data packet transmission latency.

[0051] Step 104: Based on the delay prediction results, adjust the priority of the target queue to obtain the adjusted priority of each queue.

[0052] In this step, the priority of the target queue is adjusted based on the latency prediction results to obtain the adjusted priority of each queue, so as to reduce transmission latency and improve system throughput by adjusting the priority.

[0053] In some embodiments, step 104 includes:

[0054] Step 1041: Determine the priority adjustment scheme for the target queue based on the result of the delay prediction.

[0055] Step 1042: Obtain the adjusted priorities of each queue according to the priority adjustment scheme of the target queue.

[0056] In this step, the overall priority configuration of the system (i.e., the priority adjustment scheme of the target queue) is determined based on the latency prediction results. The priority will affect the final data transmission latency and system throughput by influencing resource allocation, thus transforming the problem into a problem of solving 5G wireless resource allocation.

[0057] The adjusted priorities of each queue are obtained based on the priority of the target queue. The priorities will affect the final data transmission latency and system throughput by influencing resource allocation. Allocating resources according to the adjusted queue priorities can reduce the overall data packet transmission latency and improve system throughput.

[0058] In some embodiments, step 1041 includes:

[0059] Step 10411: Determine the time delay prediction interval range of the target queue based on the time delay prediction result.

[0060] Step 10412: Obtain a delay deterministic range table for all queues, wherein the delay deterministic range table includes the delay deterministic range for each queue.

[0061] Step 10413, based on the time delay prediction interval range Deterministic range of time delay with the target queue Determine the formula in, and Let these be the maximum and minimum delays for deterministic latency of queue i, respectively. and These are the minimum and maximum latency predicted for target queue i, respectively.

[0062] Step 10414, using the formula Determine the priority adjustment scheme for the target queue, ε i Determine the parameters for the target queue priority.

[0063] In the above scheme, based on the results of time delay prediction, samples are taken from the results of historical prediction and qualitatively compared with the upper and lower bounds of the time delay prediction interval to determine the overall priority allocation of the system (i.e., the priority adjustment scheme of the target queue).

[0064] In some embodiments, the result of the time delay prediction includes multiple steps, including step 10411:

[0065] Step 104111: Add the result of each of the time delay predictions to the time delay prediction dataset of the target queue.

[0066] Step 104112: Remove time delay prediction data from the time delay prediction dataset that does not meet the preset confidence level. Based on the time delay prediction dataset after removing the time delay prediction data that does not meet the preset confidence level, determine the time delay prediction interval range of the target queue.

[0067] In the above scheme, each latency prediction result is added to the historical dataset of latency prediction results (i.e., the latency prediction dataset of the target queue). A confidence interval with a confidence level of 1-α is set, and data statistics and sampling are performed on the historical dataset to obtain the current latency prediction interval range [Dp]. min , Dp max (i.e., the predicted time delay range of the target queue).

[0068] In some embodiments, step 10414 includes:

[0069] Step 104141, in the ε i When the target queue is within the first range, its priority is reduced.

[0070] Step 104142, in the ε i When the target queue is within the second range, its priority is increased.

[0071] In the above scheme, it is assumed that the required time delay determinism range is [T]. min T max ], compare the interval boundary value Dp of the time delay prediction min and Dp max Priority allocation is determined by the difference between the upper and lower bounds of the two intervals. A qualitative analysis of the difference between the upper and lower bounds of the two intervals is performed, such as... Figure 2-B As shown:

[0072] [1] satisfies the deterministic requirement, and the priority theoretically does not need to change, Dp min -T min If positive, Dp max -T max Negative;

[0073] [2] The overall prediction interval should be adjusted, and the transmission speed should be increased to reduce latency. Therefore, the priority needs to be increased, and Dp min -T min If positive, Dp max -T max It is positive;

[0074] [3] The overall prediction interval should be adjusted, and the transmission speed should be reduced to decrease latency. Therefore, the priority needs to be reduced, and Dp min -T min If negative, Dp max -T max Negative;

[0075] [4] The prediction delay interval is abnormal. Adjusting the sampling confidence α can achieve the overall reduction of the prediction interval, which is unrelated to the priority adjustment.

[0076] [5] explains that the current predicted latency is too small, and the transmission rate should be reduced to increase the latency. Therefore, the priority needs to be reduced, and compared to case [3], a smaller priority is needed to make a wider range of latency adjustments, Dp min -T min If negative, Dp max -T max The case where the sum is negative and the absolute value of the sum is greater than the case[3];

[0077] [6] and [5] are opposite, indicating that the current predicted delay is too large and the transmission rate should be increased to reduce the delay. Therefore, the priority needs to be increased and Dp should be increased. min -T min If positive, Dp max -T max The case where the sum of the two is positive and the absolute value of the sum is greater than the case[2].

