Resource scheduling method and system of wireless sensor network, terminal and storage medium

Abstract

The invention discloses a resource scheduling method and system for a wireless sensor network, a terminal and a storage medium, and the method comprises the steps: obtaining wireless sensor network state information of a target industrial scene, carrying out the feature extraction to obtain feature information, and carrying out the standardization processing to obtain target feature information; creating a scheduling parameter prediction model, performing optimization processing to obtain a target scheduling parameter prediction model, inputting the target feature information into the target scheduling parameter prediction model, outputting target scheduling parameters, and performing scheduling scheme generation to obtain a resource scheduling scheme; and performing performance analysis on the resource scheduling scheme to obtain a performance analysis result, performing dynamic optimization to obtain a target resource scheduling scheme, and performing resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result. The generation time of the scheduling scheme is controlled at the millisecond level, dynamic response is timely, instruction execution safety is guaranteed, multi-objective optimization can be achieved, and resource scheduling efficiency is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource scheduling, and in particular to a resource scheduling method and system for a wireless sensor network, a terminal and a computer readable storage medium. BACKGROUND

[0002] Wireless sensor networks (WSN) are widely used in industrial Internet of Things. By deploying nodes in production line equipment and warehouse systems, key data such as equipment load and vibration are collected to support predictive maintenance and production line scheduling.

[0003] The existing resource scheduling scheme is implemented by TDMA (Time Division Multiple Access) scheduling, but the existing TDMA scheduling industrial WSN system has significant defects, which directly affects the production continuity and operation and maintenance cost. The specific problems include: 1. High device energy consumption and maintenance cost: Industrial sensors are mostly deployed on mechanical arms, AGVs (Automated Guided Vehicles) and other devices, and battery replacement requires downtime. The complexity of the existing algorithm is O(n²), and the device local calculation energy consumption accounts for more than 25%. Moreover, it is not adapted to load fluctuations, and the node lifetime is less than 6 months, which is far below the industrial 12-month maintenance cycle requirement; 2. High instruction delay and safety risk: In industrial scenarios, device fault warning and other instructions need to be transmitted within 50ms, while the existing algorithm takes 3-6 seconds to generate a scheduling scheme in a 500-node network, which easily leads to device response lag, causing quality defects or safety accidents; 3. Poor adaptability and weak matching to dynamic production lines: The addition and deletion of production line nodes, AGV movement and plant electromagnetic interference cause dynamic changes in network topology and link stability. The existing algorithm is designed for fixed topology, and after the node state changes, the scheduling conflict rate exceeds 35%, resulting in data packet loss and affecting the accuracy of the MES system (a production informatization management system for the execution layer of a manufacturing enterprise workshop); 4. Imbalance of multi-objective optimization: Industrial monitoring needs to balance "low energy consumption, low delay and high reliability", but the existing method mostly optimizes energy consumption, leading to performance imbalance; 5. Insufficient dynamic response: When the production line equipment is under high load (for example, the mechanical arm is running at high speed), the sensor acquisition frequency needs to increase from 10 times per second to 100 times per second. Traditional algorithms cannot adjust the time slots in real time, resulting in congestion of burst data and difficulty in timely uploading of key working condition parameters.

[0004] Therefore, the prior art still needs to be improved and developed. SUMMARY

[0005] The main purpose of the present application is to provide a wireless sensor network resource scheduling method, system, terminal and computer readable storage medium, aiming at solving the problems of high instruction delay, multi-target optimization imbalance and insufficient dynamic response in the prior art when performing resource scheduling.

[0006] To achieve the above-mentioned purpose, the present application provides a wireless sensor network resource scheduling method, which comprises the following steps: Obtain the wireless sensor network state information of the target industrial scene, perform feature extraction on the wireless sensor network state information to obtain feature information, and perform standardization processing on the feature information to obtain target feature information; Create a scheduling parameter prediction model, optimize the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, output target scheduling parameters, and generate a scheduling scheme according to the target scheduling parameters to obtain a resource scheduling scheme; Perform performance analysis on the resource scheduling scheme to obtain performance analysis results, dynamically optimize the resource scheduling scheme according to the performance analysis results to obtain a target resource scheduling scheme, and perform resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0007] Optionally, the wireless sensor network resource scheduling method, wherein the obtaining the wireless sensor network state information of the target industrial scene, performing feature extraction on the wireless sensor network state information to obtain feature information, specifically comprises: Obtain the wireless sensor network state information of the target industrial scene, wherein the wireless sensor network state information comprises network node quantity, inter-node distance, data packet length, node initial energy and production data priority; Perform multi-dimensional feature extraction on the wireless sensor network state information to obtain feature information, wherein the feature information comprises basic features and engineering features, the basic features comprise network size features, spatial features, communication features and energy features, and the engineering features comprise network density features, priority interaction features, energy density features and complexity features.

[0008] Optionally, the wireless sensor network resource scheduling method, wherein the standardization processing of the feature information to obtain target feature information specifically comprises: Perform screening processing on the basic features of the feature information to obtain target basic features, wherein the screening processing comprises feature checking processing and redundant feature elimination processing; The engineering features of the feature information are preprocessed to obtain target engineering features, wherein the preprocessing includes missing value processing and abnormal value processing. The target feature information is obtained according to the target basic features and the target engineering features.

[0009] Optionally, the resource scheduling method of the wireless sensor network, wherein the target scheduling parameter prediction model is obtained by creating a scheduling parameter prediction model and optimizing the scheduling parameter prediction model, the target scheduling parameter is output by inputting the target feature information into the target scheduling parameter prediction model, and the resource scheduling scheme is obtained by generating a scheduling scheme according to the target scheduling parameter. The training samples are obtained, and the target training samples are obtained by performing data enhancement processing on the training samples according to a target enhancement mode, wherein the target enhancement mode includes an adaptive Gaussian noise enhancement mode, a similar sample interpolation enhancement mode, and an intelligent disturbance enhancement mode. The scheduling parameter prediction model is obtained by performing model training on the deep feedforward neural network according to the target training samples, and the target optimization strategy is obtained by matching an optimization strategy of the scheduling parameter prediction model, wherein the target optimization strategy includes a loss function optimization strategy and a regularization optimization strategy. The target scheduling parameter prediction model is obtained by optimizing the scheduling parameter prediction model according to the target optimization strategy, and the target scheduling parameter is output by inputting the target feature information into the target scheduling parameter prediction model. The allocation time slot information of the production data priority is obtained, and the resource scheduling scheme is obtained by generating a scheduling scheme for the scheduling task of the target industrial scene according to the target scheduling parameter and the allocation time slot information.

[0010] Optionally, the resource scheduling method of the wireless sensor network, wherein the target training samples are obtained by performing data enhancement processing on the training samples according to a target enhancement mode, and the target enhancement mode includes an adaptive Gaussian noise enhancement mode, a similar sample interpolation enhancement mode, and an intelligent disturbance enhancement mode. The adaptive noise is generated according to the adaptive Gaussian noise enhancement mode, and the training samples are data enhanced by the adaptive noise to obtain a first preset proportion of first enhanced training samples. The Euclidean distance calculation result is obtained by performing Euclidean distance calculation on the training samples according to the similar sample interpolation enhancement mode, and the second preset proportion of second enhanced training samples is obtained by performing interpolation processing on the training samples according to the sample with the nearest Euclidean distance in the Euclidean distance calculation result. The disturbance intensity is obtained by performing parameter sensitivity analysis according to the intelligent disturbance enhancement mode, and the third preset proportion of third enhanced training samples is obtained by performing data disturbance on the training samples according to the disturbance intensity. The first enhanced training sample, the second enhanced training sample and the third enhanced training sample are combined to obtain a target training sample.

[0011] Optionally, the resource scheduling method of the wireless sensor network, wherein the adaptive noise is generated according to the adaptive Gaussian noise enhancement mode, and specifically: ; wherein, the adaptive noise is, the noise level at the first moment, the feature vector of the training sample at the first moment, the noise level at the second moment, the feature vector of the training sample at the second moment, is a normal distribution.

[0012] Optionally, the resource scheduling method of the wireless sensor network, wherein the training sample is interpolated according to the sample with the nearest Euclidean distance in the Euclidean distance calculation result, and specifically: ; wherein, the second enhanced training sample, a random value of 0.1 to 0.9, the feature vector of the training sample at the first moment, the feature vector of the training sample at the second moment.

[0013] Optionally, the resource scheduling method of the wireless sensor network, wherein the scheduling parameter prediction model is optimized according to the target optimization strategy to obtain a target scheduling parameter prediction model, and specifically includes: According to the loss function optimization strategy, the loss value of the scheduling parameter prediction model is calculated to obtain a target loss value, and the parameters of the scheduling parameter prediction model are optimized according to the target loss value to obtain a scheduling parameter prediction optimization model; According to the regularization optimization strategy, the scheduling parameter prediction optimization model is fitted and optimized to obtain a target scheduling parameter prediction model.

[0014] Optionally, the resource scheduling method of the wireless sensor network, wherein the loss value of the scheduling parameter prediction model is calculated according to the loss function optimization strategy, and specifically: ; wherein, the mean square error loss function, a normal distribution, the noise level at the first moment, a real optimal value of a scheduling parameter at a moment, a first moment scheduling parameter prediction model outputs a predicted scheduling parameter.

