Internet of Things equipment cooperative control method based on edge computing

Abstract

The invention provides an internet-of-things equipment cooperative control method based on edge computing, which comprises the following steps of: obtaining an initial flow regulation value of each oil storage equipment by adopting a weighted average algorithm according to an initial flow difference, distributing an initial flow regulation task in edge nodes, determining a computing load of each node, and determining the computing load of each oil storage equipment according to the computing load. Determining task allocation of the edge node according to the calculated load; when the oil unloading working conditions are switched, flow data of all the oil storage devices in the oil unloading process are collected in real time, whether the current flow is balanced or not is judged according to the initial flow adjusting value in combination with a preset balance algorithm, and if not, flow adjustment is conducted; in the flow adjustment process, flow adjustment tasks are dynamically distributed to the edge nodes according to the computing power and the communication delay of the edge nodes.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a collaborative control method for Internet of Things devices based on edge computing. Background Art

[0002] In the oil depot edge computing IoT system, the coordinated control mechanism of the oil storage equipment unloading process faces a complex technical problem. When multi-point unloading operations are carried out simultaneously, the flow of each oil storage equipment unloading process needs to maintain a dynamic balance to ensure the overall operation efficiency. However, due to differences in physical properties such as the length, diameter, and number of elbows of the unloading pipeline, the initial flow of each oil storage equipment unloading process is often inconsistent. In addition, the computing resources and communication bandwidth of the edge computing nodes are limited. How to reasonably allocate flow regulation tasks among multiple edge nodes has become a key challenge. When the unloading condition switches, such as switching from single-point unloading to multi-point unloading, or from one oil product to another, the device control unit needs to quickly adjust the flow. At this time, the edge computing node needs to collect the flow data of each unloading process in real time and calculate it according to the preset balance algorithm. However, due to differences in the computing power and communication delay of the edge nodes, some nodes may not be able to complete the calculation task in time, resulting in a lag in flow regulation. This lag will further affect the flow balance of other unloading processes, forming a chain reaction. What is more complicated is that when an oil unloading process fails or needs maintenance, the control unit needs to reallocate the flow regulation tasks. At this point, the task allocation between edge computing nodes needs to be dynamically adjusted to adapt to the new operating conditions. However, this dynamic adjustment may trigger resource competition between nodes, causing some nodes to be overloaded while other nodes are idle. This unbalanced resource allocation will affect the overall operating efficiency of the system and may even cause the flow of some oil unloading processes to be out of control. Therefore, how to achieve flow balance during multi-point oil unloading and optimize the collaborative decision-making strategy of the device control unit in real time when the operating conditions switch has become a technical problem that needs to be solved in the oil depot edge computing Internet of Things system. Summary of the invention

[0003] The present invention provides an IoT device collaborative control method based on edge computing, which mainly includes:

[0004] Obtain the initial flow data of the oil unloading process of each oil storage device and the physical characteristic parameters of the connecting pipelines, calculate the hydraulic characteristics of the pipeline network based on the physical characteristic parameters of the connecting pipelines, and obtain the initial flow difference of the oil storage devices. The physical characteristic parameters of the pipelines include the length, diameter and number of elbows of the pipelines;

[0005] According to the initial flow differences, the weighted average algorithm is used to obtain the initial flow adjustment values of each oil storage device, the initial flow adjustment tasks are allocated among the edge nodes, the computing loads of each node are determined, and the task allocation of the edge nodes is determined based on the calculated loads.

[0006] When the oil unloading condition switches, the flow data of the oil unloading process of each oil storage device is collected in real time. According to the initial flow adjustment value and combined with the preset balance algorithm, it is judged whether the current flow reaches balance. If not, the flow is adjusted.

[0007] During the process of flow adjustment, according to the computing power and communication delay of the edge nodes, the flow adjustment tasks are dynamically allocated to each edge node.

[0008] When it is monitored that the flow of an oil storage device deviates from the initial flow adjustment value, it is determined that the corresponding oil storage device is faulty or under maintenance. Combining with the oil unloading condition of the corresponding oil storage device, the dynamic load balancing algorithm is used to re-determine the flow adjustment tasks allocated to other edge nodes.

[0009] The computing loads of the edge nodes and the actual flows of the oil storage devices are monitored in real time, and the flow adjustment results of the oil unloading process of each oil storage device are obtained. If the difference between the actual flow and the adjustment target value is greater than the preset threshold, it is determined that the balance state is not reached, and then the flow differences of each oil storage device are recalculated, and the task allocation of the edge nodes is adjusted until the flow balance state is achieved.

[0010] After receiving the adjusted task allocation, the edge node decomposes the task into multiple subtasks, and according to the device type and status, issues corresponding control instructions to the corresponding oil storage devices through the Internet of Things protocol. After the device executes the instructions, the real-time status data is fed back to the edge node.

[0011] Furthermore, the initial flow data of the oil unloading process of each oil storage device and the physical characteristic parameters of the connecting pipelines are obtained, the hydraulic characteristics of the pipe network are calculated according to the physical characteristic parameters of the connecting pipelines, and the initial flow differences of the oil storage devices are obtained. The physical characteristic parameters of the pipeline include the pipeline length, diameter and the number of elbows, and it includes:

[0012] The tank pressure parameters, oil viscosity indexes, tank volume sizes, and tank liquid level height data during the oil unloading process are obtained from the monitoring unit of the oil storage device. The Reynolds number calculation formula is used to judge the oil flow regime, and the liquid motion resistance value per unit length of the pipeline is calculated through the Darcy-Weisbach formula.

[0013] According to the pipeline diameter length, the number of pipeline elbows, and the ratio of pipeline diameter to wall thickness in the pipeline connection characteristics, a pipe network structure parameter matrix is constructed. The Bernoulli equation is used to solve the pressure loss of each node in the pipe network, and the overall flow loss reference value of the pipeline is calculated through the pressure loss of the pipe network nodes.

[0014] A hydraulic parameter matrix of the pipe network is established for the liquid motion resistance value per unit length of the pipeline and the benchmark value of the overall flow loss of the pipeline. The least squares method is used to fit the hydraulic parameters of the pipe network to obtain the equation of the hydraulic characteristic curve of the pipe network.

[0015] Real-time data of the unloading flow rate is obtained from the oil storage equipment monitoring unit. The theoretical flow rate value is calculated according to the equation of the hydraulic characteristic curve of the pipe network, and the local resistance correction coefficient of the pipe section is obtained by comparing the actual flow rate with the theoretical flow rate.

[0016] A corrected hydraulic characteristic data set of the pipe network is constructed for the local resistance correction coefficient of the pipe section and the equation of the hydraulic characteristic curve of the pipe network. The support vector regression algorithm is used to solve the corrected hydraulic characteristics of the pipe network to obtain the initial flow rate difference data of the oil storage equipment.

[0017] A hydraulic balance equation set of the pipe network is established according to the equation of the hydraulic characteristic curve of the pipe network and the initial flow rate difference data of the oil storage equipment. The Gaussian elimination method is used to solve the equation set to obtain the final initial flow rate difference values of each oil storage equipment.

[0018] Furthermore, the method further includes: collecting the initial flow rate during the unloading process of the oil storage equipment, calculating the average flow rate value, and at the same time calculating the hydraulic characteristics of the pipe network in combination with the physical characteristic parameters of the pipelines connected to each equipment, establishing a flow distribution model, comparing the actual flow rate of each equipment with the theoretical optimal flow rate, calculating the flow rate difference value and the difference percentage between the equipment, and forming an initial flow rate difference data table, specifically including:

[0019] The initial flow rate data of each equipment during the unloading process is obtained from the oil storage equipment monitoring unit. The arithmetic average method is used to calculate the average flow rate benchmark value of the equipment group, and the real-time flow rate data is obtained by collecting the liquid level height and the oil product flow rate index data of the storage tank through the flow rate sensor.

[0020] A basic characteristic matrix of the pipe network is constructed according to the real-time flow rate data, the pipeline pressure parameters, and the pipeline diameter length. The Reynolds number calculation formula is used to judge the flow state of the oil product, and the flow resistance coefficient per unit length of the pipeline is calculated through the Darcy-Weisbach formula.

[0021] A flow distribution calculation model is established for the flow resistance coefficient per unit length of the pipeline and the average flow rate benchmark value of the equipment group. The Bernoulli equation is used to calculate the pressure loss value of each pipe section, and the theoretical optimal flow rate value is obtained through fluid mechanics calculation.