[0078] Based on the above classification and qualitative analysis, the following reference formula can be used to design a priority adjustment scheme:

[0079]

[0080] ε i The smaller the value, the closer it is to the situation [5], and the lower the priority should be.

[0081] ε i The larger the value, the closer it is to the case[6], and the higher the priority should be.

[0082] The reference formula for the priority adjustment scheme described above demonstrates the impact of delay prediction on priority adjustment, thereby indirectly affecting the data transmission rate.

[0083] Step 105: Based on the adjusted priority, the bandwidth resource allocation model is solved using the near-end strategy optimization pruning algorithm to obtain the bandwidth resource allocation strategy.

[0084] In this step, the feasible region and objective function of problem P3 are both non-convex, so the problem has NP (Non-deterministic Polynomial) characteristics, and deep learning-related algorithms can be considered to solve the optimization problem.

[0085] The Proximal Policy Optimization (PPO) algorithm is based on the Actor-Critic framework. Compared to the traditional Policy Gradient algorithm, PPO can achieve mini-batch updates over multiple training steps, solving the problem of difficult-to-determine step size in the Policy Gradient algorithm. PPO can easily overcome the problem of unstable policy updates and low data efficiency because it uses three networks: a Critic network, an Actor network, and an old Actor network. The Actor network is responsible for parameter updates and feedback adjustment, the old Actor network interacts with the environment and makes action selection, and the Critic network provides the value of each state as a data reference for updating the network. In the system, individuals interact with the environment, starting from a state and obtaining an action space according to the importance sampling theorem, selecting an action to execute and receiving a reward.

[0086] In the design of the Proximal Policy Optimization (PPO-Clip) algorithm using the Clip algorithm to solve the bandwidth allocation policy, the state space of the PPO algorithm is S={U,Pr}, the action space is A={W(t)}, and the reward function re is designed as shown in the following formula:

[0087] re=Ly pre -Ly cur

[0088] After multiple iterations and network updates, when re approaches stability, the optimal solution to the optimization problem can be guaranteed to be output. pre and Ly cur The objective function in P3 has the following formula:

[0089]

[0090] The pseudocode for the near-end strategy optimized pruning algorithm is shown in Table 1 below:

[0091] Table 1

[0092]

[0093]

[0094] Using a proximal strategy to optimize the pruning algorithm achieves better performance and prevents the algorithm strategy from updating too quickly; therefore, the objective function L... t Designed as follows:

[0095]

[0096] Wherein, π(a t |s t ) indicates that in state s t The action 'a' is selected based on the Actor network. t The concept, This represents the estimate of the advantage function, where ε represents the parameters of the Clip algorithm. The advantage function is limited to the range [1-ε, 1+ε] to avoid excessively rapid policy updates. The formula for calculating the advantage function is shown below:

[0097]

[0098] The process of solving the problem using the near-end strategy optimization pruning algorithm is shown in Table 2:

[0099] Table 2

[0100]

[0101]

[0102] In some embodiments, step 105, the process of constructing the bandwidth resource allocation model includes:

[0103] Step 1051: The resource allocation abstract model is transformed using the Lyapunov optimization algorithm to obtain a bandwidth resource allocation model, wherein the resource allocation abstract model is a model that can calculate the amount of bandwidth resources allocated to each queue at any time.

[0104] In the above scheme, a data transmission model is constructed for the application scenario of N user terminals and 1 base station for the 5G uplink. The user terminals will randomly generate power grid safety production business data streams according to a Poisson distribution. The length of the data packets fluctuates within a certain range and is not fixed. The scheduler performs queue shaping of the data packets according to different user terminal sources, dividing them into N different queues. The total data length of the queues is Q(t) = {Q1(t), Q2(t), ..., Q...} N (t)},Q i Let Pr(t) represent the queue length of queue i at time t, and the priority of the queue at time t be represented as Pr(t) = {Pr1(t), Pr2(t), ..., Pr...}. N (t)},Pr i Let Pr(t) represent the queue priority of queue i at time t, where Pr(t) = [1, 2, ..., N]. and Z + Represents the set of positive integers.