[0015] Optionally, the resource scheduling method of the wireless sensor network, wherein the fitting optimization processing of the scheduling parameter prediction optimization model according to the regularization optimization strategy is specifically as follows: performing optimizer matching according to the regularization optimization strategy to obtain a target optimizer, and performing parameter configuration on the target optimizer to obtain an optimization parameter configuration result, wherein the optimization parameter configuration result includes an initial learning rate, a learning rate scheduling, a batch size, and a maximum number of training times; performing fitting optimization processing of the scheduling parameter prediction optimization model according to the optimization parameter configuration result to obtain a target scheduling parameter prediction model.

[0016] Optionally, the resource scheduling method of the wireless sensor network, wherein the performance analysis of the resource scheduling scheme to obtain a performance analysis result, and the dynamic optimization of the resource scheduling scheme according to the performance analysis result to obtain a target resource scheduling scheme, and the resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result, specifically includes: obtaining a scheduling task of the target industrial scene, and performing resource scheduling on the scheduling task according to the resource scheduling scheme to obtain a resource scheduling result; analyzing and processing transmission energy consumption of the resource scheduling result to obtain a transmission energy consumption analysis result, analyzing and processing reception energy consumption of the resource scheduling result to obtain a reception energy consumption analysis result, and obtaining a performance analysis result according to the transmission energy consumption analysis result and the reception energy consumption analysis result; performing comparison processing on the performance analysis result and a preset performance index of the target industrial scene to obtain a comparison result, performing dynamic optimization of the resource scheduling scheme according to the comparison result to obtain a target resource scheduling scheme, and performing resource scheduling on the scheduling task according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0017] Optionally, the resource scheduling method of the wireless sensor network, wherein the analyzing and processing of the transmission energy consumption of the resource scheduling result is specifically as follows: ; wherein, is the transmission energy consumption, is the data packet length, is the average distance between nodes, is the circuit energy consumption, Power amplifier energy for a free space model, Power amplifier energy for a multipath fading model, Threshold distance.

[0018] Optionally, the resource scheduling method of the wireless sensor network, wherein the receiving energy consumption of the resource scheduling result is analyzed and processed, and specifically: ; wherein, Receiving energy consumption, Data packet length, Circuit energy consumption.

[0019] Optionally, the resource scheduling method of the wireless sensor network, wherein the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, and then further comprising: Abnormal detection is performed on the target resource scheduling result to obtain an abnormal detection result; If there is an abnormal event in the abnormal detection result, an abnormal analysis is performed on the abnormal event to obtain an abnormal analysis result, and a warning reminder is performed according to the abnormal analysis result.

[0020] In addition, in order to achieve the above-mentioned purpose, the application also provides a resource scheduling system of a wireless sensor network, wherein the resource scheduling system of the wireless sensor network comprises: A feature extraction module is configured to acquire wireless sensor network state information of a target industrial scene, extract features from the wireless sensor network state information to obtain feature information, and perform standardization processing on the feature information to obtain target feature information; A parameter prediction module is configured to create a scheduling parameter prediction model, optimize the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, output target scheduling parameters, and generate a scheduling scheme according to the target scheduling parameters to obtain a resource scheduling scheme; A resource scheduling module is configured to analyze the performance of the resource scheduling scheme to obtain a performance analysis result, dynamically optimize the resource scheduling scheme according to the performance analysis result to obtain a target resource scheduling scheme, and perform resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0021] Optionally, the resource scheduling system of the wireless sensor network, wherein the feature extraction module comprises: An information acquisition unit is configured to acquire wireless sensor network state information of a target industrial scene; The feature extraction unit is configured to perform multi-dimensional feature extraction on the wireless sensor network state information to obtain feature information. The basic feature processing unit is configured to perform screening processing on basic features of the feature information to obtain target basic features. The engineering feature processing unit is configured to perform preprocessing on engineering features of the feature information to obtain target engineering features. The feature merging unit is configured to obtain target feature information according to the target basic features and the target engineering features.

[0022] Optionally, the resource scheduling system of the wireless sensor network comprises: The sample enhancement unit is configured to obtain training samples, perform data enhancement processing on the training samples according to a target enhancement manner, and obtain target training samples. The model training unit is configured to perform model training on a deep feedforward neural network according to the target training samples to obtain a scheduling parameter prediction model, and perform optimization strategy matching on the scheduling parameter prediction model to obtain a target optimization strategy. The model optimization unit is configured to perform optimization processing on the scheduling parameter prediction model according to the target optimization strategy to obtain a target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, and output a target scheduling parameter. The scheme generation unit is configured to obtain allocation time slot information of production data priorities, perform scheduling scheme generation on a scheduling task of the target industrial scene according to the target scheduling parameter and the allocation time slot information, and obtain a resource scheduling scheme.

[0023] Optionally, the resource scheduling system of the wireless sensor network comprises: The preliminary scheduling unit is configured to obtain a scheduling task of the target industrial scene, perform resource scheduling on the scheduling task according to the resource scheduling scheme, and obtain a resource scheduling result. The performance analysis unit is configured to perform analysis processing on transmission energy consumption of the resource scheduling result to obtain a transmission energy consumption analysis result, perform analysis processing on reception energy consumption of the resource scheduling result to obtain a reception energy consumption analysis result, and obtain a performance analysis result according to the transmission energy consumption analysis result and the reception energy consumption analysis result. The scheme optimization unit is configured to perform comparison processing on the performance analysis result and a preset performance indicator of the target industrial scene to obtain a comparison result, perform dynamic optimization on the resource scheduling scheme according to the comparison result to obtain a target resource scheduling scheme, perform resource scheduling on the scheduling task according to the target resource scheduling scheme, and obtain a target resource scheduling result.

[0024] In addition, to achieve the above object, the application further provides a terminal, wherein the terminal comprises a memory, a processor and a resource scheduling program of a wireless sensor network stored in the memory and executable on the processor, and the resource scheduling program of the wireless sensor network realizes the steps of the resource scheduling method of the wireless sensor network when executed by the processor.

[0025] In addition, to achieve the above object, the application further provides a computer readable storage medium, wherein the computer readable storage medium stores a resource scheduling program of a wireless sensor network, and the resource scheduling program of the wireless sensor network realizes the steps of the resource scheduling method of the wireless sensor network when executed by a processor.

[0026] In the application, the state information of a wireless sensor network of a target industrial scene is acquired, feature extraction is performed on the state information of the wireless sensor network to obtain feature information, and standardization processing is performed on the feature information to obtain target feature information; a scheduling parameter prediction model is created, the scheduling parameter prediction model is optimized to obtain a target scheduling parameter prediction model, the target feature information is input into the target scheduling parameter prediction model, target scheduling parameters are output, a scheduling scheme is generated according to the target scheduling parameters, and a resource scheduling scheme is obtained; performance analysis is performed on the resource scheduling scheme to obtain performance analysis results, the resource scheduling scheme is dynamically optimized according to the performance analysis results to obtain a target resource scheduling scheme, resource scheduling is performed according to the target resource scheduling scheme, and a target resource scheduling result is obtained. The scheduling scheme generation time of the application is controlled in the millisecond level, the dynamic response is timely, the safety of instruction execution is ensured, multi-target optimization can be realized, and the resource scheduling efficiency is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of a preferred embodiment of the resource scheduling method of the wireless sensor network of the application; Figure 2 is a schematic diagram of an overall scheme architecture in the preferred embodiment of the resource scheduling method of the wireless sensor network of the application; Figure 3 is a flowchart of the specific implementation process of step S10 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the application; Figure 4 is a flowchart of the specific implementation process of step S20 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the application; Figure 5 is a flowchart of the specific implementation process of step S22 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the application; Figure 6 is a flow chart of the specific implementation process of step S23 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 7 is a flow chart of the specific implementation process of step S232 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 8 is a flow chart of the optimization of the resource scheduling scheme in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 9 is a flow chart of the specific implementation process of step S30 in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 10 is a flow chart of the verification of the optimized parameter configuration in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 11 is a flow chart of the abnormality detection of the target resource scheduling result in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 12 is another flow chart of the abnormality detection of the target resource scheduling result in the preferred embodiment of the resource scheduling method of the wireless sensor network of the present application; Figure 13 is a structural schematic diagram of the resource scheduling system of the wireless sensor network of the present application; Figure 14 is another structural schematic diagram of the resource scheduling system of the wireless sensor network of the present application; Figure 15 is a structural diagram of the preferred embodiment of the terminal of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical scheme and advantages of the present application more clear and definite, the present application is further described in detail below with reference to the drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0029] WSN (Wireless Sensor Networks, wireless sensor network) is widely used in industrial Internet of Things. By deploying nodes in production line equipment, warehouse systems and other parts, key data such as equipment load and vibration are collected to support predictive maintenance and production line scheduling.