[0022] A node pressure distribution matrix of the pipe network is constructed according to the pressure loss values of each pipe section. The least squares method is used to fit the hydraulic characteristics of the pipe network, and the hydraulic balance equation of the pipe network is obtained through calculation.

[0023] A flow difference calculation model is established for the pipe network hydraulic balance equation and the theoretical optimal flow rate value. The support vector regression algorithm is used to compare the actual flow rate with the theoretical flow rate, and the initial value of the flow difference of the oil storage equipment is obtained.

[0024] A flow difference data table is established based on the initial value of the flow difference of the oil storage equipment, and the Gaussian elimination method is used to calculate the flow difference data to obtain the percentage value of the flow difference of each oil storage equipment.

[0025] Furthermore, according to the initial flow difference, the weighted average algorithm is used to obtain the initial flow adjustment value of each oil storage equipment, the initial flow adjustment tasks are allocated among the edge nodes, the computing load of each node is determined, and the task allocation of the edge nodes is determined according to the calculated load, including:

[0026] Obtain the flow difference values of each oil storage equipment from the flow difference database, use the equipment flow weight ratio to perform weighted calculation on the flow difference, and obtain the initial adjustment reference value by comparing the actual flow value of the equipment with the average flow value.

[0027] Construct an edge node computing task matrix according to the initial adjustment reference value, use the least squares method to solve the upper limit of the processing capacity of the edge nodes, and obtain the edge node processing capacity distribution value through task volume normalization calculation.

[0028] Establish a computing load benchmark matrix for the edge node processing capacity distribution value, use the random forest algorithm to predict the node computing load, and obtain the upper limit value of the task capacity of each edge node through calculation.

[0029] Construct a task allocation optimization matrix according to the upper limit value of the edge node task capacity, use the analytic hierarchy process to sort the task priorities, and obtain the edge node computing task allocation sequence through calculation.

[0030] Establish a flow adjustment execution matrix for the edge node computing task allocation sequence, use the Gaussian elimination method to solve the adjustment execution value, and obtain the initial flow adjustment value of the oil storage equipment through calculation.

[0031] Construct an adjustment task allocation table according to the initial flow adjustment value of the oil storage equipment, use the support vector regression algorithm to predict the edge node computing load, and obtain the final edge node task allocation plan.

[0032] Furthermore, when the unloading operation condition changes, the flow data of the unloading process of each oil storage equipment is collected in real time. According to the initial flow adjustment value and combined with the preset balance algorithm, it is judged whether the current flow reaches balance. If not, the flow is adjusted, including:

[0033] Obtain the information of the switching point of the oil unloading working condition from the oil storage equipment monitoring unit, use a data collector to collect the oil unloading flow data of each oil storage equipment, and obtain the flow change data at the moment of working condition switching by comparing the flow values before and after switching.

[0034] Construct a flow fluctuation value matrix based on the flow change data at the moment of oil unloading working condition switching, use the least squares method to calculate the flow fluctuation trend, and obtain the flow fluctuation characteristic curve through time series analysis.

[0035] Establish a flow balance judgment matrix for the flow fluctuation characteristic curve and the initial flow adjustment value, use the neural network algorithm to predict the flow balance state, and judge whether the current flow is in the balance interval through a preset balance threshold.

[0036] Construct a flow adjustment instruction matrix according to the judgment result of the flow balance state, use the support vector regression algorithm to optimize the adjustment instruction, and obtain the flow adjustment parameters of each oil storage equipment through calculation.

[0037] Establish an adjustment execution control matrix for the flow adjustment parameters, use the proportional integral algorithm to refine the calculation of the adjustment amount, and obtain the valve opening adjustment value through the output of the controller.

[0038] Construct a flow adjustment feedback matrix according to the valve opening adjustment value, use the Gaussian elimination method to verify the adjustment result, and obtain the balance judgment result after flow adjustment through calculation.

[0039] Furthermore, the method further includes: comparing the real-time flow data of each oil storage equipment collected in real time with its corresponding initial flow adjustment value, calculating the difference between the actual flow of each equipment and the adjustment target value, evaluating the balance degree of the overall system flow distribution through the variance analysis or mean square error calculation method in the preset balance algorithm, setting the allowable deviation range threshold, and when the flow differences of all equipment are less than the deviation range threshold and the flow ratio between equipment meets the design requirements, determining that the system reaches the flow balance state, specifically including:

[0040] Obtain the real-time flow data from the oil storage equipment monitoring unit, use a data collector to monitor the oil unloading flow of each oil storage equipment, and obtain the equipment flow deviation data by calculating the difference between the actual flow and the adjustment target value.

[0041] Construct a mean square error calculation matrix according to the equipment flow deviation data, use the variance analysis method to standardize the flow deviation values of each equipment, and obtain the flow distribution standard deviation and root mean square error through calculation.

[0042] Establish a balance judgment benchmark matrix for the flow distribution standard deviation and the allowable deviation range threshold, use the support vector machine algorithm to classify the flow deviation data, and obtain the preliminary judgment result of the flow balance state through calculation.

[0043] Construct an equipment flow ratio matrix according to the initial judgment result of the flow balance state, calculate the flow ratio between equipment by using a linear regression algorithm, and obtain a flow distribution rationality judgment value by comparing with the designed flow ratio range.

[0044] Establish an equilibrium degree evaluation matrix for the flow distribution rationality judgment value, predict the overall flow distribution of the equipment group by using a neural network algorithm, and obtain the flow equilibrium degree index of the oil storage equipment group through calculation.

[0045] Construct a balance state judgment table according to the flow equilibrium degree index of the oil storage equipment group, verify the equilibrium judgment conditions by using the Gaussian elimination method, and obtain the final flow balance state judgment result through calculation.

[0046] Further, during the process of flow adjustment, according to the computing power and communication delay of edge nodes, dynamically allocate flow adjustment tasks to each edge node, including:

[0047] Obtain node processing rate and communication response time data from the edge node monitoring unit, monitor the usage rate of the central processing unit, memory occupancy rate, and network delay time of each node by using a data collector, and obtain edge node performance evaluation data through calculation.

[0048] Construct a resource state matrix according to the edge node performance evaluation data, fit the node computing load and network transmission delay by using the least squares method, and obtain the node processing capacity score value through calculation.

[0049] Establish a task allocation benchmark matrix for the node processing capacity score value, predict the node resource utilization rate by using a support vector machine algorithm, and obtain the data transmission capacity index through the network bandwidth monitoring value.

[0050] Construct a communication delay evaluation matrix according to the data transmission capacity index, calculate the data synchronization delay between nodes by using a time series analysis method, and obtain the node communication performance level by comparison.

[0051] Establish a task priority matrix for the node communication performance level, calculate the task weight by using the analytic hierarchy process, and obtain the initial task allocation scheme through a resource scheduling algorithm.

[0052] Construct a dynamic scheduling matrix according to the initial task allocation scheme, predict the node load change trend by using a neural network algorithm, and obtain the dynamic allocation sequence of the flow adjustment task through real-time calculation.

[0053] Further, when it is monitored that the flow of the oil storage equipment deviates from the initial flow adjustment value, it is determined that the corresponding oil storage equipment is faulty or under maintenance. Combining with the unloading condition of the corresponding oil storage equipment, use a dynamic load balancing algorithm to re-determine the flow adjustment tasks assigned to other edge nodes, including:

[0054] Obtain real-time flow data and oil unloading condition parameters from the oil storage equipment monitoring unit, use a data collector to monitor the flow deviation values of each oil storage equipment, and obtain the equipment operation status matrix by comparing with the initial flow adjustment value.

[0055] Construct a fault determination reference table according to the equipment operation status matrix, use the least squares method to calculate the flow fluctuation amplitude and duration of the equipment, and obtain the equipment abnormal state data by judging through the operation stability threshold.

[0056] Establish an equipment status evaluation matrix for the equipment abnormal state data and oil unloading condition parameters, use the support vector machine algorithm to predict the equipment operation trend, and obtain the equipment fault type determination result through calculation.

[0057] Construct a node task reallocation matrix according to the equipment fault type determination result, use the analytic hierarchy process to evaluate the computing load of the edge nodes, and obtain the available computing resource data through the node performance index.

[0058] Establish a load balancing reference matrix for the available computing resource data, use the time series analysis method to evaluate the node processing ability, and obtain the node load allocation weight through the task volume calculation.