[0105] Data transmission is performed using a priority-based scheduling method between different queues. Each time a new data packet arrives, a multiple linear regression model or an RNN model is used to predict latency, and the queue priorities in the system are dynamically adjusted according to the following formula:

[0106]

[0107] Queue Q i The queue update process for (t) is as follows:

[0108] Q i (t+1)=max{Q i (t)-b i (t), 0}+a i (t)

[0109] Among them, b i (t) represents the amount of data in queue i at time t, b i (t)=T*R i (t), R i (t) represents the data transmission rate of queue i at time t; T represents the fixed time required for one scheduled transmission action, i.e., the time slot; a i (t) represents the amount of data generated by device i at time t, that is, the rate of new data packets arriving in queue i at time t. In order to simulate the randomness and suddenness of the power grid safety production business flow, a Poisson distribution with parameter λ is used to simulate the generation of business flow data.

[0110] Let the queue bandwidth allocated to N user terminals be W(t) = {W1(t), W2(t), ..., W...} N (t)},W i (t) represents the amount of bandwidth allocated to queue i at time t. And satisfy 0≤W i (t)≤BW max.

[0111] According to the improved Shannon formula, the transmission processing rate of queue i is as follows:

[0112]

[0113] W i (t) represents the bandwidth allocated to queue i, P represents the transmit power, c represents the receive signal-to-noise ratio, vk represents the channel dispersion, L represents the data packet length in bits, and fq -1 express The inverse function of γ, where γ is the confidence level of the deterministic delay estimate, and 1-γ represents the probability of packet loss, i.e., the transmission error rate. This represents the signal-to-noise ratio of the receiving channel and is usually a fixed value.

[0114] The system throughput can be determined by the transmission rate R. i The sum of (t) is expressed as:

[0115]

[0116] To represent the determinism of time delay in power grid safe production operations, the confidence level γ of the time delay deterministic estimation is used, γ = P{Dp}. min <D<Dp max}

[0117] Since the delay of a data packet includes the channel transmission delay x and the queuing interference delay W, i.e., D = x + w, the following formula applies:

[0118] γ=P{Dp min <x+w<Dp max}

[0119] =P{Dp min <x<βDp max}P{0<w<(1-β)Dp max}

[0120] β represents the delay scaling factor.

[0121] The randomness of transmission delay mainly comes from the fading variables of the wireless channel. Therefore, we can assume that the preferred transmission channel parameter in this disclosure is a Rayleigh random channel with δ, and x represents a function of the channel fading random variable. Thus, the interval probability of transmission delay x can be obtained from the Shannon channel capacity formula and the cumulative distribution function of the channel random variable, as shown below:

[0122]

[0123] In this formula:

[0124]

[0125]

[0126] L represents the current data packet length, β represents the delay scaling factor, and W... i (t) represents the channel bandwidth of queue i, and t represents the time variable.

[0127] In the assumed data transmission model, due to the randomness of data volume and scheduling optimization issues, it is difficult to obtain the statistical envelope of the cumulative arrival volume of services and the cumulative service volume of the system in advance. Furthermore, it is certain that the service arrival process and the service process of the data transmission model are independent of each other. Therefore, stochastic network calculus based on the Moment Generating Function (MGF) is chosen for network analysis. The interval probability of queuing delay can be solved according to the theory of stochastic network calculus based on the Moment Generating Function (MGF), yielding the following inequality:

[0128]

[0129] θ satisfies the constraint: E[e θA(1) E[e -θS(1) If ≤1, the inequality is close to 1, then the equality in the above equation is approximately true.

[0130] A(t) and S(t) represent the network arrival process and service process within the time interval (0, t). The arrival process A(t) follows a Poisson distribution, and the service process S(t) is the cumulative service volume, defined using the Shannon formula.

[0131] Therefore, the following inequality can be derived:

[0132]

[0133] From the above formula:

[0134]

[0135]

[0136]

[0137] And under the aforementioned θ constraint, we can obtain the following equation:

[0138]

[0139] To address the latency determinism issue of bursty data, it is necessary to ensure the boundedness of latency, which means allocating bandwidth resources to make γ approximately equal to 1.

[0140] Based on the above analysis, the following optimization problem P1 can be proposed:

[0141] max wi(t) {U(W, γ)}

[0142] stC1: 0 < γ < 1

[0143]

[0144]

[0145] C1 ensures the deterministic requirement of data transmission, C2 guarantees bandwidth allocation, and C3 guarantees the stability of each queue.

[0146] Analysis of the properties of problem P1 reveals that its optimization constraints are nonlinear, and both the feasible set and the objective function are non-concave. Therefore, it is difficult to solve using general optimization algorithms. To facilitate problem solving, the Lyapunov optimization algorithm can be used to transform the problem.

[0147] Controlling the Lyapunov function to approach negative zero can achieve system stability. Lyapunov optimization algorithms can ensure the stability of systems of different forms.