[0030] The existing resource scheduling scheme is implemented by TDMA (Time Division Multiple Access) scheduling, but the existing TDMA scheduling industrial WSN system has significant defects, which directly affects the production continuity and operation and maintenance cost, and the specific problems include: 1. High device energy consumption and maintenance cost: industrial sensors are mostly deployed on mechanical arms, AGVs (Automated Guided Vehicle, Automated Guided Vehicle) and other devices, and battery replacement requires downtime. The complexity calculated by the existing algorithm is O(n²), the energy consumption of local calculation of the device accounts for more than 25%, and it is not adapted to load fluctuation, and the node life is less than 6 months, which is far lower than the requirement of 12 months maintenance cycle in industry; 2. High instruction delay, safety risk: in industrial scenarios, device failure warning and other instructions need to be transmitted within 50ms, while the existing algorithm takes 3-6 seconds to generate a scheduling scheme in a 500-node network, which is easy to cause device response lag, leading to quality defects or safety accidents; 3. Poor adaptability and weak matching to dynamic production line: the increase and decrease of nodes in the production line, the movement of AGVs and the electromagnetic interference in the workshop make the network topology and link stability change dynamically. The existing algorithm is designed for fixed topology, and the scheduling conflict rate is more than 35% after the node state changes, resulting in data packet loss and affecting the accuracy of the MES system (a production informatization management system for the execution layer of the manufacturing enterprise workshop); 4. Multi-objective optimization imbalance: industrial monitoring needs to balance "low energy consumption, low delay and high reliability", but the existing method mostly optimizes energy consumption, leading to performance imbalance, for example, an algorithm scheduling system improves the node life by 8%, but the device failure data packet loss rate reaches 10%, which cannot meet the predictive maintenance demand; 5. Insufficient dynamic response: when the production line equipment is under high load (for example, the mechanical arm is running at high speed), the sensor acquisition frequency needs to increase from 10 times / s to 100 times / s, and the traditional algorithm cannot adjust the time slot in real time, leading to congestion of burst data and difficulty in timely uploading of key working condition parameters.

[0031] Therefore, the present application provides a resource scheduling method for a wireless sensor network to overcome the problems of high instruction delay, multi-objective optimization imbalance and insufficient dynamic response in the prior art.

[0032] The resource scheduling method for a wireless sensor network according to the preferred embodiment of the present application comprises the following steps: Figure 1 As shown in the accompanying drawings, the resource scheduling method for a wireless sensor network comprises the following steps: Step S10, obtaining the state information of the wireless sensor network of the target industrial scene, extracting features from the state information of the wireless sensor network, obtaining feature information, and standardizing the feature information to obtain target feature information.

[0033] Specifically, the overall scheme for implementing the resource scheduling method for a wireless sensor network is as shown in the accompanying drawings Figure 2As shown, the corresponding system architecture includes a data acquisition module, a feature engineering module, a deep learning prediction module, a scheduling optimization module, and a performance monitoring module. The data acquisition module is used to collect network topology, node energy, communication demand, and other information. The feature engineering module is used to extract and construct network features. The deep learning prediction module is used for scheduling parameter prediction based on deep neural networks. The scheduling optimization module is used to generate an optimal TDMA scheduling scheme. The performance monitoring module is used to monitor network performance in real time and provide feedback for optimization.

[0034] First, the wireless sensor network state information of the target industrial scene is obtained, wherein the wireless sensor network state information includes industrial network scale (i.e., the number of network nodes), inter-node distance, data packet length, node initial energy, and production data priority.

[0035] Second, multi-dimensional feature extraction is performed on the wireless sensor network state information to obtain feature information, wherein the feature information includes basic features and engineering features. The basic features include network scale features, spatial features, communication features, and energy features. The engineering features include network density features, priority interaction features, energy density features, and complexity features.

[0036] Finally, the feature information needs to be standardized to obtain target feature information. Specifically, the basic features of the feature information are filtered to obtain target basic features, wherein the filtering includes feature verification processing and redundant feature elimination processing. The engineering features of the feature information are preprocessed to obtain target engineering features, wherein the preprocessing includes missing value processing and outlier processing. The target feature information is obtained based on the target basic features and the target engineering features.

[0037] As shown in Figure 3 , the step S10 specifically includes: Step S11, obtaining wireless sensor network state information of a target industrial scene, wherein the wireless sensor network state information includes industrial network scale, inter-node distance, data packet length, node initial energy, and production data priority.

[0038] Specifically, step S11 is implemented as: Obtaining wireless sensor network state information of a target industrial scene, wherein the wireless sensor network state information includes industrial network scale (i.e., the number of network nodes), inter-node distance (the actual communication distance after correction of workshop electromagnetic interference), data packet length (dynamically adjusted according to the type of production data, with increased fault data length), node initial energy (Battery parameters adapted to high-temperature workshop environments) and production data priority (When the equipment load exceeds 80% or the vibration exceeds the standard, then) Otherwise, if the equipment load does not exceed 80% or the vibration does not exceed the standard, ).

[0039] Step S12: Perform multi-dimensional feature extraction on the wireless sensor network state information to obtain feature information, wherein the feature information includes basic features and engineering features. The basic features include network scale features, spatial features, communication features and energy features. The engineering features include network density features, priority interaction features, energy density features and complexity features.

[0040] Specifically, step S12 is implemented as follows: Feature extraction is required for the wireless sensor network status information. In this embodiment of the invention, a multi-dimensional feature extraction method is designed. Specifically, multi-dimensional feature extraction is performed on the wireless sensor network status information to obtain feature information. The feature information includes basic features and engineering features. The basic features include network scale features (i.e., the total number of nodes in the workshop monitoring area), spatial features (i.e., the average distance between nodes, combined with workshop electromagnetic interference correction to exclude invalid distances caused by metal obstruction), communication features (i.e., data packet length features, with fault warning data packets being 5000 bits long and regular production data packets being 800 bits long), and energy features (i.e., the initial energy of the nodes). The characteristics include: initial node energy of 1.2 joules in high-temperature environments and 0.6 joules in normal-temperature environments; network density characteristics (i.e., node density ratio, which is the ratio of the number of network nodes to the effective production area of ​​the workshop); priority interaction characteristics (i.e., scale-distance-priority interaction items, highlighting the scheduling weight of high-priority devices); energy density characteristics (i.e., average node energy, reflecting the workshop network's endurance, which is the ratio of the node's initial energy to the number of network nodes); and complexity characteristics (i.e., communication complexity factor, which is larger for higher priority and closer distance, requiring priority scheduling).

[0041] Step S13: Filter the basic features of the feature information to obtain the target basic features. The filtering process includes feature verification and redundant feature removal.

[0042] Specifically, step S13 is implemented as follows: The feature information needs to be standardized to obtain target feature information. Specifically, the basic features of the feature information are filtered to obtain target basic features. The filtering process includes feature verification and redundant feature removal.

[0043] For feature checking processing: (1) Dimension consistency: all features have no units, and the numerical range meets the selected method (for example, Min-Max: 0-1); (2) Feature discrimination: the original core information is not lost; (3) Application adaptability: the subsequent task effect meets the standard (for example, the fault nodes are correctly grouped in clustering, and the convergence speed of the optimization algorithm is improved).

[0044] For redundant feature elimination processing: (1) Redundant feature identification: redundant features are identified through "business logic + data indicators" to avoid misjudgment of pure data-driven; (2) Redundant feature influence evaluation: evaluate the "irreplaceability" of redundant features to avoid eliminating "seemingly redundant but unique business value" features; (3) Redundant feature elimination execution: according to the principle of "business priority, data assistance", the redundant features are finally eliminated.

[0045] Step S14, preprocessing the engineering features of the feature information to obtain target engineering features, wherein the preprocessing includes missing value processing and abnormal value processing.

[0046] Among them, step S14 is specifically implemented as: Because WSN data is easily affected by communication packet loss, sensor failure, and environmental interference, there are missing values and abnormal values, which need to be processed first, otherwise it will cause standardization distortion, so the engineering features of the feature information are preprocessed to obtain target engineering features, wherein the preprocessing includes missing value processing and abnormal value processing.

[0047] For missing value processing: (1) Mean filling: suitable for uniform feature distribution (for example, communication success rate), replace missing values with the mean of all valid samples of the feature; (2) Median filling: suitable for skewed distribution features (for example, residual energy, some nodes may have very low energy consumption due to high load); (3) KNN filling (K-Nearest Neighbors Imputation, a similarity-based missing value filling method): use the spatial correlation of WSN nodes (i.e. similar state of adjacent nodes), select the k nearest normal nodes, and fill them with weighted average.

[0048] For abnormal value processing: (1) 3σ principle: if the feature is approximately normally distributed (for example, communication delay), remove values outside [μ 3σ, μ + 3σ] (μ is the mean, σ is the standard deviation); (2) IQR method (Interquartile Range): suitable for non-normal distribution (for example, residual energy), remove values outside [Q1 1.5IQR, Q3 + 1.5IQR] (Q1 is the first quartile, Q3 is the third quartile, IQR = Q3-Q1); (3) Physical constraint correction: remove according to the physical meaning of WSN (for example, residual energy cannot be negative, communication success rate cannot exceed 100%).

[0049] Step S15, obtaining target feature information according to the target basic feature and the target engineering feature.

[0050] Among them, step S15 is specifically implemented as: The target network size feature, the target space feature, the target communication feature, and the target energy feature, the target network density feature, the target priority interaction feature, the target energy density feature, and the target complexity feature are combined and processed to obtain target feature information.