[0059] Construct a dynamic scheduling matrix according to the node load allocation weight, use the neural network algorithm to optimize the task allocation scheme, and obtain the updated flow adjustment task allocation sequence through real-time calculation.

[0060] Further, the computing load of the real-time monitoring edge nodes and the actual flow of the oil storage equipment are monitored, and the flow adjustment results of each oil storage equipment during the oil unloading process are obtained. If the difference between the actual flow and the adjustment target value is greater than the preset threshold, it is determined that the balance state is not reached, and then the flow difference of each oil storage equipment is recalculated, and the task allocation of the edge nodes is adjusted until the flow balance state is achieved, including:

[0061] Obtain node computing load data and oil unloading flow data from the edge node monitoring unit, use a data collector to monitor the actual flow of each oil storage equipment and the adjustment target value, and obtain the flow adjustment execution result matrix by calculating the deviation value.

[0062] Construct a balance determination reference table according to the flow adjustment execution result matrix, use the least squares method to fit the flow deviation value, and obtain the balance state determination result by comparing with the preset balance threshold.

[0063] If the balance state determination result shows that the flow difference is greater than the preset threshold, use the analytic hierarchy process to calculate the flow difference data, and obtain the new flow adjustment parameters through weighted average.

[0064] Establish a node task allocation matrix for the new traffic regulation parameters, evaluate the computing resources of edge nodes using the support vector machine algorithm, and obtain the node load allocation scheme through calculation.

[0065] Construct a scheduling optimization matrix according to the node load allocation scheme, evaluate the task execution efficiency using the time series analysis method, and obtain the edge node task adjustment sequence through calculation.

[0066] Establish a load balancing matrix for the edge node task adjustment sequence, predict the adjustment execution effect using the neural network algorithm, and obtain a new task allocation scheme through real-time calculation.

[0067] Obtain the traffic regulation result according to the new task allocation scheme, and repeat steps 2 to 6 using the cyclic iteration method until the traffic difference is less than the preset balance threshold.

[0068] Further, after receiving the adjusted task allocation, the edge node decomposes the task into multiple subtasks, and according to the device type and status, issues corresponding control instructions to the corresponding oil storage devices through the Internet of Things protocol. After the device executes the instructions, it feeds back the real-time status data to the edge node, including:

[0069] Obtain the adjusted task allocation data from the edge node, decompose the control task using the task grading method, and obtain the subtask priority matrix by judging the device model and working condition parameters.

[0070] Construct an instruction generation table according to the subtask priority matrix, match the device operation parameters with the control target value using the least squares method, and obtain the execution instruction sequence through optimization calculation.

[0071] Establish a communication protocol conversion table for the execution instruction sequence, identify the communication interface of the oil storage device using the configuration parsing method, and obtain the standard Internet of Things communication message through the protocol converter.

[0072] Construct a device control matrix according to the standard Internet of Things communication message, predict the device operation state using the support vector machine algorithm, and obtain the control instruction execution parameters through real-time monitoring.

[0073] Establish a data feedback channel for the control instruction execution parameters, collect the device status data using the time series analysis method, and obtain the real-time operation status value through parsing.

[0074] Construct a device monitoring matrix according to the real-time operation status value, analyze the status parameters using the neural network algorithm, and obtain the device operation feedback data through calculation.

[0075] The decomposition of the control tasks for oil storage equipment and the issuance of instructions involve multiple links. Taking 6 oil storage equipment in a petrochemical enterprise as an example, the task allocation data received by the edge node includes equipment type, operating condition parameters, and control objectives.

[0076] The control tasks are decomposed into three categories: flow regulation, pressure control, and temperature monitoring through task hierarchical processing. Among them, the flow regulation has the highest priority, the pressure control comes second, and the temperature monitoring has the lowest priority, forming a sub-task priority matrix.

[0077] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0078] The present invention discloses an Internet of Things device collaborative control method based on edge computing. The method calculates the hydraulic characteristics of the pipe network and the initial flow difference by obtaining the initial flow data of the oil storage equipment and the physical characteristics parameters of the pipeline. The weighted average algorithm is used to determine the initial flow adjustment value, and the computing tasks are allocated at the edge node. When the unloading working condition is switched, the flow data is collected in real time, and the flow is adjusted in combination with the preset balance algorithm. According to the computing power and communication delay of the edge node, the flow regulation tasks are dynamically allocated. When a device failure or maintenance is detected, the dynamic load balancing algorithm is used to reallocate the tasks. Through continuous monitoring and adjustment until the flow balance state is reached. The edge node decomposes the tasks and issues control instructions to the oil storage equipment to achieve precise flow control. The present invention effectively solves the problem of flow imbalance during the unloading process of multiple oil storage equipment, and improves the unloading efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a flowchart of an Internet of Things device collaborative control method based on edge computing according to the present invention.

[0080] Figure 2 It is a schematic diagram of an Internet of Things device collaborative control method based on edge computing according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] In the following description, specific details such as system structures and calculation methods are proposed for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present invention. However, those skilled in the art should clearly understand that the present invention can also be implemented in other embodiments without these specific details.

[0082] Such as Figure 1-2 , in the embodiment of the present invention, the Internet of Things device collaborative control method based on edge computing realizes the flow balance of the oil storage equipment through the edge node. The following is a detailed description through specific steps:

[0083] S101. Collect the initial flow rate data during the oil unloading process of each oil storage device and the physical characteristic parameters of the connecting pipelines. Calculate the hydraulic characteristics of the pipeline network based on the physical characteristic parameters of the connecting pipelines to obtain the initial flow rate differences of the oil storage devices. The physical characteristic parameters of the pipeline include the pipeline length, diameter, and the number of elbows, and relevant data is obtained from the monitoring unit for further analysis of the flow rate characteristics.

[0084] During the oil unloading process, the oil storage device provides flow rate data and status parameters in real time through the monitoring unit. In the embodiment of the present invention, the initial flow rate data is first collected from each oil storage device, and at the same time, the physical characteristic parameters of the connecting pipelines are recorded, such as the pipeline length, diameter, and the number of elbows, which directly affect the fluid movement resistance.

[0085] S1011. Obtain the tank pressure parameter, oil product viscosity index, and tank liquid level height data during the oil unloading process from the monitoring unit of the oil storage device. Use the Reynolds number calculation formula to judge the flow state of the oil product based on the tank pressure parameter. Calculate the liquid movement resistance value per unit length of the pipeline using the Darcy-Weisbach formula based on the flow state of the oil product. Construct a pipeline network structure parameter matrix according to the pipeline diameter, length, and the number of elbows, and use the Bernoulli equation to calculate the benchmark value of the overall flow loss of the pipeline to determine the hydraulic characteristics of the pipeline network.

[0086] Specifically, the tank pressure parameter reflects the dynamic conditions during oil unloading, the oil product viscosity index determines the fluid resistance characteristics, and the liquid level height data is related to the flow rate change. The Reynolds number calculation can judge whether the oil product is in a turbulent state. For example, when the Reynolds number is greater than 4000, the flow state is turbulent, and at this time, the resistance increases with the square of the flow velocity. The Darcy-Weisbach formula calculates the resistance value per unit length according to the flow state and pipeline characteristics. Further, in combination with the pipeline diameter and the number of elbows, a multi-dimensional pipeline network structure parameter matrix is constructed, and the Bernoulli equation is used to solve the pressure loss at each node, and finally, a benchmark value of the flow loss is generated to provide a basis for subsequent analysis.

[0087] S1012. Establish a pipeline network hydraulic characteristic curve equation based on the benchmark value of the overall flow loss of the pipeline. Calculate the theoretical flow rate value through the pipeline network hydraulic characteristic curve equation. Compare the theoretical flow rate value with the actual flow rate to obtain the local resistance correction coefficient of the pipe section. Construct a corrected pipeline network hydraulic characteristic data set for the local resistance correction coefficient of the pipe section. Use the support vector regression algorithm to solve the corrected pipeline network hydraulic characteristic data set to obtain the initial flow rate difference data of the oil storage device. Then, use the Gaussian elimination method to solve the pipeline network hydraulic balance equation set according to the initial flow rate difference data to determine the final initial flow rate difference value of each oil storage device.