[0148] In some embodiments, step 1051 includes:

[0149] Step 10511: Based on the constraints of the resource allocation abstract model, the Lyapunov optimization algorithm is used to transform the resource allocation abstract model to obtain the first model. The constraints of the resource allocation abstract model include: ensuring the deterministic requirement of data transmission (0 < γ < 1) and guaranteeing the allocation of overall bandwidth. And ensure the stability of each queue Wherein, w i (t) represents the bandwidth allocated to queue i, γ represents the confidence level of the deterministic delay estimate, and BW max Q represents the maximum value of bandwidth resources, T represents the duration of a fixed time slot, and Q represents the maximum value of bandwidth resources. i (t) represents the queue length of queue i at time t, and n represents the number of bandwidth resource partitions.

[0150] Step 10512: Transform the resource allocation abstract model into the first model min. w(t) {E[G(t)|Q(t)]-VU}, where V represents the control coefficient, U represents the overall throughput, E[G(t)|Q(t)] represents the expectation of G(t) under the condition of Q(t), Q(t) represents the queue length at time t, G(t) is an intermediate function, γ represents the confidence level of the time delay deterministic estimate, and w(t) represents the bandwidth allocated to the queue at time t.

[0151] Step 10513: Separate the bandwidth w(t) from the first model to obtain the bandwidth resource allocation model. Among them, E[Q i (t)b i (t)|Q i [t] represents Q. i (t) Q under condition i (t)b iThe expectation of (t), Q i (t) represents the queue length of queue i at time t, b i w(t) represents the amount of data scheduled by queue i at time t, V represents the control coefficient, U represents the overall throughput, N represents the number of queues allocated bandwidth resources, and w(t) represents the bandwidth allocated to queue i at time t.

[0152] In the above scheme, based on the algorithm model in problem P1, and considering constraint C3, the queue stability can be represented by a Lyapunov function, L(t), as shown below:

[0153]

[0154] According to the relevant theory of Lyapunov's algorithm, the Lyapunov drift function Δ(Q(t)) is defined as follows:

[0155] Δ(Q(t))=L(Q(t+1))-L(Q(t))

[0156] Δ(Q(t)) is often difficult to solve, therefore it is necessary to analyze the upper bound of the drift function, according to the theorem formula of Lyapunov optimization algorithm:

[0157] Δ(Q(t))=L(Q(t+1))-L(Q(t))≤D+E[G(t)|Q(t)]

[0158] We can obtain the following definition:

[0159]

[0160]

[0161] Minimizing the quadratic Lyapunov drift function may lead to a stable backpressure routing algorithm, also known as the maximum weight algorithm. This algorithm adds a weighted penalty term to the Lyapunov drift and minimizes the sum, thus achieving both network stability and penalty minimization. The Lyapunov drift penalty function is designed as follows:

[0162] Δ(Q(t))-VU≤D+E[G(t)|Q(t)]-VU

[0163] V is the control coefficient, which controls the impact of the penalty function on the overall system. By adjusting the value of V, the function can be optimized. U is the objective function in P1, which is the system throughput.

[0164] At this point, the problem is transformed into minimizing the Lyapunov drift penalty function, therefore the problem becomes P2:

[0165] min w(t){E(G(t)|Q(t)]-VU}

[0166] stC1: 0 < γ < 1

[0167]

[0168] Constraint C1 ensures the deterministic requirement of data transmission, and C2 guarantees the allocation of system bandwidth. This disclosure preferably focuses on solving the allocation strategy for the bandwidth vector W(t), separating W from the objective function of P2. i (t), further transforming problem P2 into problem P3:

[0169]

[0170] C1∶0<γ<1

[0171] The constraints of the bandwidth resource allocation model include: ensuring the deterministic requirement of data transmission (0 < γ < 1) and guaranteeing the overall bandwidth allocation. Among them, E[Q i (t)b i (t)|Q i [t] represents Q. i (t) Q under condition i (t)b i The expectation of (t), Q i (t) represents the queue length of queue i at time t, b i (t) represents the amount of data scheduled by queue i at time t, V represents the control coefficient, U represents the overall throughput, N represents the number of queues allocated bandwidth resources, and w i (t) represents the bandwidth allocated to queue i at time t, γ represents the confidence level of the deterministic delay estimate, and BW max This represents limited bandwidth resources, and n represents the number of user terminals.

[0172] Step 106: Based on the bandwidth resource allocation strategy, obtain the scheduling method for power grid safe production services.

[0173] In this step, after solving the bandwidth resource allocation strategy using the near-end policy optimization pruning algorithm, the data transmission rate of each queue is calculated according to the improved Shannon formula. Data packet scheduling of the queues is performed within a fixed time slot to obtain the scheduling method of power grid safe production services.