[0051] Step S20, creating a scheduling parameter prediction model, optimizing the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, inputting the target feature information into the target scheduling parameter prediction model, outputting a target scheduling parameter, and generating a scheduling scheme according to the target scheduling parameter to obtain a resource scheduling scheme.

[0052] Specifically, a training sample is acquired, data augmentation processing is performed on the training sample according to a target augmentation manner, and a target training sample is obtained, wherein the target augmentation manner includes an adaptive Gaussian noise augmentation manner, a similar sample interpolation augmentation manner and an intelligent disturbance augmentation manner; a deep feedforward neural network is trained according to the target training sample, and a scheduling parameter prediction model is obtained, wherein the network architecture of the deep feedforward neural network includes an input layer (8 neurons), 5 hidden layers (a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and a fifth hidden layer, wherein the first hidden layer contains 128 neurons, the second hidden layer contains 96 neurons, the third hidden layer contains 64 neurons, the fourth hidden layer contains 32 neurons, and the fifth hidden layer contains 16 neurons, each hidden layer is equipped with batch normalization, an activation function and a Dropout layer, wherein Dropout is defined as a regularization technique for reducing overfitting by randomly turning off part of the neurons, and a probability value represents the proportion of neurons randomly discarded in the layer (for example, 0.4 means that 40% of the neurons do not participate in the current training), and an output layer (1 neuron outputs scheduling parameters); the scheduling parameter prediction model is matched with an optimization strategy to obtain a target optimization strategy, wherein the target optimization strategy includes a loss function optimization strategy and a regularization optimization strategy; the scheduling parameter prediction model is optimized according to the target optimization strategy to obtain a target scheduling parameter prediction model, and the target feature information is input into the target scheduling parameter prediction model to output a target scheduling parameter; the allocation time slot information of the production data priority is acquired, and a scheduling scheme of the scheduling task of the target industrial scene is generated according to the target scheduling parameter and the allocation time slot information to obtain a resource scheduling scheme.

[0053] The application proposes an adaptive data augmentation method based on parameter sensitivity, combines an adaptive Gaussian noise augmentation manner, a similar sample interpolation augmentation manner and an intelligent disturbance augmentation manner, and the data augmentation ratio reaches 30%, which significantly improves the model generalization capability. A 6-layer deep neural network is also designed, the parameter quantity is about 15000, a hierarchical decreasing Dropout strategy (0.4→0.2) is adopted, and multiple regularization techniques such as L2 regularization, batch normalization and gradient clipping are integrated to improve the model prediction efficiency.

[0054] As shown in Figure 4 the step S20 specifically includes: Step S21, a training sample is acquired, data augmentation processing is performed on the training sample according to a target augmentation manner, and a target training sample is obtained, wherein the target augmentation manner includes an adaptive Gaussian noise augmentation manner, a similar sample interpolation augmentation manner and an intelligent disturbance augmentation manner.

[0055] As shown in Figure 5As shown, step S21 specifically includes: Step S211, generating adaptive noise according to the adaptive Gaussian noise enhancement mode, and performing data enhancement on the training sample by the adaptive noise to obtain a first preset proportion of first enhanced training samples.

[0056] Wherein, step S211 is specifically implemented as: In order to improve the model generalization ability, three data enhancement modes are adopted, which are adaptive Gaussian noise enhancement mode, similar sample interpolation enhancement mode and intelligent disturbance enhancement mode.

[0057] For the adaptive Gaussian noise enhancement mode: adaptive noise is generated according to the adaptive Gaussian noise enhancement mode, and the adaptive noise is used to perform data enhancement on the training sample to obtain a first preset proportion (for example, 40%) of first enhanced training samples, and the expression for is: ; Wherein, is the adaptive noise, is the noise level at the first moment, is the feature vector of the training sample at the first moment, is the feature vector of the training sample at the second moment. Step S212, performing Euclidean distance calculation on the training sample according to the similar sample interpolation enhancement mode to obtain Euclidean distance calculation result, and performing interpolation processing on the training sample according to the sample with the nearest Euclidean distance in the Euclidean distance calculation result to obtain a second preset proportion of second enhanced training samples.

[0058] Wherein, step S212 is specifically implemented as: For the similar sample interpolation enhancement mode: the Euclidean distance calculation is performed on the training sample according to the similar sample interpolation enhancement mode to obtain the Euclidean distance calculation result, and the interpolation processing is performed on the training sample according to the sample with the nearest Euclidean distance in the Euclidean distance calculation result to obtain a second preset proportion (for example, 30%) of second enhanced training samples, and the expression for is:

[0059] ; Wherein, is the second enhanced training sample, is a random value of 0.1 to 0.9, is the feature vector of the training sample at the first moment, is the feature vector of the training sample at the second moment.

[0060] ​​Step S213: Perform parameter sensitivity analysis according to the intelligent perturbation enhancement method to obtain the perturbation intensity, and perform data perturbation on the training samples according to the perturbation intensity to obtain a third enhanced training sample with a third preset ratio.

[0061] Specifically, step S213 is implemented as follows: For the intelligent perturbation enhancement method: perform parameter sensitivity analysis according to the intelligent perturbation enhancement method to obtain the perturbation intensity, and perturb the training samples according to the perturbation intensity to obtain a third enhanced training sample with a third preset ratio (e.g., 30%).

[0062] Step S214: Combine the first enhanced training sample, the second enhanced training sample, and the third enhanced training sample to obtain the target training sample.

[0063] Specifically, step S213 is implemented as follows: After obtaining the first, second, and third enhanced training samples, the first, second, and third enhanced training samples are combined to obtain the target training sample.

[0064] Step S22: Train the deep feedforward neural network model based on the target training samples to obtain a scheduling parameter prediction model, and perform optimization strategy matching on the scheduling parameter prediction model to obtain a target optimization strategy, wherein the target optimization strategy includes a loss function optimization strategy and a regularization optimization strategy.

[0065] Specifically, step S22 is implemented as follows: The deep feedforward neural network is trained based on the target training samples to obtain a scheduling parameter prediction model. Specifically, the training process involves inputting a set of 8-dimensional features from the target training samples into the deep feedforward neural network; the deep feedforward neural network performs fusion processing on the 8-dimensional features to obtain target fused features, and then predicts parameters based on these target fused features to obtain predicted scheduling parameters; subsequently, the deep feedforward neural network is corrected based on the actual scheduling parameters corresponding to the predicted scheduling parameters; finally, the next set of 8-dimensional features is input into the deep feedforward neural network, and the above process is repeated (details omitted here); this continues until the training status of the deep feedforward neural network meets a preset condition, which is the number of training iterations. This preset condition can be determined based on the accuracy of the scheduling parameter prediction model (details omitted here). The preset number of iterations can be the maximum number of training iterations for the deep feedforward neural network, for example, 2000 iterations, etc., to finally obtain a trained scheduling parameter prediction model.

[0066] After obtaining the scheduling prediction model, it is necessary to optimize the scheduling prediction model. This requires first matching the best optimization strategy, i.e., the target optimization strategy. Specifically, the scheduling parameter prediction model is matched with optimization strategies to obtain the target optimization strategy, which includes a loss function optimization strategy and a regularization optimization strategy.

[0067] Step S23: Optimize the scheduling parameter prediction model according to the target optimization strategy to obtain the target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, and output the target scheduling parameters.

[0068] Among them, such as Figure 6 As shown, step S23 specifically includes: Step S231: Calculate the loss value of the scheduling parameter prediction model according to the loss function optimization strategy to obtain the target loss value, and optimize the parameters of the scheduling parameter prediction model according to the target loss value to obtain the scheduling parameter prediction optimization model.

[0069] Specifically, step S231 is implemented as follows: After matching the optimal target optimization strategy, the scheduling parameter prediction model is optimized according to the loss function optimization strategy and the regularization optimization strategy in the target optimization strategy. Specifically, for the loss function optimization strategy: the loss value of the scheduling parameter prediction model is calculated according to the loss function optimization strategy to obtain the target loss value, and the expression for it is: ; in, Let the mean squared error loss function be . It follows a normal distribution. For the first The true optimal value of the scheduling parameters at each time point. For the first The predicted scheduling parameters are output by the scheduling parameter prediction model at each time point; the parameters of the scheduling parameter prediction model are optimized based on the target loss value to obtain the optimized scheduling parameter prediction model.

[0070] Step S232: Perform fitting optimization processing on the scheduling parameter prediction optimization model according to the regularization optimization strategy to obtain the target scheduling parameter prediction model.

[0071] Specifically, regarding the regularization optimization strategy: the regularization optimization strategy is implemented by using regularization techniques, including L2 regularization and hierarchical Dropout.

[0072] For L2 regularization is a weight parameter of the L2 regularization term, used to balance the model fitting ability and the overfitting risk, by imposing a penalty on the neural network weight parameters, to suppress the model overfitting caused by too large weights, and to ensure that the model adapts to the scene data distribution of different industrial workshops.

[0073] For hierarchical Dropout, 0.4→0.35→0.3→0.25→0.2, this "0.4→0.35→0.3→0.25→0.2" is a process of decreasing the Dropout probability of the five hidden layers in the deep neural network of the application, and the core role is to suppress overfitting while ensuring the model's feature learning ability and adapting to the industrial gateway's computing power needs. Dropout is defined as a regularization technique that reduces overfitting by randomly turning off some neurons, and the probability value represents the proportion of neurons that are randomly discarded (for example, 0.4 means that 40% of the neurons do not participate in this round of training). According to the regularization optimization strategy, the scheduling parameter prediction optimization model is fitted and optimized to obtain a target scheduling parameter prediction model.