[0088] In the embodiment of the present invention, the pipeline network hydraulic characteristic curve equation is generated by fitting the flow loss reference value and is used to predict the theoretical flow rate. The deviation between the actual flow rate and the theoretical flow rate reflects the influence of local resistance, and the calculated correction coefficient can optimize the hydraulic characteristic data. The support vector regression algorithm uses the radial basis function as the kernel to train the multi-dimensional data set and outputs the initial flow rate difference. The Gaussian elimination method accurately quantifies the flow rate differences of each device by solving the equilibrium equations. For example, the initial flow rates of three oil storage devices are 300, 280, and 260 cubic meters per hour respectively, and the differences are caused by the liquid level height and the pipeline network structure.

[0089] In addition, when collecting the initial flow rate, the average flow rate value of the device group is also calculated. Combining the real-time flow rate data and the pipeline pressure parameters, a pipeline network basic characteristic matrix is constructed, and the theoretical optimal flow rate is generated through fluid mechanics calculation. The comparison between the actual flow rate and the theoretical flow rate forms a flow rate difference data table, which provides a basis for subsequent adjustment. Taking four oil storage devices as an example, the initial flow rates are 320, 290, 310, and 300 cubic meters per hour respectively, the average value is 305 cubic meters per hour, and the difference percentage is calculated by the algorithm.

[0090] In the embodiment of the present invention, the pipeline network hydraulic characteristic analysis shows that the flow rate difference is closely related to the pressure loss, the local resistance correction coefficient, and the liquid level height. By reasonably adjusting the opening degree of the node valve, the flow rate can be effectively balanced. For example, when the liquid level height difference is less than 500 millimeters, adjusting the valve can control the flow rate deviation within 5%. This step lays a data foundation for the edge node task allocation and ensures the accuracy of subsequent adjustment.

[0091] It can be understood that the embodiment of the present invention does not overly limit the acquisition method of the flow rate data and the specific implementation of the algorithm, and can be adjusted by technicians according to the actual scenario. The subsequent steps will achieve dynamic flow balance based on this data, which will not be elaborated here.

[0092] S102. Calculate the initial flow rate adjustment values of each oil storage device by using the weighted average algorithm according to the initial flow rate differences, and assign the adjustment tasks to the edge nodes. At the same time, determine the calculation load of each node, and optimize the task allocation scheme of the edge nodes through load analysis to achieve the flow rate balance during the oil unloading process.

[0093] First, extract the initial flow rate difference data of each device from the flow rate difference database of the oil storage device. These data reflect the deviation between the actual flow rate and the expected value of each device during the oil unloading process and are the basis for subsequent adjustment. Taking an oil depot as an example, assume that the flow rate differences of 6 oil storage devices are +15, -12, +8, -10, +6, and -7 cubic meters per hour respectively, indicating that the flow rate distribution of each device is uneven.

[0094] S1021. In an embodiment of the present invention, the actual flow values and average flow values of each oil storage device are obtained. The initial adjustment reference value is calculated by comparing the two, and an edge node computing task matrix is constructed based on the initial adjustment reference value. The upper limit of the processing capacity of the edge nodes is analyzed using the least squares method, and then the processing capacity distribution values of each node are calculated through task volume normalization to provide a basis for task allocation.

[0095] Specifically, the actual flow values are collected in real time by the monitoring unit, and the average flow value is obtained by the arithmetic mean of the flows of all devices. Taking 6 devices as an example, assuming the initial flows are 315, 288, 308, 290, 306, and 293 cubic meters per hour respectively, the average flow is 300 cubic meters per hour. The initial adjustment reference value is calculated from the difference between the actual flow and the average flow. For example, for the first device, it is 15 cubic meters per hour, and for the second device, it is -12 cubic meters per hour. Then, these reference values are combined with the device weights, which can be determined according to the device capacity or pipeline characteristics, such as 0.25, 0.15, 0.18, 0.14, 0.16, and 0.12 respectively, and the weighted adjustment values are calculated as 3.75, -1.8, 1.44, -1.4, 0.96, and -0.84 cubic meters per hour. Subsequently, an edge node computing task matrix is constructed, and the matrix includes dimensions such as adjustment values, the number of nodes, and task complexity. The least squares method is used to fit the upper limits of the processing capacities of 8 edge nodes. Assuming the upper limits are 120, 110, 105, 100, 95, 90, 85, and 80 megahertz respectively, after normalization, the capacity distribution values of each node are obtained. For example, the proportion of the 1st node is 13.5%, and the proportion of the 8th node is 9.0%, reflecting the differences in computing resources among the nodes.

[0096] S1022. In an embodiment of the present invention, a computing load benchmark matrix is established based on the edge node processing capacity distribution values. The random forest algorithm is used to predict the computing load of each node and determine the upper limit value of the task capacity. Subsequently, an optimization matrix for task allocation is constructed for the upper limit value of the task capacity, and the task priorities are sorted through the analytic hierarchy process. Finally, an edge node computing task allocation sequence is generated and the initial flow adjustment value is calculated.

[0097] In this step, the computing load benchmark matrix integrates multi-dimensional data such as node processing capabilities, task volumes, and response times. The random forest algorithm predicts the load peak by constructing multiple decision trees. For example, the prediction results are 102, 95, 90, 88, 82, 78, 75, 70 MHz, indicating the bearing limits of each node under high load. The upper limit value of the task capacity is determined based on this prediction to ensure that task allocation does not exceed the node capabilities. Then, an optimized task allocation matrix is constructed. The analytic hierarchy process is used to divide the tasks into three layers: the top layer is the overall traffic balance goal, the middle layer is the allocation criteria such as load balancing and response speed, and the bottom layer is the specific nodes. After priority sorting, high-load tasks are allocated to Node 1 and Node 2, and low-load tasks are allocated to the end nodes. Finally, the initial flow regulation value of each device is calculated by the weighted average algorithm combining the flow difference and weights, and the regulation range is controlled within -3 to +3 cubic meters per hour, for example.

[0098] For the execution of the regulation value, a flow regulation execution matrix is constructed to reflect the mapping relationship between tasks and nodes. The Gaussian elimination method is used to solve the matrix equation to ensure the accuracy of the regulation value. For example, the regulation value of the first device is +3 cubic meters per hour, and the second device is -2 cubic meters per hour. This method optimizes the allocation accuracy of the regulation value through iterative calculations of linear equations.

[0099] In addition, in the task allocation of edge nodes, a regulation task allocation table is constructed to record information such as task numbers, execution nodes, and computing loads. The support vector regression algorithm is used to further predict the load change trend. For example, during the task peak period, the load rates of 8 nodes are 85%, 82%, 78%, 75%, 72%, 70%, 68%, 65% respectively. The support vector regression selects the radial basis kernel function to fit the non-linear relationship between the load and the task volume, ensuring that the prediction results are close to the actual operating state.

[0100] In the embodiment of the present invention, the task allocation scheme fully considers the computing power differences and communication requirements among edge nodes. Through dynamic adjustment of priorities and load prediction, efficient utilization of resources is achieved. For example, when the load of a certain node approaches the upper limit, tasks can be transferred to low-load nodes to avoid single-point overload. Practical operation shows that this allocation strategy highly matches the regulation requirements of oil storage equipment. Especially when the flow fluctuates greatly, it can maintain the stability and response speed of the system.

[0101] It can be understood that the present invention does not impose excessive limitations on the specific parameters of the algorithm or the matrix construction method, which can be adjusted by technicians according to the actual scenario. The setting of the initial flow regulation value provides a reliable basis for subsequent working condition switching and at the same time optimizes the configuration efficiency of edge computing resources. The subsequent steps will further achieve dynamic balance based on this, which will not be elaborated here.

[0102] S103. When the unloading operation condition switches, collect the flow data of each oil storage device in real time and combine it with the initial flow regulation value. Use a preset balance algorithm to judge the flow balance state. If the balance is not achieved, perform dynamic adjustment. At the same time, generate accurate flow regulation parameters through multi-level analysis and optimization algorithms to ensure the efficient operation of the system under complex working conditions.

[0103] First, use the data collector in the oil storage device monitoring unit to collect the real-time flow data during the unloading process at a high frequency and record the switching point information when the working condition switches. For example, in the scenario where 4 oil storage devices in a petrochemical enterprise switch from Tank No. 1 to Tank No. 2, the switching point occurs at the 1200th second of operation, the collection interval is 0.5 second, the flow rates before switching are 320, 290, 310, and 300 cubic meters per hour, and after switching, they become 285, 275, 295, and 280 cubic meters per hour. By comparing the flow values before and after switching, calculate the flow change data to reflect the impact of the working condition switch on each device.