[0174] In some embodiments, step 106 includes:

[0175] Step 1061: Calculate the data transmission rate of each first queue using the first formula according to the bandwidth resource allocation strategy.

[0176] The expression for the first formula is:

[0177] Among them, R i (t) represents the data transmission rate of queue i, W i (t) represents the bandwidth allocated to queue i at time t, P represents the transmit power, c represents the receive signal-to-noise ratio, vk represents the channel dispersion, L represents the data packet length, and fq -1 yes The inverse function of , where t represents the time parameter, γ is the confidence level of the time delay deterministic estimate, and 1-γ represents the probability of packet loss.

[0178] Step 1062: Based on the data transmission rate of each first queue, schedule the data packets entering each first queue within a fixed time slot to obtain the scheduling method for power grid safety production operations.

[0179] In the above scheme, after solving the bandwidth resource allocation strategy using the near-end strategy optimization pruning algorithm, the data transmission rate of each queue is calculated according to the improved Shannon formula. Data packet scheduling of the queues is performed within a fixed time slot to obtain the scheduling method of power grid safety production business. This improves data transmission efficiency and resource utilization, thereby avoiding the problems of network congestion and excessively long scheduling time, and meeting the transmission requirements.

[0180] The above scheme uses a recurrent neural network model to predict the latency of new data packets sent by the target user terminal. Based on the latency prediction results, a priority adjustment scheme for the target queue is determined, and the adjusted queue priority is obtained. Then, the Lyapunov optimization algorithm is used to transform the resource allocation abstract model to obtain a bandwidth resource allocation model, which simplifies the resource allocation abstract model. Based on the adjusted queue priority, the near-end policy optimization pruning algorithm is used to solve the bandwidth resource allocation model to obtain the bandwidth resource allocation strategy. Finally, based on the bandwidth resource allocation strategy, the scheduling method for power grid safety production services is obtained. Scheduling according to this method can effectively reduce the overall transmission latency, improve system throughput and transmission rate, and solve the problems of network congestion and excessively long scheduling time to a certain extent, thus meeting the transmission requirements of latency-sensitive services.

[0181] Based on the same inventive concept, specific descriptions are given of the application scenarios corresponding to the power grid safety production business scheduling method of the above embodiments, such as... Figure 2-C As shown, the details are as follows:

[0182] In the entire scheduling system, data packets are randomly generated according to a Poisson distribution and enter different queues (Q1...Q1) based on their origin from the user equipment (i.e., the user terminal). nAfterwards, packet delay prediction is performed. The current queue length is L, and the current queue priority is Pr. Based on the packet delay prediction results, the queue dynamic priority is adjusted. The system bandwidth allocation is completed using the dynamic priority of the queue. The transmission speed of queue scheduling is obtained based on the bandwidth allocation results. Packet scheduling is performed on the queue. The pseudocode of the overall system scheduling algorithm is shown in Table 3.

[0183] Table 3

[0184]

[0185]

[0186] In the entire scheduling system, the algorithm's complexity primarily depends on the near-end policy optimization pruning algorithm, which can obtain the optimal bandwidth allocation strategy in each decision cycle, and its algorithm complexity is O(n^2). 2 Therefore, the overall scheduling algorithm has a time complexity of O(n^2). 2 ).

[0187] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0188] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0189] Based on the same inventive concept, corresponding to any of the above embodiments, this disclosure also provides a power grid safety production business dispatching device.

[0190] refer to Figure 3 The power grid safety production dispatching device includes: a queue determination module, a data packet determination module, a delay prediction module, a priority determination module, a bandwidth resource allocation strategy determination module, and the dispatch determination module.

[0191] The queue determination module 301 is configured to acquire data packets sent by multiple user terminals, and integrate the multiple data packets into multiple queues according to the device information of the user terminals, wherein the multiple queues have different priorities.

[0192] The data packet determination module 302 is configured to add the new data packet to the target queue corresponding to the target user terminal when at least one target user terminal among a plurality of user terminals sends a new data packet, wherein the target queue is at least one of the plurality of queues;

[0193] The latency prediction module 303 is configured to use a recurrent neural network model to predict the latency of the new data packet and obtain the latency prediction result.

[0194] The priority determination module 304 is configured to adjust the priority of the target queue based on the result of the delay prediction, so as to obtain the adjusted queue priority.

[0195] The bandwidth resource allocation strategy determination module 305 is configured to solve the bandwidth resource allocation model using a near-end strategy optimization pruning algorithm based on the adjusted priority to obtain the bandwidth resource allocation strategy.

[0196] The scheduling determination module 306 is configured to determine the scheduling method for power grid safe production services based on the bandwidth resource allocation strategy.