[0074] As shown in Figure 7 the step S232 specifically includes: Step S2321, according to the regularization optimization strategy, an optimizer is matched to obtain a target optimizer, and the target optimizer is parameter configured to obtain an optimization parameter configuration result.

[0075] wherein, the step S2321 is specifically implemented as: According to the regularization optimization strategy, the scheduling parameter prediction optimization model is fitted and optimized to obtain a target scheduling parameter prediction model, specifically, according to the regularization optimization strategy, an optimizer is matched to obtain a target optimizer, and the target optimizer is parameter configured to obtain an optimization parameter configuration result, wherein the optimization parameter configuration result includes an initial learning rate (for example, 0.005), a learning rate schedule (decreased by 20% every 15 rounds), a batch size (for example, 128; the batch size is a key hyperparameter of the optimizer when training a machine learning model, which defines the number of training samples used in each parameter update step, and the batch size determines how much data the optimizer processes at a time when calculating gradients and adjusting model weights), and a maximum number of training times (for example, 80).

[0076] Step S2322, according to the optimization parameter configuration result, the scheduling parameter prediction optimization model is fitted and optimized to obtain a target scheduling parameter prediction model.

[0077] wherein, the step S2321 is specifically implemented as: The scheduling parameter prediction optimization model is fitted and optimized based on the optimization parameter configuration results to obtain the target scheduling parameter prediction model; then, the target feature information is input into the target scheduling parameter prediction model to output the target scheduling parameters.

[0078] Step S24: Obtain the allocation time slot information of the production data priority, and generate a scheduling scheme for the scheduling task of the target industrial scenario based on the target scheduling parameters and the allocation time slot information to obtain a resource scheduling scheme.

[0079] Specifically, step S24 is implemented as follows: After obtaining the target scheduling parameters, a corresponding resource scheduling scheme needs to be generated. Specifically, the allocation time slot information of the production data priority is obtained, and a scheduling scheme is generated for the scheduling tasks of the target industrial scenario based on the target scheduling parameters and the allocation time slot information to obtain the resource scheduling scheme. This resource scheduling scheme can shorten the time slot interval of high-priority nodes by 50%.

[0080] The corresponding advantages are as follows: (1) Computational efficiency: The time complexity is reduced from Reduce to ; (2) Real-time performance: Prediction time is reduced from seconds to milliseconds; (3) Generalization capability: Supports networks of different sizes and topologies; (4) Accuracy: The accuracy rate is over 90% within the ±1 error range; (5) Scalability: Supports network scale of 15 to 700 nodes.

[0081] Step S30: Perform performance analysis on the resource scheduling scheme to obtain performance analysis results. Based on the performance analysis results, dynamically optimize the resource scheduling scheme to obtain a target resource scheduling scheme. Then, perform resource scheduling based on the target resource scheduling scheme to obtain the target resource scheduling result.

[0082] Specifically, after obtaining the resource scheduling plan, it is necessary to determine whether this resource scheduling plan is the optimal scheduling plan, such as... Figure 8 As shown, resource scheduling is performed through the resource scheduling scheme to obtain resource scheduling results, and performance analysis is performed on the resource scheduling results to obtain performance analysis results; it is determined whether the performance analysis results meet the preset performance indicators. If the performance analysis results do not meet the preset performance indicators, the resource scheduling scheme is dynamically optimized based on the performance analysis results to obtain the target resource scheduling scheme.

[0083] Specifically, a scheduling task of the target industrial scene is acquired, and resource scheduling is performed on the scheduling task according to the resource scheduling scheme to obtain a resource scheduling result; transmission energy consumption of the resource scheduling result is analyzed and processed to obtain transmission energy consumption analysis result, reception energy consumption of the resource scheduling result is analyzed and processed to obtain reception energy consumption analysis result, and performance analysis result is obtained according to the transmission energy consumption analysis result and the reception energy consumption analysis result; the performance analysis result is compared with a preset performance index of the target industrial scene to obtain a comparison result, the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, and resource scheduling is performed on the scheduling task according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0084] As shown in Figure 9 , the step S30 specifically includes: Step S31, a scheduling task of the target industrial scene is acquired, and resource scheduling is performed on the scheduling task according to the resource scheduling scheme to obtain a resource scheduling result.

[0085] In which, step S31 is specifically implemented as: After obtaining the resource scheduling scheme, resource scheduling needs to be performed according to the resource scheduling scheme first, so as to perform performance analysis subsequently, specifically, a scheduling task of the target industrial scene is acquired, and resource scheduling is performed on the scheduling task according to the resource scheduling scheme to obtain a resource scheduling result.

[0086] Step S32, transmission energy consumption of the resource scheduling result is analyzed and processed to obtain transmission energy consumption analysis result, reception energy consumption of the resource scheduling result is analyzed and processed to obtain reception energy consumption analysis result, and performance analysis result is obtained according to the transmission energy consumption analysis result and the reception energy consumption analysis result.

[0087] In which, step S32 is specifically implemented as: After obtaining the resource scheduling result, energy consumption analysis needs to be performed on the resource scheduling result to analyze whether the current resource scheduling scheme is the best scheduling scheme, specifically, transmission energy consumption of the resource scheduling result is analyzed and processed to obtain transmission energy consumption analysis result, and the corresponding expression is: ; In which, is the transmission energy consumption, is the data packet length, is the average distance between nodes, is the circuit energy consumption, is the power amplifier energy of the free space model, is the power amplifier energy of the multipath fading model, is a threshold distance.

[0088] The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result, and the corresponding expression is: ; wherein, is the receiving energy consumption, is the data packet length, is the circuit energy consumption; then, the performance analysis result is obtained according to the transmission energy consumption analysis result and the receiving energy consumption analysis result.

[0089] In step S33, the performance analysis result is compared with a preset performance index of the target industrial scene to obtain a comparison result, the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, and the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0090] In step S33, the performance analysis result is compared with a preset performance index of the target industrial scene to obtain a comparison result, the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, and the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result. After obtaining the performance analysis result, it is necessary to judge whether the resource scheduling scheme needs to be optimized, specifically, the performance analysis result is compared with a preset performance index of the target industrial scene to obtain a comparison result, if the performance analysis result does not exceed the preset performance index of the target industrial scene, the resource scheduling scheme does not need to be optimized; if the performance analysis result exceeds the preset performance index of the target industrial scene, the resource scheduling scheme needs to be optimized, and the corresponding optimization process is that the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, and the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0091] Through experimental verification, the present application has the following improvements compared with the prior art: (1) Calculation performance: Calculation time: compared with the second level of the prior art, it is improved to the millisecond level; Time complexity: compared with the prior art , it is improved to ; Memory consumption: compared with the prior art, it is reduced by 80%; (2) Prediction accuracy: Mean absolute error: compared with the prior art, it is reduced to 2.0; Root mean square error: compared with the prior art, it is reduced to 3.0; R² determination coefficient: compared with the prior art, it is improved to 0.85; Accuracy: improved to 90% compared with prior art; (3) Network performance: Network lifetime: improved by 15-25% compared with prior art; Energy balance: improved by 20-30% compared with prior art; Communication delay: reduced by 10-15% compared with prior art; (4) Applicable scope: Network size: 15-700 nodes; Node distribution: 1.5-35 meters apart; Data packet length: 500-5000 bits; Initial energy: 0.1-1.0 joules.

[0092] Further, as shown in Figure 10 According to the regularization optimization strategy, the optimizer matching is optimized to obtain the target optimizer, and the target optimizer is parameter configured to obtain the optimization parameter configuration result, and then further includes: Step S41, verifying the optimization parameter configuration result to obtain a parameter verification result; Step S42, if the parameter verification result has an abnormal configuration parameter, marking the abnormal configuration parameter to obtain a marking result, and adjusting the abnormal configuration parameter according to the marking result.

[0093] Specifically, after obtaining the optimization parameter configuration result, it is necessary to verify whether the configuration is successful, specifically, verifying the optimization parameter configuration result to obtain a parameter verification result; if the parameter verification result has an abnormal configuration parameter, marking the abnormal configuration parameter to obtain a marking result, and adjusting the abnormal configuration parameter according to the marking result.

[0094] Further, as shown in Figure 11 According to the target resource scheduling scheme, the scheduling task is resource scheduled to obtain a target resource scheduling result, and then further includes: Step S51, detecting the target resource scheduling result for abnormalities to obtain an abnormality detection result; Step S52, if the abnormality detection result has an abnormal event, analyzing the abnormal event for abnormalities to obtain an abnormality analysis result, and warning according to the abnormality analysis result.

[0095] Specifically, after the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result, the target resource scheduling result needs to be abnormally detected, specifically, the target resource scheduling result is abnormally detected to obtain an abnormal detection result, if there is an abnormal event in the abnormal detection result, the abnormal event is abnormally analyzed to obtain an abnormal analysis result, and a warning is reminded according to the abnormal analysis result, so as to timely handle the abnormal situation.