[0104] Next, construct a flow fluctuation value matrix based on the flow change data. The matrix records the flow change trend within 60 seconds before and after switching. Use the least squares method to fit the matrix, analyze the flow fluctuation trend, and obtain that the fluctuation amplitude is within plus or minus 15%. Time series analysis further generates a flow fluctuation characteristic curve, showing that the flow rate of Device No. 1 drops the fastest, decreasing by 35 cubic meters per hour within 10 seconds, while the change of Device No. 4 is relatively slow, only dropping by 20 cubic meters per hour within 60 seconds. This curve provides a dynamic characteristic basis for subsequent balance judgment.

[0105] S1031. Construct a flow balance judgment matrix for the flow fluctuation characteristic curve and the initial flow regulation value. Use the neural network algorithm to predict the balance state and make a judgment in combination with a preset threshold. Then, generate a flow regulation instruction matrix according to the judgment result, optimize the regulation parameters through the support vector regression algorithm, and finally refine them to the valve opening control value to achieve precise regulation.

[0106] Specifically, the flow balance judgment matrix integrates multi-dimensional data such as real-time flow, change rate, and initial adjustment value. The neural network algorithm adopts a three-layer structure. The input layer includes the flow and fluctuation characteristics of each device. The hidden layer is set with 20 neurons to capture non-linear relationships. The output layer judges whether it is balanced. The preset balance threshold stipulates that the flow difference between adjacent devices does not exceed 30 cubic meters per hour, and the total flow fluctuation does not exceed 8%. The prediction result shows that the system is unbalanced after switching and needs to be adjusted. Then, a flow adjustment instruction matrix is constructed. The support vector regression algorithm optimizes the adjustment parameters with a radial basis kernel function and calculates the adjustment amounts of 4 devices to be +25, +15, +5, and +10 cubic meters per hour. To ensure smooth adjustment, the proportional-integral algorithm is used to refine the adjustment amounts and convert them into valve opening adjustment values. For example, Device 1 increases by 12%, Device 2 increases by 8%, Device 3 increases by 3%, and Device 4 increases by 5%. The controller outputs signals according to these values to complete the adjustment.

[0107] During this process, a flow adjustment feedback matrix is constructed to record the flow changes after adjustment. For example, the flow rates after adjustment reach 310, 290, 300, and 290 cubic meters per hour. The Gaussian elimination method is used to verify the adjustment result. The calculation shows that the flow difference between adjacent devices is reduced to within 20 cubic meters per hour, and the total fluctuation is reduced to 5%, meeting the balance requirements. The entire adjustment takes 90 seconds, demonstrating the fast response ability.

[0108] S1032. Calculate the device flow deviation through the difference between the real-time flow and the adjustment target value, evaluate the flow distribution balance using variance analysis and support vector machine algorithms, and combine neural networks to predict the overall balance degree of the device group. Finally, generate a flow balance state determination result through multi-index verification to provide a basis for subsequent optimization.

[0109] Furthermore, obtain real-time flow data from the monitoring unit and calculate the deviation by comparing it with the adjustment target value. Taking 6 devices as an example, the real-time flows are 320, 290, 310, 300, 305, and 295 cubic meters per hour, and the target value is 300 cubic meters per hour. The deviations are +20, -10, +10, 0, +5, and -5 cubic meters per hour respectively. Construct a mean square error calculation matrix. The standard deviation calculated by variance analysis is 10.8 cubic meters per hour, and the root mean square error is 11.2 cubic meters per hour, indicating a low distribution dispersion. The support vector machine algorithm classifies the deviation data, sets a threshold of 15 cubic meters per hour, and judges that the system is close to balance. Subsequently, construct a device flow ratio matrix, and linearly regress to calculate the ratios of adjacent devices, such as 1.10, 0.94, etc., all within the designed range of 0.9 to 1.1. The neural network predicts the balance degree with 16 hidden layer neurons, and the output index is 0.92, higher than the threshold of 0.85. The Gaussian elimination method verifies the comprehensive index and confirms that the system reaches balance.

[0110] The realization of flow balance benefits from real-time monitoring and multi-algorithm collaboration. After adjustment, the total flow fluctuation of the device group is controlled within 5%, reflecting the stability and accuracy of the system during working condition switching. This method provides technical support for dynamic flow management and can be widely applied to the Internet of Things scenarios in oil depots. The subsequent steps will further optimize the control strategy using this result, which will not be elaborated here.

[0111] S104. During the flow adjustment process, dynamically allocate flow adjustment tasks to each node according to the computing power and communication delay of the edge nodes to ensure task execution efficiency and system real-time performance. At the same time, achieve reasonable resource scheduling through multi-level data analysis and optimization algorithms. Here, the computing power reflects the speed of the node to process tasks, and the communication delay affects the timeliness of data transmission and synchronization.

[0112] First, collect performance data from the edge node monitoring unit, and use the data collector to obtain the central processor utilization rate, memory occupancy rate, and network latency of each node in real time. Taking 8 edge nodes in an oil depot as an example, the central processor utilization rates are 65%, 58%, 72%, 45%, 62%, 53%, 68%, 50% respectively, the memory occupancy rates are 55%, 48%, 62%, 42%, 58%, 45%, 60%, 46% respectively, and the network latency times are 25, 28, 32, 22, 27, 24, 30, 23 milliseconds respectively. These data reflect the operating status of the nodes under the current working conditions and lay the foundation for subsequent evaluation.

[0113] S1041. In the embodiment of the present invention, calculate the edge node performance evaluation index through the collected performance data and construct a resource status matrix. Use the least squares method to fit the relationship between the computing load and network latency to generate a node processing capacity score value. Subsequently, based on the score value, use the support vector machine algorithm to predict the resource utilization rate and combine the network bandwidth data to evaluate the transmission capacity, providing a quantitative basis for task allocation.

[0114] Specifically, the performance evaluation index is obtained by weighted calculation of the central processor utilization rate, memory occupancy rate, and network latency time. For example, the weights are set to 0.4, 0.3, 0.3. After calculation, the performance evaluation values of each node reflect their comprehensive capabilities. The resource status matrix integrates these indicators into a multi-dimensional data set. The least squares method fits the relationship between the computing load and latency through linear regression to generate a processing capacity score value, such as 85, 78, 70, 92, 80, 88, 75, 90. A high score value indicates that the node is more suitable for high-load tasks. Then, the support vector machine algorithm predicts the trend of resource utilization rate with the Gaussian kernel function. For example, the prediction shows that the load of node 2 will rise to 75% within 30 minutes, while node 4 remains stable. Combining the network bandwidth monitoring values, such as the bandwidth range is between 10 and 15 megabytes per second, calculate the data transmission capacity index to further quantify the communication efficiency between nodes.

[0115] On this basis, a communication delay evaluation matrix is constructed to record the data synchronization time delay between nodes, and the timing analysis reveals the delay fluctuation characteristics. For example, the maximum delay between Node 1 and Node 3 reaches 35 milliseconds, and the minimum is 20 milliseconds. Nodes with larger fluctuations need to reduce synchronous intensive tasks. The communication performance is accordingly divided into four levels: Nodes 4 and 8 are at the first level with the best performance; Nodes 1, 2, and 5 are at the second level; Node 6 is at the third level; and Nodes 3 and 7 are at the fourth level.

[0116] S1042. In the embodiment of the present invention, a task priority matrix is constructed according to the communication performance level and the transmission capacity index, and the analytic hierarchy process is used to determine the task weights. The neural network predicts the load change trend and combines with the resource scheduling algorithm to generate a dynamic task allocation sequence, ensuring the real-time matching of task allocation and node capabilities.

[0117] The task priority matrix is constructed using the analytic hierarchy process and is divided into three layers: The top layer is the overall goal of traffic regulation, the middle layer is the scheduling criteria such as computing power and communication delay, and the bottom layer is each node. After calculating the task weights, high-priority tasks such as real-time traffic calculation are assigned to the first-level and second-level nodes, and low-priority tasks such as data recording are assigned to the third-level and fourth-level nodes. The neural network predicts the load trend with a three-layer structure. The input layer includes performance evaluation values and delay data, the hidden layer has 10 neurons to process the non-linear relationship, and the output layer gives the future load peak value, such as reaching 85% during the task peak period. The resource scheduling algorithm generates an initial task allocation plan according to the prediction result and triggers dynamic adjustment when the load exceeds 90% or the delay exceeds 40 milliseconds. For example, the task is migrated from Node 3 to Node 8.