[0197] In some embodiments, the priority determination module 304 includes:

[0198] The priority adjustment scheme determination unit is configured to determine the priority adjustment scheme of the target queue based on the result of the delay prediction.

[0199] The adjusted priority determination unit is configured to obtain the adjusted queue priority of each queue according to the priority adjustment scheme of the target queue.

[0200] In some embodiments, the priority adjustment scheme determination unit includes:

[0201] The delay prediction interval range determination subunit is configured to determine the delay prediction interval range of the target queue based on the delay prediction result;

[0202] The delay deterministic range table acquisition subunit is configured to acquire the delay deterministic range table of all the queues, wherein the delay deterministic range table includes the delay deterministic range of each queue;

[0203] The formula determines the sub-unit, which is configured to predict the time delay range based on the given time delay interval. Deterministic range of time delay with the target queue Determine the formula in, and Let these be the maximum and minimum delays for deterministic latency of queue i, respectively. and These are the minimum and maximum latency predicted for target queue i, respectively.

[0204] The adjustment scheme determines the sub-unit, which is configured to utilize the formula. Determine the priority adjustment scheme for the target queue, ε i Determine the parameters for the target queue priority.

[0205] In some embodiments, the result of the time delay prediction includes multiple time delay prediction interval range determination sub-units, which are specifically configured as follows:

[0206] The result of each of the aforementioned time delay predictions is added to the time delay prediction dataset of the target queue;

[0207] Time delay prediction data that does not meet the preset confidence level is removed from the time delay prediction dataset. Based on the time delay prediction dataset after removing the time delay prediction data that does not meet the preset confidence level, the time delay prediction interval range of the target queue is determined.

[0208] In some embodiments, the adjustment scheme determines the sub-unit, specifically configured as follows:

[0209] In the ε i When the target queue is within the first range, its priority is reduced.

[0210] In the ε i When the target queue is within the second range, its priority is increased.

[0211] In some embodiments, the delay prediction module 303 is specifically configured as follows:

[0212] The recurrent neural network model is trained, and the parameters of the recurrent neural network model are modified in each training session until the accuracy of the output result of the recurrent neural network model reaches a preset threshold, at which point the training of the recurrent neural network model is stopped.

[0213] The latency prediction results are obtained by using a trained recurrent neural network model to predict the latency of the new data packets.

[0214] In some embodiments, the power grid safety production dispatching device further includes a bandwidth resource allocation strategy determination module, configured to:

[0215] The Lyapunov optimization algorithm is used to transform the resource allocation abstract model to obtain a bandwidth resource allocation model, wherein the resource allocation abstract model is a model that can calculate the amount of bandwidth resources allocated to each queue at any time.

[0216] In some embodiments, the bandwidth resource allocation strategy determination module is specifically configured as follows:

[0217] Based on the constraints of the resource allocation abstract model, the Lyapunov optimization algorithm is used to transform the resource allocation abstract model to obtain the first model. The constraints of the resource allocation abstract model include: ensuring the deterministic requirement of data transmission (0 < γ < 1) and guaranteeing the allocation of overall bandwidth. And ensure the stability of each queue Wherein, w i (t) represents the bandwidth allocated to queue i, γ represents the confidence level of the deterministic delay estimate, and BW max Q represents the maximum value of bandwidth resources, T represents the duration of a fixed time slot, and Q represents the maximum value of bandwidth resources. i (t) represents the queue length of queue i at time t, and n represents the number of bandwidth resource partitions;

[0218] The resource allocation abstract model is transformed into the first model min. w(t) {E[G(t)|Q(t)]-VU}, where V represents the control coefficient, U represents the overall throughput, E[G(t)|Q(t)] represents the expectation of G(t) under the condition of Q(t), Q(t) represents the queue length at time t, G(t) is an intermediate function, γ represents the confidence level of the time delay deterministic estimate, and w(t) represents the bandwidth allocated to the queue at time t;

[0219] By separating the bandwidth w(t) from the first model, we obtain the bandwidth resource allocation model. Among them, E[Q i (t)b i (t)|Q i [t] represents Q. i (t) Q under condition i (t)b i The expectation of (t), Q i (t) represents the queue length of queue i at time t, b i w(t) represents the amount of data scheduled by queue i at time t, V represents the control coefficient, U represents the overall throughput, N represents the number of queues allocated bandwidth resources, and w(t) represents the bandwidth allocated to queue i at time t.