[0096] Further, as shown in the figure, Figure 12 After the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result, the following steps are further included: Step S61, the target resource scheduling result is abnormally detected to obtain an abnormal detection result; Step S62, if there is no abnormal event in the abnormal detection result, an abnormal detection period is set, and the target resource scheduling result is abnormally detected according to the abnormal detection period.

[0097] Specifically, after the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result, the target resource scheduling result needs to be abnormally detected, specifically, the target resource scheduling result is abnormally detected to obtain an abnormal detection result, if there is no abnormal event in the abnormal detection result, an abnormal detection period is set, and the target resource scheduling result is abnormally detected according to the abnormal detection period, and the purpose of setting the abnormal detection period is to detect whether there is an abnormal situation in the resource scheduling process in real time, so as to ensure the normal operation of the whole resource scheduling process.

[0098] The technical effects that can be brought by the application are as follows: (1) Reduce the energy consumption of equipment: through The scheduling method with low computational complexity reduces the energy consumption of edge devices, and the optimized transmission energy consumption model prolongs the life of industrial sensor nodes to more than 12 months, matching the production and maintenance cycle.

[0099] (2) Real-time response: the scheduling scheme generation time is controlled within milliseconds to ensure that critical instructions such as device fault warning and emergency shutdown are uploaded to the execution unit within 50ms, meeting the production safety requirements.

[0100] (3) Improve environmental adaptability: the algorithm can adapt to scenarios such as increase or decrease of production line nodes, device movement, and load fluctuation, without manual parameter adjustment, and the scheduling conflict rate is controlled below 5%.

[0101] (4) Multi-objective optimization is realized: the device endurance time is improved, the control instruction delay is reduced, the production data reliability is ensured, and the core requirements of industrial scenes are met.

[0102] (5) Large-scale deployment is supported: the industrial network scale of 15-800 nodes can be adapted, and an extensible solution is provided for large intelligent manufacturing plants, smart warehouses and other scenes.

[0103] Further, as shown in Figure 13 the resource scheduling method of the wireless sensor network, the application also correspondingly provides a resource scheduling system of a wireless sensor network, wherein the resource scheduling system of the wireless sensor network comprises: a feature extraction module 50, configured to acquire wireless sensor network state information of a target industrial scene, perform feature extraction on the wireless sensor network state information to obtain feature information, and perform standardization processing on the feature information to obtain target feature information; a parameter prediction module 60, configured to create a scheduling parameter prediction model, perform optimization processing on the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, output a target scheduling parameter, and perform scheduling scheme generation according to the target scheduling parameter to obtain a resource scheduling scheme; a resource scheduling module 70, configured to perform performance analysis on the resource scheduling scheme to obtain a performance analysis result, perform dynamic optimization on the resource scheduling scheme according to the performance analysis result to obtain a target resource scheduling scheme, and perform resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0104] As shown in Figure 14 another embodiment of the resource scheduling system of the wireless sensor network in the embodiment of the application, in this embodiment, the feature extraction module 50 comprises: an information acquisition unit 501, configured to acquire wireless sensor network state information of a target industrial scene; a feature extraction unit 502, configured to perform multi-dimensional feature extraction on the wireless sensor network state information to obtain feature information; a basic feature processing unit 503, configured to perform screening processing on the basic features of the feature information to obtain target basic features; an engineering feature processing unit 504, configured to perform preprocessing on the engineering features of the feature information to obtain target engineering features; a feature merging unit 505, configured to obtain target feature information according to the target basic features and the target engineering features.

[0105] In this embodiment, the parameter prediction module 60 comprises: The sample enhancement unit 601 is configured to obtain a training sample, perform data enhancement processing on the training sample according to a target enhancement manner, and obtain a target training sample. The model training unit 602 is configured to perform model training on a deep feedforward neural network according to the target training sample, obtain a scheduling parameter prediction model, and perform optimization strategy matching on the scheduling parameter prediction model, to obtain a target optimization strategy. The model optimization unit 603 is configured to perform optimization processing on the scheduling parameter prediction model according to the target optimization strategy, obtain a target scheduling parameter prediction model, and input the target feature information into the target scheduling parameter prediction model, to output a target scheduling parameter. The scheme generation unit 604 is configured to obtain allocation time slot information of a production data priority, perform scheduling scheme generation on a scheduling task of the target industrial scene according to the target scheduling parameter and the allocation time slot information, and obtain a resource scheduling scheme.

[0106] In this embodiment, the resource scheduling module 70 includes: The preliminary scheduling unit 701 is configured to obtain a scheduling task of the target industrial scene, and perform resource scheduling on the scheduling task according to the resource scheduling scheme, to obtain a resource scheduling result. The performance analysis unit 702 is configured to perform analysis processing on transmission energy consumption of the resource scheduling result, to obtain a transmission energy consumption analysis result, perform analysis processing on reception energy consumption of the resource scheduling result, to obtain a reception energy consumption analysis result, and obtain a performance analysis result according to the transmission energy consumption analysis result and the reception energy consumption analysis result. The scheme optimization unit 703 is configured to perform comparison processing on the performance analysis result and a preset performance index of the target industrial scene, to obtain a comparison result, perform dynamic optimization on the resource scheduling scheme according to the comparison result, to obtain a target resource scheduling scheme, and perform resource scheduling on the scheduling task according to the target resource scheduling scheme, to obtain a target resource scheduling result.

[0107] The scheduling scheme generation time of the present application is controlled in the millisecond level, the dynamic response is timely, the safety of instruction execution is guaranteed, multi-target optimization can be realized, and the resource scheduling efficiency is greatly improved.

[0108] Further, as shown in Figure 15 Based on the above resource scheduling method and system of the wireless sensor network, the present application further provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 15 Only part of the components of the terminal are shown, but it should be understood that all the shown components are not required to be implemented, and more or less components can be alternatively implemented.

[0109] The memory 20 can be an internal storage unit of the terminal in some embodiments, such as a hard disk or a memory of the terminal. The memory 20 can also be an external storage device of the terminal in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal. Further, the memory 20 can include both the internal storage unit and the external storage device of the terminal. The memory 20 is used to store application software and various data installed on the terminal, such as program codes of the terminal, etc. The memory 20 can also be used to temporarily store data that has been output or will be output. In an embodiment, the memory 20 stores a wireless sensor network resource scheduling program 40, which can be executed by the processor 10 to implement the wireless sensor network resource scheduling method in the present application.

[0110] The processor 10 can be a central processing unit (CPU), a microprocessor or other data processing chip in some embodiments, which is used to run program codes or process data stored in the memory 20, such as to execute the wireless sensor network resource scheduling method, etc.

[0111] The display 30 can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. in some embodiments. The display 30 is used to display information of the terminal and to display a visual user interface. The processor 10, the memory 20 and the display 30 of the terminal communicate with each other through a system bus.

[0112] In an embodiment, the following steps are implemented when the processor 10 executes the wireless sensor network resource scheduling program 40 in the memory 20: Obtaining wireless sensor network state information of a target industrial scene, performing feature extraction on the wireless sensor network state information to obtain feature information, and performing standardization processing on the feature information to obtain target feature information; Creating a scheduling parameter prediction model, performing optimization processing on the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, inputting the target feature information into the target scheduling parameter prediction model, outputting target scheduling parameters, and generating a scheduling scheme according to the target scheduling parameters to obtain a resource scheduling scheme; Performance analysis is performed on the resource scheduling scheme to obtain performance analysis results, the resource scheduling scheme is dynamically optimized according to the performance analysis results to obtain a target resource scheduling scheme, and resource scheduling is performed according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0113] The wireless sensor network state information of the target industrial scene is obtained, and feature information is obtained by performing feature extraction on the wireless sensor network state information, specifically including: The wireless sensor network state information of the target industrial scene is obtained, and feature information is obtained by performing feature extraction on the wireless sensor network state information, specifically including: The wireless sensor network state information of the target industrial scene is obtained, and feature information is obtained by performing feature extraction on the wireless sensor network state information, specifically including:

[0114] The feature information is standardized to obtain target feature information, specifically including: The feature information is standardized to obtain target feature information, specifically including: The feature information is standardized to obtain target feature information, specifically including: The feature information is standardized to obtain target feature information, specifically including:

[0115] The feature information is standardized to obtain target feature information, specifically including: The training sample is obtained, and the training sample is data enhanced according to a target enhancement mode to obtain a target training sample, wherein the target enhancement mode includes an adaptive Gaussian noise enhancement mode, a similar sample interpolation enhancement mode, and an intelligent disturbance enhancement mode. The feature information is standardized to obtain target feature information, specifically including: According to the target optimization strategy, the scheduling parameter prediction model is optimized to obtain a target scheduling parameter prediction model, and the target feature information is input into the target scheduling parameter prediction model to output a target scheduling parameter. obtain the allocation time slot information of the production data priority, and generate a scheduling scheme for the scheduling task of the target industrial scene according to the target scheduling parameter and the allocation time slot information to obtain a resource scheduling scheme.