[0118] The dynamic scheduling matrix records the allocation sequence and makes real-time adjustments to ensure load balance. For example, the initial plan allocates 50% of the tasks to Nodes 4 and 8, 30% to Nodes 1 and 2, and the rest are distributed to other nodes. In actual operation, the load balance is improved by 20%, the response time is shortened by 15%, and the resource utilization rate is stabilized at 70% to 80%, reflecting the efficiency of dynamic allocation under complex working conditions.

[0119] It can be understood that the present invention does not strictly limit the specific weights or algorithm parameters, and those skilled in the art can adjust them according to actual needs. The dynamic allocation mechanism ensures the efficient utilization of edge computing resources and the stability of traffic regulation through multi-dimensional analysis and real-time optimization, providing reliable support for the subsequent steps, which will not be elaborated here.

[0120] S105. When it is monitored that the flow rate of the oil storage equipment deviates from the initial adjustment value, it is determined that there may be a fault or it is in a maintenance state, and the flow regulation task is redistributed to other edge nodes through the dynamic load balancing algorithm in combination with the oil unloading working condition to ensure the continuity and stability of the system operation.

[0121] First, obtain real-time flow data and operating conditions parameters from the oil storage equipment monitoring unit, and use the data collector to monitor the flow deviation of each device. Taking six oil storage devices in an oil depot as an example, assume that the flow rate of device No. 2 drops from 300 cubic meters per hour to 240 cubic meters per hour, deviating from the initial adjustment value by 20%. At this time, the pressure under the high-pressure oil unloading condition is 0.4 MPa. By comparing the real-time flow rate with the initial adjustment value, construct an equipment operation state matrix, which includes multi-dimensional parameters such as flow rate fluctuation, pressure change, and temperature, providing a data basis for fault analysis.

[0122] S1051. In the embodiment of the present invention, according to the equipment operation state matrix, the least squares method is used to analyze the flow rate fluctuation amplitude and duration and combine with the stability threshold to determine the abnormal state. Subsequently, the support vector machine algorithm is used to predict the operation trend and determine the fault type, providing a decision basis for task reallocation.

[0123] Specifically, the least squares method calculates the fluctuation characteristics by fitting the flow rate data. For example, the fluctuation amplitude of device No. 2 reaches 20% and lasts for 180 seconds. The operation stability threshold is set as the fluctuation not exceeding 15% and the duration less than 120 seconds. If it exceeds this range, it is determined as abnormal. The support vector machine algorithm is trained based on historical data, uses the radial basis kernel function to predict the operation trend, combines with the fault determination reference table to analyze the cause of the abnormality, and determines that device No. 2 may be in the maintenance state rather than a hardware fault. This determination is based on the smoothness of the flow rate decrease and the matching degree of the operating conditions parameters, avoiding misjudgment.

[0124] Then, construct an equipment state evaluation matrix according to the abnormal state and operating conditions parameters, integrate information such as flow rate deviation and pressure, and further verify the fault type. The prediction result shows that tasks need to be reallocated to maintain the system function.

[0125] In task reallocation, the performance of the 8 edge nodes on site becomes the key. Monitoring shows that the load rates of nodes No. 1, No. 3, and No. 5 are 65%, 58%, and 72% respectively, with room to accept tasks. The analytic hierarchy process is used to evaluate the node performance, considering the processor usage rate, memory occupancy, and network latency, and the comprehensive scores are calculated as 85, 82, and 78 respectively. The high scores indicate that these nodes are suitable for taking on additional loads.

[0126] S1052. In the embodiment of the present invention, for the fault determination result, a neural network is used to optimize the task allocation scheme and combine with time series analysis to evaluate the node processing ability. Through the dynamic scheduling matrix and the load balancing algorithm, an updated task allocation sequence is generated to ensure the high efficiency of resource utilization and the smooth transition of task execution.

[0127] Further, a node task reallocation matrix is constructed to record the available computing resource data. The time series analysis shows that the load fluctuations of nodes 1, 3, and 5 in the past 30 minutes are less than 10%, indicating strong stability. The load balancing benchmark matrix calculates the allocation weights based on the scores. For example, the weight of node 1 is 0.35, node 3 is 0.33, and node 5 is 0.32. The neural network optimizes the allocation with a three-layer structure. The input layer includes the node performance metrics and the current load. The hidden layer with 16 neurons processes complex relationships, and the output layer gives the task allocation ratio. For example, the tasks of device 2 are allocated to these three nodes at 0.4, 0.35, and 0.25. After optimization, the load rate of the node group is maintained at about 70%, avoiding overload.

[0128] The dynamic scheduling matrix adjusts the task allocation in real time. During actual operation, the load rates of nodes 1, 3, and 5 rise to 78%, 75%, and 82% respectively, still within the safe range. This mechanism maintains the continuity of traffic regulation through smooth migration, and at the same time, there are no communication congestion or resource waste phenomena.

[0129] It can be understood that the present invention does not strictly limit the thresholds or algorithm parameters, and those skilled in the art can adjust according to the actual scenario. The dynamic nature and accuracy of task reallocation provide guarantee for the stable operation of the system in the fault scenario. The subsequent steps will optimize the control process based on this, which will not be elaborated here.

[0130] S106. By real-time monitoring the computing load of edge nodes and the actual flow rate of oil storage equipment and analyzing the flow regulation results, if the deviation between the actual flow rate and the regulation target value exceeds the preset threshold, it is determined that the balance state is not reached. Then, the flow rate difference is recalculated and the task allocation of edge nodes is adjusted until the flow balance is achieved.

[0131] First, the data collector is used to obtain the actual flow rate data of each device from the monitoring unit of the oil storage equipment and compare it with the regulation target value. Taking 6 oil storage devices in an oil depot as an example, the first-round monitoring shows that the actual flow rates are 320, 290, 310, 300, 305, and 295 cubic meters per hour respectively, while the target value is 300 cubic meters per hour for all, and the deviation values are +20, -10, +10, 0, +5, and -5 cubic meters per hour respectively. At the same time, the computing loads of 8 edge nodes are 75%, 68%, 72%, 65%, 70%, 63%, 69%, and 64% respectively, reflecting the current resource occupancy.

[0132] Next, a device flow regulation execution result matrix is constructed to integrate the deviation data between the actual flow rate and the target value. The least squares method is used to fit the deviation trend and calculate the balance state determination result. The preset balance threshold is that the deviation does not exceed 15 cubic meters per hour. The analysis shows that the deviation of device 1 exceeds the threshold, indicating that the system is not yet balanced. The least squares method quantifies the deviation distribution through linear regression, providing a basis for subsequent regulation.

[0133] S1061. In the embodiments of the present invention, if it is determined that the balance has not been reached, the analytic hierarchy process is used to recalculate the flow difference to generate new adjustment parameters, and the support vector machine algorithm is used to evaluate the edge node resources and optimize the task allocation to ensure the efficient progress of the adjustment process.

[0134] Specifically, the analytic hierarchy process constructs a three-layer evaluation system: the top layer is the flow balance target, the middle layer is the adjustment priority such as the deviation size and the device importance, and the bottom layer is the flow difference of each device. After calculating the weights, it is determined that Device 1 needs to reduce by 12 cubic meters per hour, Device 2 needs to increase by 8 cubic meters per hour, and the remaining devices are slightly adjusted. Then, a device task allocation matrix is constructed, and the support vector machine algorithm is used to evaluate the node resources. Taking the node computing power, communication delay, and margin as features, the Gaussian kernel function is used to predict the availability, and it is identified that Nodes 3, 4, and 7 are suitable for undertaking the new tasks because their loads are low and their performances are stable.

[0135] The time series analysis further evaluates the task execution efficiency, showing that the efficiency fluctuations of these nodes within the past 30 minutes are less than 5%, indicating reliable processing capabilities. Based on this, a task adjustment sequence is generated, sorted from low to high according to the load, and tasks are preferentially allocated to nodes with surplus resources.