[0220] In some embodiments, the scheduling determination module 306 is specifically configured as follows:

[0221] Based on the bandwidth resource allocation strategy, the data transmission rate of each first queue is calculated using the first formula:

[0222] The expression for the first formula is:

[0223] Among them, R i (t) represents the data transmission rate of queue i, W i (t) represents the bandwidth allocated to queue i at time t, P represents the transmit power, c represents the receive signal-to-noise ratio, vk represents the channel dispersion, L represents the data packet length, and fq -1 yes The inverse function of , where t represents the time parameter, γ is the confidence level of the time delay deterministic estimate, and 1-γ represents the probability of packet loss;

[0224] Based on the data transmission rate of each first queue, the data packets entering each first queue are scheduled within a fixed time slot to obtain the scheduling method for power grid safety production operations.

[0225] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0226] The apparatus described above is used to implement the corresponding power grid safety production business scheduling method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0227] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the power grid safety production business scheduling method described in any of the above embodiments.

[0228] Figure 4 This illustration shows a more specific hardware structure diagram of an electronic device provided in this embodiment. The device may include: a processor 401, a memory 402, an input / output interface 403, a communication interface 404, and a bus 405. The processor 401, memory 402, input / output interface 403, and communication interface 404 are interconnected internally via the bus 405.

[0229] The processor 401 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0230] The memory 402 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 402 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 402 and is called and executed by the processor 401.

[0231] Input / output interface 403 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0232] Communication interface 404 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0233] Bus 405 includes a pathway for transmitting information between various components of the device (e.g., processor 401, memory 402, input / output interface 403, and communication interface 404).

[0234] It should be noted that although the above-described device only shows the processor 401, memory 402, input / output interface 403, communication interface 404, and bus 405, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0235] The electronic devices described above are used to implement the corresponding power grid safety production business scheduling methods in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0236] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to execute the power grid safety production business scheduling method as described in any of the above embodiments.

[0237] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0238] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the power grid safety production business scheduling method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0239] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0240] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0241] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0242] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for dispatching power grid safety production operations, characterized in that, The method includes: Data packets sent by multiple user terminals are acquired, and the multiple data packets are integrated into multiple queues according to the device information of the user terminals, wherein the multiple queues have different priorities; When at least one target user terminal among multiple user terminals sends a new data packet, the new data packet is added to the target queue corresponding to the target user terminal, wherein the target queue is at least one of the multiple queues; The latency prediction result is obtained by using a recurrent neural network model to predict the latency of the new data packet. Based on the results of the latency prediction, the priority of the target queue is adjusted to obtain the adjusted priority of each queue; Based on the adjusted priority, the pre-built bandwidth resource allocation model is solved using the near-end strategy optimization pruning algorithm to obtain the bandwidth resource allocation strategy. Based on the bandwidth resource allocation strategy, the scheduling method for power grid safety production services is obtained; The construction process of the bandwidth resource allocation model includes: The resource allocation abstract model is transformed using the Lyapunov optimization algorithm to obtain a bandwidth resource allocation model, wherein the resource allocation abstract model is a model that can calculate the amount of bandwidth resources allocated to each queue at any time, including: Based on the constraints of the resource allocation abstract model, the Lyapunov optimization algorithm is used to transform the resource allocation abstract model to obtain the first model. The constraints of the resource allocation abstract model include: ensuring the deterministic requirement of data transmission. Ensure overall bandwidth allocation And ensure the stability of each queue , wherein Queue The allocated bandwidth The confidence level represents the time delay deterministic estimate. Represents the maximum value of bandwidth resources. Represents the duration of a fixed time slot. Queue exist The queue length at any given time. Indicates the number of bandwidth resource segments; By separating the bandwidth from the first model, a bandwidth resource allocation model is obtained.

2. The method according to claim 1, characterized in that, The step of adjusting the priority of the target queue based on the delay prediction result to obtain the adjusted priority of each queue includes: Based on the results of the latency prediction, a priority adjustment scheme for the target queue is determined; Based on the priority adjustment scheme of the target queue, the adjusted priorities of each queue are obtained.

3. The method according to claim 2, characterized in that, The step of determining the priority adjustment scheme for the target queue based on the delay prediction result includes: Based on the results of the latency prediction, the latency prediction interval range of the target queue is determined; Obtain a delay deterministic range table for all the queues, wherein the delay deterministic range table includes the delay deterministic range for each queue; Based on the time delay prediction interval range Deterministic range of time delay with the target queue Determine the formula ,in, and Let these be the minimum and maximum delays for deterministic delay of queue i, respectively. and These are the minimum and maximum latency predicted for target queue i, respectively. Using the formula Determine the priority adjustment scheme for the target queue. Determine the parameters for the target queue priority.