[0116] According to the target enhancement mode, the training sample is data enhanced to obtain a target training sample, specifically including: According to the adaptive Gaussian noise enhancement mode, adaptive noise is generated, and the adaptive noise is used for data enhancement of the training sample to obtain a first enhanced training sample with a first preset proportion; According to the similar sample interpolation enhancement mode, the training sample is subjected to Euclidean distance calculation to obtain a Euclidean distance calculation result, and the training sample is subjected to interpolation processing according to the sample with the nearest Euclidean distance in the Euclidean distance calculation result to obtain a second enhanced training sample with a second preset proportion; According to the intelligent perturbation enhancement mode, parameter sensitivity analysis is performed to obtain a perturbation intensity, and the training sample is subjected to data perturbation according to the perturbation intensity to obtain a third enhanced training sample with a third preset proportion; The first enhanced training sample, the second enhanced training sample and the third enhanced training sample are combined to obtain a target training sample.

[0117] According to the adaptive Gaussian noise enhancement mode, adaptive noise is generated, specifically as follows: ; wherein, is adaptive noise, is a noise level at the first moment, is a noise level at the second moment, is a feature vector of the training sample at the first moment, is a feature vector of the training sample at the second moment, and is a normal distribution.

[0118] According to the adaptive Gaussian noise enhancement mode, adaptive noise is generated, specifically as follows: ; wherein, is a second enhanced training sample, is a random value of 0.1 to 0.9, is a feature vector of the training sample at the first moment, is a feature vector of the training sample at the second moment.

[0119] Specifically, optimizing the scheduling parameter prediction model according to the target optimization strategy to obtain the target scheduling parameter prediction model includes: The loss value of the scheduling parameter prediction model is calculated according to the loss function optimization strategy to obtain the target loss value, and the parameters of the scheduling parameter prediction model are optimized according to the target loss value to obtain the scheduling parameter prediction optimization model. The scheduling parameter prediction optimization model is fitted and optimized according to the regularization optimization strategy to obtain the target scheduling parameter prediction model.

[0120] Specifically, the step of calculating the loss value of the scheduling parameter prediction model based on the loss function optimization strategy includes: ; in, Let the mean squared error loss function be . It follows a normal distribution. For the first The true optimal value of the scheduling parameters at each moment. For the first The predicted scheduling parameters output by the scheduling parameter prediction model at each time point.

[0121] Specifically, the step of fitting and optimizing the scheduling parameter prediction optimization model according to the regularization optimization strategy to obtain the target scheduling parameter prediction model includes: Optimizer matching is performed according to the regularization optimization strategy to obtain the target optimizer, and the parameters of the target optimizer are configured to obtain the optimization parameter configuration result, wherein the optimization parameter configuration result includes the initial learning rate, learning rate scheduling, batch size and maximum number of training iterations; The scheduling parameter prediction optimization model is fitted and optimized based on the optimization parameter configuration results to obtain the target scheduling parameter prediction model.

[0122] Specifically, the process of performing performance analysis on the resource scheduling scheme to obtain performance analysis results, dynamically optimizing the resource scheduling scheme based on the performance analysis results to obtain a target resource scheduling scheme, and performing resource scheduling based on the target resource scheduling scheme to obtain a target resource scheduling result includes: Obtain the scheduling task of the target industrial scenario, and perform resource scheduling on the scheduling task according to the resource scheduling scheme to obtain the resource scheduling result; The transmission energy consumption of the resource scheduling result is analyzed and processed to obtain a transmission energy consumption analysis result, the receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result, and a performance analysis result is obtained according to the transmission energy consumption analysis result and the receiving energy consumption analysis result. The performance analysis result is compared with a preset performance index of the target industrial scene to obtain a comparison result, the resource scheduling scheme is dynamically optimized according to the comparison result to obtain a target resource scheduling scheme, the scheduling task is scheduled according to the target resource scheduling scheme to obtain a target resource scheduling result.

[0123] The transmission energy consumption of the resource scheduling result is analyzed and processed to obtain a transmission energy consumption analysis result, the receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result, and a performance analysis result is obtained according to the transmission energy consumption analysis result and the receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result.

[0124] The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result.

[0125] The receiving energy consumption of the resource scheduling result is analyzed and processed to obtain a receiving energy consumption analysis result. The target resource scheduling result is subjected to abnormality detection to obtain an abnormality detection result. If there is an abnormal event in the abnormality detection result, the abnormal event is subjected to abnormality analysis to obtain an abnormality analysis result, and a warning is given according to the abnormality analysis result.

[0126] The application further provides a computer readable storage medium, wherein the computer readable storage medium stores a resource scheduling program of a wireless sensor network, and the resource scheduling program of the wireless sensor network, when executed by a processor, implements the steps of the resource scheduling method of the wireless sensor network.

[0127] ​​​​​​In summary, the present application provides a resource scheduling method and system for a wireless sensor network, a terminal and a computer readable storage medium, the method comprising: obtaining wireless sensor network state information of a target industrial scene, extracting features from the wireless sensor network state information to obtain feature information, and performing standardization processing on the feature information to obtain target feature information; creating a scheduling parameter prediction model, optimizing the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, inputting the target feature information into the target scheduling parameter prediction model, outputting target scheduling parameters, and generating a scheduling scheme according to the target scheduling parameters to obtain a resource scheduling scheme; performing performance analysis on the resource scheduling scheme to obtain performance analysis results, dynamically optimizing the resource scheduling scheme according to the performance analysis results to obtain a target resource scheduling scheme, and performing resource scheduling according to the target resource scheduling scheme to obtain a target resource scheduling result. The scheduling scheme generation time of the present application is controlled within milliseconds, the dynamic response is timely, the safety of instruction execution is ensured, multi-objective optimization can be achieved, and the resource scheduling efficiency is greatly improved.

[0128] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusions, such that processes, methods, articles, or terminals including a series of elements not only include those elements, but also include other elements not explicitly listed, or include elements inherent to such processes, methods, articles, or terminals. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article, or terminal including the element.

[0129] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing relevant hardware (such as a processor, a controller, etc.) to complete, and the program can be stored in a computer readable computer readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The computer readable storage medium can be a memory, a magnetic disc, an optical disc, etc.

[0130] It should be understood that the application of the present application is not limited to the above examples, and those skilled in the art can improve or modify the above description, all of which should be within the scope of the appended claims of the present application.

Claims

1. A resource scheduling method for wireless sensor networks, characterized in that, The resource scheduling method for the wireless sensor network includes: The wireless sensor network status information of the target industrial scene is acquired, the wireless sensor network status information is subjected to feature extraction to obtain feature information, and the feature information is standardized to obtain target feature information. A scheduling parameter prediction model is created, and the scheduling parameter prediction model is optimized to obtain a target scheduling parameter prediction model. The target feature information is input into the target scheduling parameter prediction model, the target scheduling parameters are output, and a scheduling scheme is generated based on the target scheduling parameters to obtain a resource scheduling scheme. The resource scheduling scheme is subjected to performance analysis to obtain performance analysis results. Based on the performance analysis results, the resource scheduling scheme is dynamically optimized to obtain a target resource scheduling scheme. Resource scheduling is then performed according to the target resource scheduling scheme to obtain the target resource scheduling result.

2. The resource scheduling method for wireless sensor networks according to claim 1, characterized in that, The process of acquiring the wireless sensor network status information of the target industrial scenario and extracting features from the wireless sensor network status information to obtain feature information specifically includes: Acquire wireless sensor network status information of the target industrial scenario, wherein the wireless sensor network status information includes the number of network nodes, the distance between nodes, the data packet length, the initial energy of the nodes, and the priority of production data; Multidimensional feature extraction is performed on the state information of the wireless sensor network to obtain feature information, which includes basic features and engineering features. The basic features include network scale features, spatial features, communication features, and energy features. The engineering features include network density features, priority interaction features, energy density features, and complexity features.

3. The resource scheduling method for wireless sensor networks according to claim 1, characterized in that, The standardization process for the feature information to obtain target feature information specifically includes: The basic features of the feature information are filtered to obtain the target basic features, wherein the filtering process includes feature verification and redundant feature removal. The engineering features of the feature information are preprocessed to obtain the target engineering features, wherein the preprocessing includes missing value processing and outlier processing; Target feature information is obtained based on the target basic features and the target engineering features.

4. The resource scheduling method for wireless sensor networks according to claim 2, characterized in that, The process of creating a scheduling parameter prediction model, optimizing the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, inputting the target feature information into the target scheduling parameter prediction model, outputting target scheduling parameters, and generating a scheduling scheme based on the target scheduling parameters to obtain a resource scheduling scheme specifically includes: Acquire training samples, and perform data augmentation processing on the training samples according to the target augmentation method to obtain target training samples. The target augmentation method includes adaptive Gaussian noise augmentation, similar sample interpolation augmentation, and intelligent perturbation augmentation. The deep feedforward neural network is trained based on the target training samples to obtain a scheduling parameter prediction model. The scheduling parameter prediction model is then matched with an optimization strategy to obtain a target optimization strategy, wherein the target optimization strategy includes a loss function optimization strategy and a regularization optimization strategy. The scheduling parameter prediction model is optimized according to the target optimization strategy to obtain the target scheduling parameter prediction model. The target feature information is then input into the target scheduling parameter prediction model to output the target scheduling parameters. Obtain the allocation time slot information of the production data priority, and generate a scheduling scheme for the scheduling task of the target industrial scenario based on the target scheduling parameters and the allocation time slot information to obtain a resource scheduling scheme.

