Intelligent connection management method and system for park edge side equipment

Abstract

The invention discloses a park edge side device intelligent connection management method and system, and the method comprises the following steps: predicting the power consumption trend of each load device in a future set time period; determining a preliminary connection threshold value of the load equipment based on the power consumption trend; and analyzing the current network condition, and dynamically adjusting the connection threshold by combining the power consumption trend, the network load and the real-time change of the equipment data. The method effectively solves the problems of high power consumption and low efficiency caused by the fact that a traditional connection mode is triggered only based on single time or event. Based on electric quantity monitoring and analysis and real-time sensing of the network load, the equipment can intelligently adjust the connection threshold according to the own electric quantity state and the network environment, so that the energy consumption of the equipment in the connection process is remarkably reduced.

Description

Technical Field

[0001] The present invention belongs to the field of Internet of Things, and in particular relates to a method and system for intelligent connection management of edge-side equipment at a park. Background Art

[0002] With the advancement of zero-carbon park construction, a large number of IoT asset devices are deployed on the edge of the park, including smart meters, smart sensors, low-power surveillance cameras, etc. They play a key role in realizing intelligent park management, energy optimization scheduling, and environmental monitoring. However, most of these devices are battery-powered or low-power types, and they face severe power consumption challenges when connecting to the park wireless network for data transmission.

[0003] Traditional device connection management methods are mainly divided into two types: time-triggered and event-triggered. In the time-triggered connection method, the device attempts to connect to the network and transmit data at a preset fixed time interval. The disadvantage of this method is that regardless of the current power status of the device, the urgency of the data, and the busyness of the network, it will connect on time, which may easily lead to unnecessary connection operations when the device power is low, accelerating power consumption. At the same time, when the network load is high, a large number of devices connecting at fixed times will further aggravate network congestion and affect data transmission efficiency and stability.

[0004] Although the connection method based on event triggering avoids the blindness of connection at fixed time intervals to a certain extent, it only considers the single factor of event occurrence. For example, when the sensor detects that the data change exceeds the threshold, it immediately connects to the network to transmit data, but does not comprehensively consider the remaining power of the device itself and the overall load of the current network. If the device is low on power, a single connection transmission may cause its power to drop sharply or even fail to work properly; if the network is under high load, the connection request may be delayed or fail, resulting in data loss or transmission delays, affecting the timely acquisition and processing of asset equipment data by relevant systems in the park, which is not conducive to the efficient operation and precise energy management of the zero-carbon park. Therefore, there is an urgent need for a new intelligent connection management strategy to solve the above problems and realize reliable and low-power communication of asset equipment on the edge of the zero-carbon park. Summary of the invention

[0005] In view of the above problems, the technical solution adopted by the present invention is: a method for intelligent connection management of campus edge devices, the management method comprising the following steps:

[0006] Predict the power consumption trend of each load device within a set time period in the future;

[0007] Determining a preliminary connection threshold of the load device based on the power consumption trend;

[0008] Analyze the current network status, and dynamically adjust the connection threshold in combination with the power consumption trend, network load, and real-time changes in device data.

[0009] Optionally, in the step of predicting the power consumption trend of each load device within a future set time period, it includes:

[0010] Obtain the current power and battery health data of the load device in real time, and retrieve the historical energy consumption data of the load device;

[0011] Analyze through the long short-term memory network deep learning algorithm, establish a power consumption prediction model with the current power, battery health data, and historical energy consumption data as inputs, and use the adaptive moment estimation optimization algorithm to adjust the weights and learning rate of the power consumption prediction model during the training process.

[0012] Optionally, in the step of determining the preliminary connection threshold of the load device, it includes:

[0013] Judge the power range of the load device based on the power consumption trend;

[0014] Determine the preliminary connection threshold based on the power range, where the higher the power range, the lower the preliminary connection threshold.

[0015] Optionally, the management method further includes the following steps:

[0016] Deploy network detection devices at key nodes on the edge side network of the park;

[0017] Collect network load data in real time based on the network detection devices;

[0018] Summarize and analyze the network load data, and construct a network load evaluation model based on the big data analysis algorithm.

[0019] Optionally, in the step of dynamically adjusting the connection threshold, it includes:

[0020] Classify and evaluate the data collected by each load device in the park, determine the basic importance weight of each type of data for the park operation, and combine the basic information of the roles played by various types of devices in the park to assign an initial connection priority to each load device;

[0021] Monitor the power status, network load status, and real-time changes in data of the load device, and re-evaluate the connection priority of the device according to the changed data when the set conditions are met;

[0022] Based on the correspondence between the preset priority and the connection threshold adjustment strategy, the connection threshold of the load device after priority adjustment is dynamically adjusted; wherein the adjustment strategy is that when the priority increases, the connection threshold decreases, and vice versa.

[0023] Optionally, the management method further includes the following steps:

[0024] After the load device completes the connection and transmits data, collect the data transmission effect after this connection and the impact on the park operation decision, and adjust the data basic importance weight.