4. The method according to claim 3, characterized in that, The latency prediction results include multiple parameters, and determining the latency prediction interval range of the target queue based on the latency prediction results includes: The result of each of the aforementioned time delay predictions is added to the time delay prediction dataset of the target queue; Time delay prediction data that does not meet the preset confidence level is removed from the time delay prediction dataset. Based on the time delay prediction dataset after removing the time delay prediction data that does not meet the preset confidence level, the time delay prediction interval range of the target queue is determined. .

5. The method according to claim 3, characterized in that, The formula used Determine the priority adjustment scheme for the target queue, wherein Parameters for determining the priority of the target queue include: In the When the target queue is within the first range, its priority is reduced. In the When the target queue is within the second range, its priority is increased.

6. The method according to claim 1, characterized in that, The step of using a recurrent neural network model to predict the latency of the new data packet, and obtaining the latency prediction result, includes: The recurrent neural network model is trained, and the parameters of the recurrent neural network model are modified in each training session until the accuracy of the output result of the recurrent neural network model reaches a preset threshold, at which point the training of the recurrent neural network model is stopped. The latency prediction results are obtained by using a trained recurrent neural network model to predict the latency of the new data packets.

7. The method according to claim 1, characterized in that, The bandwidth resource allocation is performed using the Lyapunov optimization algorithm to obtain a bandwidth resource allocation model. This abstract model is capable of calculating the amount of bandwidth resources allocated to each queue at any given time, and includes: Transform the resource allocation abstract model into a first model. ,in, Indicates the control coefficient. Indicates total throughput. Indicates in under conditions Expectations express The queue length at any given time. For intermediate functions, The confidence level represents the time delay deterministic estimate. Indicates the queue is The bandwidth allocated at any given time; The bandwidth in the first model Separate the bandwidth resource allocation model to obtain the model. ,in, Representative at under conditions Expectations Queue exist The queue length at any given time. Queue exist The amount of data to be scheduled at any given time Indicates the control coefficient. Indicates total throughput. This represents the number of queues allocated bandwidth resources. Indicates the queue is The bandwidth allocated at any given time.

8. The method according to claim 1, characterized in that, The step of obtaining the scheduling method for power grid safe production services based on the bandwidth resource allocation strategy includes: Based on the bandwidth resource allocation strategy, the data transmission rate of each first queue is calculated using the first formula: The expression for the first formula is: ; in, For queue Data transmission rate, Queue exist The bandwidth allocated at any given time. Indicates the transmission power. This represents the received signal-to-noise ratio coefficient. Indicates channel dispersion, Indicates the length of the data packet. yes inverse function, Indicates time parameter, It is the confidence level of the deterministic estimation of time delay. Indicates the probability of data packet loss; Based on the data transmission rate of each first queue, the data packets entering each first queue are scheduled within a fixed time slot to obtain the scheduling method for power grid safety production operations.

9. A power grid safety production dispatching device, characterized in that, include: Queue determination module, packet determination module, latency prediction module, priority determination module, bandwidth resource allocation strategy determination module, and scheduling determination module: The queue determination module is configured to acquire data packets sent by multiple user terminals, and integrate the multiple data packets into multiple queues according to the device information of the user terminals, wherein the multiple queues have different priorities. The data packet determination module is configured to add the new data packet to the target queue corresponding to the target user terminal when at least one target user terminal among a plurality of user terminals sends a new data packet, wherein the target queue is at least one of the plurality of queues; The latency prediction module is configured to use a recurrent neural network model to predict the latency of the new data packet and obtain the latency prediction result. The priority determination module is configured to adjust the priority of the target queue based on the result of the delay prediction, so as to obtain the adjusted priority of each queue. The bandwidth resource allocation strategy determination module is configured to solve a pre-built bandwidth resource allocation model using a near-end strategy optimization pruning algorithm based on the adjusted priority to obtain a bandwidth resource allocation strategy; and to transform the resource allocation abstract model using a Lyapunov optimization algorithm to obtain a bandwidth resource allocation model, wherein the resource allocation abstract model is a model that can calculate the amount of bandwidth resources allocated to each queue at any time, including: transforming the resource allocation abstract model using a Lyapunov optimization algorithm based on the constraints of the resource allocation abstract model to obtain a first model, wherein the constraints of the resource allocation abstract model include ensuring the deterministic requirement of data transmission. Ensure overall bandwidth allocation And ensure the stability of each queue , wherein Queue The allocated bandwidth The confidence level represents the time delay deterministic estimate. Represents the maximum value of bandwidth resources. Represents the duration of a fixed time slot. Queue exist The queue length at any given time. This represents the number of bandwidth resource partitions; by separating the bandwidth in the first model, a bandwidth resource allocation model is obtained; The scheduling determination module is configured to determine the scheduling method for power grid safe production services based on the bandwidth resource allocation strategy.

Images