5. The resource scheduling method for wireless sensor networks according to claim 4, characterized in that, The step of performing data augmentation processing on the training samples according to the target augmentation method to obtain the target training samples specifically includes: Adaptive noise is generated according to the adaptive Gaussian noise enhancement method, and the adaptive noise is used to enhance the training samples to obtain a first enhanced training sample with a first preset ratio. The training samples are subjected to Euclidean distance calculation according to the similar sample interpolation enhancement method to obtain the Euclidean distance calculation result. The training samples are then interpolated according to the sample with the closest Euclidean distance in the Euclidean distance calculation result to obtain the second enhanced training sample with the second preset ratio. Based on the intelligent perturbation enhancement method, parameter sensitivity analysis is performed to obtain the perturbation intensity, and the training samples are perturbed according to the perturbation intensity to obtain a third enhanced training sample with a third preset ratio. The first enhanced training sample, the second enhanced training sample, and the third enhanced training sample are combined to obtain the target training sample.

6. The resource scheduling method for wireless sensor networks according to claim 5, characterized in that, The process of generating adaptive noise according to the adaptive Gaussian noise enhancement method specifically involves: ； in, For adaptive noise, For the first The noise level at any given moment. For the first The feature vector of the training sample at time 1. It follows a normal distribution.

7. The resource scheduling method for wireless sensor networks according to claim 5, characterized in that, The step of interpolating the training samples based on the sample with the closest Euclidean distance from the Euclidean distance calculation results is specifically as follows: ； in, As the second augmented training sample, A random value between 0.1 and 0.

9. Let be the feature vector of the training sample at time 1. This is the feature vector of the training sample at the second time step.

8. The resource scheduling method for wireless sensor networks according to claim 4, characterized in that, The step of optimizing the scheduling parameter prediction model according to the target optimization strategy to obtain the target scheduling parameter prediction model specifically includes: The loss value of the scheduling parameter prediction model is calculated according to the loss function optimization strategy to obtain the target loss value, and the parameters of the scheduling parameter prediction model are optimized according to the target loss value to obtain the scheduling parameter prediction optimization model. The scheduling parameter prediction optimization model is fitted and optimized according to the regularization optimization strategy to obtain the target scheduling parameter prediction model.

9. The resource scheduling method for wireless sensor networks according to claim 8, characterized in that, The step of calculating the loss value of the scheduling parameter prediction model according to the loss function optimization strategy is as follows: ； in, Let the mean squared error loss function be . It follows a normal distribution. For the first The true optimal value of the scheduling parameters at each moment. For the first The predicted scheduling parameters output by the scheduling parameter prediction model at each time point.

10. The resource scheduling method for wireless sensor networks according to claim 8, characterized in that, The step of fitting and optimizing the scheduling parameter prediction optimization model according to the regularization optimization strategy to obtain the target scheduling parameter prediction model specifically includes: Optimizer matching is performed according to the regularization optimization strategy to obtain the target optimizer, and the parameters of the target optimizer are configured to obtain the optimization parameter configuration result, wherein the optimization parameter configuration result includes the initial learning rate, learning rate scheduling, batch size and maximum number of training iterations; The scheduling parameter prediction optimization model is fitted and optimized based on the optimization parameter configuration results to obtain the target scheduling parameter prediction model.

11. The resource scheduling method for wireless sensor networks according to claim 1, characterized in that, The process of performing performance analysis on the resource scheduling scheme to obtain performance analysis results, dynamically optimizing the resource scheduling scheme based on the performance analysis results to obtain a target resource scheduling scheme, and then performing resource scheduling based on the target resource scheduling scheme to obtain a target resource scheduling result, specifically includes: Obtain the scheduling task of the target industrial scenario, and perform resource scheduling on the scheduling task according to the resource scheduling scheme to obtain the resource scheduling result; The transmission energy consumption of the resource scheduling result is analyzed and processed to obtain the transmission energy consumption analysis result. The reception energy consumption of the resource scheduling result is analyzed and processed to obtain the reception energy consumption analysis result. The performance analysis result is obtained based on the transmission energy consumption analysis result and the reception energy consumption analysis result. The performance analysis results are compared with the preset performance indicators of the target industrial scenario to obtain the comparison results. The resource scheduling scheme is dynamically optimized based on the comparison results to obtain the target resource scheduling scheme. The scheduling task is then scheduled according to the target resource scheduling scheme to obtain the target resource scheduling result.

12. The resource scheduling method for wireless sensor networks according to claim 11, characterized in that, The analysis and processing of the transmission energy consumption of the resource scheduling results specifically includes: ； in, For transmission energy consumption, For data packet length, The average distance between nodes. For circuit power consumption, For the power amplifier energy in the free space model, For the power amplifier energy in the multipath fading model, The threshold distance.

13. The resource scheduling method for wireless sensor networks according to claim 11, characterized in that, The analysis and processing of the energy consumption of receiving the resource scheduling results specifically includes: ； in, To receive energy consumption, For data packet length, This refers to the power consumption of the circuit.

14. The resource scheduling method for wireless sensor networks according to claim 11, characterized in that, The step of dynamically optimizing the resource scheduling scheme based on the comparison results to obtain the target resource scheduling scheme further includes: Anomaly detection is performed on the target resource scheduling results to obtain anomaly detection results; If an abnormal event is found in the anomaly detection results, anomaly analysis is performed on the abnormal event to obtain the anomaly analysis results, and an early warning is issued based on the anomaly analysis results.

15. A resource scheduling system for a wireless sensor network, characterized in that, The resource scheduling system for the wireless sensor network includes: The feature extraction module is used to acquire the wireless sensor network status information of the target industrial scene, extract features from the wireless sensor network status information to obtain feature information, and standardize the feature information to obtain target feature information. The parameter prediction module is used to create a scheduling parameter prediction model, optimize the scheduling parameter prediction model to obtain a target scheduling parameter prediction model, input the target feature information into the target scheduling parameter prediction model, output the target scheduling parameters, and generate a scheduling scheme based on the target scheduling parameters to obtain a resource scheduling scheme. The resource scheduling module is used to perform performance analysis on the resource scheduling scheme, obtain performance analysis results, dynamically optimize the resource scheduling scheme based on the performance analysis results, obtain a target resource scheduling scheme, and perform resource scheduling based on the target resource scheduling scheme to obtain a target resource scheduling result.

16. The resource scheduling system for wireless sensor networks according to claim 15, characterized in that, The feature extraction module includes: The information acquisition unit is used to acquire the status information of the wireless sensor network in the target industrial scenario; The feature extraction unit is used to perform multi-dimensional feature extraction on the wireless sensor network state information to obtain feature information; The basic feature processing unit is used to filter the basic features of the feature information to obtain the target basic features; An engineering feature processing unit is used to preprocess the engineering features of the feature information to obtain the target engineering features; The feature merging unit is used to obtain target feature information based on the target basic features and the target engineering features.

17. The resource scheduling system for wireless sensor networks according to claim 15, characterized in that, The parameter prediction module includes: The sample augmentation unit is used to acquire training samples and perform data augmentation processing on the training samples according to the target augmentation method to obtain the target training samples. The model training unit is used to train the deep feedforward neural network based on the target training samples to obtain a scheduling parameter prediction model, and to perform optimization strategy matching on the scheduling parameter prediction model to obtain a target optimization strategy. The model optimization unit is used to optimize the scheduling parameter prediction model according to the target optimization strategy to obtain the target scheduling parameter prediction model, and input the target feature information into the target scheduling parameter prediction model to output the target scheduling parameters. The scheme generation unit is used to obtain the allocation time slot information of production data priority, and generate a scheduling scheme for the scheduling task of the target industrial scenario according to the target scheduling parameters and the allocation time slot information to obtain a resource scheduling scheme.

18. The resource scheduling system for wireless sensor networks according to claim 15, characterized in that, The resource scheduling module includes: The preliminary scheduling unit is used to acquire the scheduling tasks of the target industrial scenario, and to perform resource scheduling on the scheduling tasks according to the resource scheduling scheme to obtain the resource scheduling results. The performance analysis unit is used to analyze and process the transmission energy consumption of the resource scheduling result to obtain the transmission energy consumption analysis result, analyze and process the reception energy consumption of the resource scheduling result to obtain the reception energy consumption analysis result, and obtain the performance analysis result based on the transmission energy consumption analysis result and the reception energy consumption analysis result. The scheme optimization unit is used to compare the performance analysis results with the preset performance indicators of the target industrial scenario to obtain the comparison results, dynamically optimize the resource scheduling scheme based on the comparison results to obtain the target resource scheduling scheme, and perform resource scheduling on the scheduling task based on the target resource scheduling scheme to obtain the target resource scheduling result.

19. A terminal, characterized in that, The terminal includes: a memory, a processor, and a resource scheduler for a wireless sensor network stored in the memory and executable on the processor. When the resource scheduler for the wireless sensor network is executed by the processor, it implements the steps of the resource scheduling method for a wireless sensor network as described in any one of claims 1-14.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a resource scheduler for a wireless sensor network, which, when executed by a processor, implements the steps of the resource scheduling method for a wireless sensor network as described in any one of claims 1-14.

Images