[0136] During this process, a scheduling optimization matrix is constructed to record the task allocation scheme. The neural network predicts the adjustment effect with a three-layer structure. The input layer contains node performance, task volume, and delay data. The hidden layer has 16 neurons to capture the non-linear relationship, and the output layer gives optimization suggestions. After the second round of adjustment, the flow rates are adjusted to 308, 298, 305, 300, 302, and 297 cubic meters per hour, and the maximum deviation is reduced to 8 cubic meters per hour, approaching the balanced state.

[0137] For the new scheme, a load balancing matrix is constructed and the cyclic iteration method is used to repeat the monitoring and adjustment. After three rounds of iteration, the flow rates finally stabilize at 302, 301, 299, 300, 301, and 298 cubic meters per hour, all deviations are less than the threshold, and the node loads converge to 68% - 72%, achieving balanced allocation.

[0138] It can be understood that the present invention does not make a rigid stipulation on the threshold or the number of iterations, and can be adjusted according to actual needs. The multi-round dynamic adjustment ensures the flow accuracy and resource efficiency, providing support for the continuous optimization of the system. The subsequent steps will be carried out based on this and will not be elaborated.

[0139] S107. After receiving the adjusted task allocation, the edge node decomposes it into multiple subtasks and issues control instructions to the corresponding devices through the Internet of Things protocol according to the type and status of the oil storage devices. At the same time, after the devices execute the instructions, they feedback the real-time status to the edge node to form a closed-loop control.

[0140] First, the edge node extracts information from the task assignment data and decomposes the control task into several subtasks using the task classification method. Taking six oil storage devices in an oil depot as an example, the tasks are decomposed into three categories: flow regulation, pressure control, and temperature monitoring. According to the equipment model and operating conditions parameters such as flow demand and pressure conditions, it is determined that the flow regulation has the highest priority, followed by pressure control, and temperature monitoring has the lowest priority, generating a subtask priority matrix to provide a basis for subsequent instruction allocation.

[0141] Next, an instruction generation table is constructed based on the subtask priority matrix, and the equipment operating parameters are matched with the control target values through the least squares method. For example, the real-time flows are 320, 290, 310, 300, 305, 295 cubic meters per hour, and the target value is 300 cubic meters per hour. The least squares method generates an execution instruction sequence by fitting the deviation, including instructions such as valve opening adjustment, flow feedback, and status query, to ensure the precise alignment of the instructions with the target.

[0142] S1071. In the embodiment of the present invention, for the execution instruction sequence, the configuration analysis is used to identify the device communication interface and generate a standard Internet of Things communication message through a protocol converter. Subsequently, a device control matrix is constructed and the support vector machine algorithm is used to predict the operating state to optimize the instruction execution parameters and improve the control accuracy.

[0143] Specifically, the communication protocol conversion table identifies the device interface based on the configuration analysis. Among them, 4 devices use the ModBus protocol and 2 devices use the OPC protocol. The protocol converter unifies these heterogeneous protocols into the Internet of Things standard message in the MQTT format to ensure the compatibility of instruction transmission. Then, a device control matrix is constructed to record the instruction type and execution parameters. The support vector machine algorithm predicts the device state with the Gaussian kernel function, and the analysis shows that the adjustment range of the valve opening should be controlled within plus or minus 5% to avoid overshoot causing flow fluctuations.

[0144] During this process, the real-time monitoring data after the instruction is issued reflects the execution effect. For example, the valve opening of device No. 1 is adjusted from 65% to 60%, and the flow gradually decreases; the valve opening of device No. 2 is increased from 55% to 58%, and the flow slowly increases. This fine adjustment benefits from the prediction optimization, ensuring the process stability.

[0145] For the feedback link, a data feedback channel is established, and the device status data is collected using the time series analysis method with a period of 0.5 seconds and a transmission delay of less than 50 milliseconds. The data includes parameters such as flow, pressure, and temperature, and a device monitoring matrix is constructed. The neural network analyzes the state changes with a three-layer structure. The input layer integrates the real-time data, the hidden layer with 20 neurons processes the non-linear relationship, and the output layer gives the operation feedback. For example, 20 seconds after the instruction is executed, the flow of device No. 1 drops to 308 cubic meters per hour, and the flow of device No. 2 rises to 296 cubic meters per hour, indicating that the adjustment takes effect gradually.

[0146] It is understandable that the present invention does not strictly limit the protocol type or decomposition method, which can be adjusted according to the actual scenario. Through hierarchical decomposition, standardized communication and real-time feedback, the closed-loop control solves the complexity of multi-device collaboration and improves the accuracy and reliability of flow regulation. Subsequent optimizations will be carried out based on this data, which will not be elaborated here.

Claims

1. An Internet of Things device collaborative control method based on edge computing, characterized in that, The method includes: Obtain the initial flow data of each oil storage device during the oil unloading process and the physical characteristic parameters of the connecting pipelines. Calculate the hydraulic characteristics of the pipe network based on the physical characteristic parameters of the connecting pipelines to obtain the initial flow differences of the oil storage devices. The physical characteristic parameters of the pipelines include pipeline length, diameter, and the number of elbows; According to the initial flow differences, use the weighted average algorithm to obtain the initial flow adjustment values of each oil storage device. Allocate the initial flow adjustment tasks among the edge nodes, determine the computing loads of each node, and determine the task allocation of the edge nodes based on the calculated loads; When the oil unloading condition changes, collect the flow data of each oil storage device during the oil unloading process in real time. According to the initial flow adjustment values and in combination with a preset balance algorithm, determine whether the current flow is balanced. If it is not balanced, perform flow adjustment; During the process of performing flow adjustment, dynamically allocate the flow adjustment tasks to each edge node according to the computing capabilities and communication delays of the edge nodes; When it is detected that the flow of an oil storage device deviates from the initial flow adjustment value, it is determined that the corresponding oil storage device is faulty or under maintenance. In combination with the oil unloading condition of the corresponding oil storage device, use the dynamic load balancing algorithm to re-determine the flow adjustment tasks allocated to other edge nodes; Monitor the computing loads of the edge nodes and the actual flows of the oil storage devices in real time, and obtain the flow adjustment results of each oil storage device during the oil unloading process. If the difference between the actual flow and the adjustment target value is greater than the preset threshold, it is determined that the balanced state has not been reached. Then recalculate the flow differences of each oil storage device and adjust the task allocation of the edge nodes until the flow balanced state is achieved; After receiving the adjusted task allocation, the edge node decomposes the task into multiple subtasks, and according to the device type and status, issues corresponding control instructions to the corresponding oil storage devices through the Internet of Things protocol in a coordinated manner. After the device executes the instruction, it feeds back the real-time status data to the edge node.

2. The method according to claim 1, characterized in that The obtaining of the initial flow data of each oil storage device during the oil unloading process and the physical characteristic parameters of the connecting pipelines, calculating the hydraulic characteristics of the pipe network based on the physical characteristic parameters of the connecting pipelines to obtain the initial flow differences of the oil storage devices, where the physical characteristic parameters of the pipelines include pipeline length, diameter, and the number of elbows, includes: Obtain the tank pressure parameters, oil product viscosity index, and tank liquid level height data from the monitoring unit of the oil storage device. Use the Reynolds number calculation formula to judge the oil flow state based on the tank pressure parameters, and calculate the liquid motion resistance value per unit length of the pipeline using the Darcy-Weisbach formula through the oil flow state; For the liquid motion resistance value per unit length of the pipeline, construct a pipe network structure parameter matrix according to the pipeline diameter length and the number of pipeline elbows, and use the Bernoulli equation to calculate the benchmark value of the overall flow loss of the pipeline for the pipe network structure parameter matrix; Establish a pipe network hydraulic characteristic curve equation based on the benchmark value of the overall flow loss of the pipeline, calculate the theoretical flow value through the pipe network hydraulic characteristic curve equation, and obtain the local resistance correction coefficient of the pipe section by comparing the theoretical flow value with the actual flow. Construct a corrected pipeline network hydraulic characteristic dataset for the local resistance correction coefficient of the pipe section, and use the support vector regression algorithm to solve the corrected pipeline network hydraulic characteristic dataset to obtain the initial flow difference data of the oil storage equipment. Solve the pipeline network hydraulic balance equations using the Gaussian elimination method based on the initial flow difference data of the oil storage equipment.

3. The method according to claim 2, wherein It further includes: Collect the initial flow rate during the oil unloading process of the oil storage equipment, calculate the average flow rate value, and at the same time calculate the pipeline network hydraulic characteristics in combination with the physical characteristic parameters of the connecting pipelines of each equipment. Establish a flow distribution model, compare the actual flow rate of each equipment with the theoretical optimal flow rate, calculate the flow difference value and the difference percentage between the equipment, and form an initial flow difference data table, specifically including: Obtain the flow rate data of each equipment during the oil unloading process from the oil storage equipment monitoring unit, calculate the average flow rate reference value of the equipment group using the arithmetic mean method, and obtain the real-time flow rate data by collecting the liquid level height data of the storage tank through a flow sensor. Construct a pipeline network basic characteristic matrix based on the real-time flow rate data and pipeline pressure parameters, and judge the flow state of the oil product using the Reynolds number calculation formula to obtain the flow resistance coefficient per unit length of the pipeline. Establish a flow distribution calculation model for the flow resistance coefficient per unit length of the pipeline and the average flow rate reference value of the equipment group, and calculate the theoretical optimal flow rate value using the Bernoulli equation. Establish a flow difference calculation model for the theoretical optimal flow rate value, and use the support vector regression algorithm to compare the actual flow rate with the theoretical flow rate to obtain the flow difference percentage value of the oil storage equipment.

4. The method according to claim 1, characterized in that, Based on the initial flow difference, use the weighted average algorithm to obtain the initial flow adjustment value of each oil storage equipment, allocate the initial flow adjustment task among the edge nodes, determine the calculation load of each node, and determine the task allocation of the edge nodes according to the calculated load, including: Obtain the actual flow rate value and the average flow rate value of the oil storage equipment, and obtain the initial adjustment reference value according to the comparison relationship between the actual flow rate value and the average flow rate value. Construct an edge node calculation task matrix for the initial adjustment reference value, use the least squares method to solve the upper limit of the edge node processing capacity, and calculate the edge node processing capacity distribution value through task volume normalization. Establish a calculation load reference matrix based on the edge node processing capacity distribution value, use the random forest algorithm to predict the node calculation load, and calculate the upper limit value of the edge node task capacity through calculation. Construct a task allocation optimization matrix for the upper limit value of the edge node task capacity, use the analytic hierarchy process to sort the task priorities, and calculate the edge node calculation task allocation sequence through calculation.

5. The method according to claim 1, wherein When the oil unloading working condition switches, collect the flow rate data of each oil storage equipment during the oil unloading process in real time. According to the initial flow adjustment value and in combination with a preset balance algorithm, judge whether the current flow rate reaches balance. If not, perform flow adjustment, including: Obtain the oil unloading working condition switching point information from the oil storage equipment monitoring unit. The oil storage equipment monitoring unit uses a data collector to collect the flow rate data, and obtains the flow rate change data at the moment of working condition switching by comparing the flow rate values before and after switching. Construct a flow fluctuation value matrix based on the said flow change data, calculate the flow fluctuation value matrix using the least squares method, and obtain a flow fluctuation characteristic curve through time series analysis; Establish a flow balance judgment matrix for the said flow fluctuation characteristic curve, process the flow balance judgment matrix using a neural network algorithm, and obtain a flow balance state through a preset balance threshold; Construct a flow regulation instruction matrix according to the said flow balance state, optimize the flow regulation instruction matrix using a support vector regression algorithm, and obtain flow regulation parameters of the oil storage equipment through calculation.

6. The method according to claim 5, characterized in that, It also includes: Compare the flow data of each oil storage equipment collected in real time with its corresponding initial flow regulation value, calculate the difference between the actual flow of each equipment and the regulation target value, evaluate the balance degree of the overall system flow distribution through the variance analysis or mean square deviation calculation method in the preset balance algorithm, set the allowable deviation range threshold, and when the flow differences of all equipment are less than the deviation range threshold and the flow ratio between equipment meets the design requirements, determine that the system reaches the flow balance state, specifically including: Use a data collector to obtain real-time oil unloading flow data from the oil storage equipment monitoring unit, and obtain equipment flow deviation data by calculating the difference between the real-time oil unloading flow and the regulation target value; Construct a mean square deviation calculation matrix according to the said equipment flow deviation data, and perform standardization processing on the deviation data using the variance analysis method to obtain the flow distribution standard deviation; Establish a balance judgment benchmark matrix for the said flow distribution standard deviation, and classify the standard deviation using a support vector machine algorithm to obtain a preliminary judgment result of the flow balance state; Construct an equipment flow ratio matrix according to the preliminary judgment result of the flow balance state, and predict the overall flow distribution of the equipment group using a neural network algorithm to obtain the flow balance degree index of the oil storage equipment group.

7. The method according to claim 1, wherein During the process of flow adjustment, dynamically allocate flow regulation tasks to each edge node according to the computing power and communication delay of the edge node, including: Use a data collector to obtain the central processor utilization rate information of the edge node, where the central processor utilization rate information of the edge node includes memory occupancy data and network delay time data, and obtain edge node performance evaluation data through calculation; Construct a resource status matrix according to the said edge node performance evaluation data, fit the node computing load and network transmission delay using the least squares method, and obtain the node processing ability score value through calculation; Establish a task allocation benchmark matrix for the said node processing ability score value, predict the node resource utilization rate using a support vector machine algorithm, and obtain the data transmission ability index through the network bandwidth monitoring value; Construct a communication delay evaluation matrix according to the said data transmission ability index, calculate the data synchronization delay between nodes using the time series analysis method, and obtain an initial task allocation plan through a resource scheduling algorithm.

8. The method according to claim 1, wherein When it is monitored that the flow of the oil storage equipment deviates from the initial flow regulation value, it is determined that the corresponding oil storage equipment is faulty or under maintenance. Combining the oil unloading working conditions of the corresponding oil storage equipment, use a dynamic load balancing algorithm to re-determine the flow regulation tasks assigned to other edge nodes, including: Obtain the real-time flow data in the oil storage equipment monitoring unit, and use a data collector to monitor the flow deviation value of the oil storage equipment according to the real-time flow data, and obtain the equipment operation status matrix by comparing the initial flow adjustment value; For the equipment operation status matrix, use the least squares method to calculate the equipment flow fluctuation amplitude and duration, and obtain the equipment abnormal status data by judging through the operation stability threshold; According to the equipment abnormal status data, use the support vector machine algorithm to predict the equipment operation trend, and obtain the equipment fault type determination result by calculating through the fault determination reference table; For the equipment fault type determination result, use the neural network algorithm to optimize the task allocation scheme, and obtain the flow adjustment task allocation sequence by calculating through the node performance index.

9. The method according to claim 1, wherein The real-time monitoring edge node calculates the computing load and the actual flow of the oil storage equipment, and obtains the flow adjustment result of each oil storage equipment during the oil unloading process. If the difference between the actual flow and the adjustment target value is greater than the preset threshold, it is determined that the balance state has not been reached, and then the flow difference of each oil storage equipment is recalculated, and the task allocation of the edge node is adjusted until the flow balance state is achieved, including: Obtain the actual flow of the oil storage equipment and the adjustment target value, and the adjustment target value is obtained by the data collector monitoring the oil storage equipment; Establish a flow adjustment execution result matrix according to the actual flow and the adjustment target value, and the flow adjustment execution result matrix uses the least squares method to fit the flow deviation value to obtain the balance state determination result; If the balance state determination result shows that the flow difference is greater than the preset balance threshold, use the analytic hierarchy process to calculate the flow difference data to obtain new flow adjustment parameters; For the new flow adjustment parameters, use the neural network algorithm to predict the adjustment execution effect to obtain a new task allocation scheme, and the new task allocation scheme is used to update the adjustment target value.

10. The method according to claim 1, wherein The edge node receives the adjusted task allocation, decomposes the task into multiple subtasks, and according to the equipment type and status, issues corresponding control instructions to the corresponding oil storage equipment through the Internet of Things protocol. After the equipment executes the instructions, it feeds back the real-time status data to the edge node, including: Use the task grading method to decompose the control task, and obtain the subtask priority matrix by judging according to the equipment model and working condition parameters; Construct an instruction generation table for the subtask priority matrix, and match the equipment operation parameters and the control target value by using the least squares method to obtain the execution instruction sequence; Conduct configuration analysis on the communication interface of the oil storage equipment according to the execution instruction sequence, and obtain the standard Internet of Things communication message by judging through the protocol converter; Construct an equipment control matrix for the standard Internet of Things communication message, and predict and analyze the equipment operation status by using the support vector machine algorithm to obtain the control instruction execution parameters.

Images