[0025] Optionally, the management method further includes the following steps:

[0026] After the device completes the connection, collect the actual energy consumption data and / or network response data during this connection process;

[0027] Feed the actual energy consumption data and / or network response data back to the power consumption prediction model and / or network load assessment model, and use the stochastic gradient descent algorithm to update the model parameters;

[0028] Establish a loss function L(θ), and calculate the gradient of the loss function L(θ) with respect to the model parameter θ

[0029] Update the model parameters according to the gradient, and the update formula is where η is the learning rate;

[0030] After each round of parameter update, use the validation set data to evaluate the updated model to obtain evaluation indicators, and judge whether the evaluation indicators meet the preset conditions after several consecutive rounds of updates;

[0031] If not, reduce the learning rate and continue training until the evaluation indicators meet the preset conditions.

[0032] Optionally, the management method further includes the following steps:

[0033] Regard each load device as an agent, the action space includes adjusting the connection priority and the connection threshold, and the state space includes the device power state, network load state, data importance, and the connection decisions of surrounding agents;

[0034] Using the Q-Leaming algorithm, after the agent selects an action according to the current state, an immediate reward is obtained, where the reward function comprehensively considers the timeliness of data transmission, energy consumption savings, and the impact on the overall network stability;

[0035] After multiple rounds of iteration, each agent determines the final connection priority and connection threshold.

[0036] And, an intelligent connection management system for edge devices in a park, the management system comprising:

[0037] A power consumption monitoring module, configured to predict the power consumption trend of each load device within a set future time period;

[0038] A network load monitoring module, configured to analyze the current network condition and evaluate the real-time change of network load;

[0039] A device connection decision module, configured to determine a preliminary connection threshold for the load device according to the power consumption trend, and further configured to dynamically adjust the connection threshold in combination with the power consumption trend, network load, and real-time change of device data.

[0040] Optionally, the device connection decision module includes:

[0041] A device data processing unit, configured to classify and evaluate the data collected by the load device, and determine the basic importance weight of each type of data for park operation;

[0042] A priority evaluation unit, configured to assign an initial connection priority to each load device according to the device data processing unit, and further configured to re-evaluate the connection priority of the device according to the power condition, network load status, and real-time change of data of the load device when the set conditions are met;

[0043] A connection decision unit, configured to dynamically adjust the connection threshold of the load device after priority adjustment according to the preset correspondence between priority and connection threshold adjustment strategy; wherein the adjustment strategy is that when the priority increases, the connection threshold decreases, and vice versa.

[0044] Optionally, the device connection decision module further includes:

[0045] A decision optimization unit, configured to further optimize the connection priority and connection threshold determined by the priority evaluation unit and the connection decision unit; the optimization steps of the decision optimization unit include: regarding each load device as an agent, the action space includes adjusting the connection priority and connection threshold, and the state space includes the device power state, network load state, data importance, and connection decisions of surrounding agents; using the Q-Learning algorithm, after the agent selects an action according to the current state, it obtains an immediate reward, where the reward function comprehensively considers the timeliness of data transmission, energy consumption savings, and the impact on the overall network stability; after multiple rounds of iteration, each agent determines the final connection priority and connection threshold.

[0046] Due to the adoption of the above technical solution, the present invention has the following beneficial effects: effectively solving the problems of high power consumption and low efficiency caused by the traditional connection method based only on a single time or event trigger. Based on the power consumption monitoring and analysis and the real-time perception of network load, the device can intelligently adjust the connection threshold according to its own power state and network environment, thereby significantly reducing the energy consumption during the connection process. The battery life of the device is effectively extended, especially suitable for a large number of Internet of Things devices relying on battery power and requiring long-term stable operation in zero-carbon parks, such as smart meters, environmental sensors, etc.

[0047] With the help of the intelligent connection decision-making and priority dynamic adjustment method, fully considering the type, update frequency of the data collected by the device and its importance to the park operation, and combining the real-time state of the device and network conditions, the dynamic optimization of connection priority is realized. This ensures that key data (such as abnormal energy supply data) can be preferentially transmitted at the appropriate time, improving the timeliness and effectiveness of data transmission. Taking the energy management in the park as an example, when there is an energy supply fluctuation, the relevant monitoring devices can quickly increase the connection priority and transmit the data to the management system in time, so as to take control measures in time to avoid energy waste or supply interruption, greatly improving the intelligent level and reliability of the park energy management, and thus contributing to the realization of the overall zero-carbon goal of the park.

[0048] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structure pointed out in the specification and claims. Detailed implementation manners

[0049] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] The intelligent connection management method for edge-side devices in the park according to the embodiments of the present invention, the management method includes the following steps:

[0051] S1: Predict the power consumption trend of each load device within a future set time period. Specifically, it includes:

[0052] S101: Use a high-precision power sensor to obtain the current power of the load device in real time at a set frequency, and combine the chip data built into the battery to obtain battery health data, including internal resistance (accurate to 0.01 ohm), temperature (accurate to 0.1 degree Celsius), etc.; at the same time, retrieve the historical energy consumption data of the load device, including the energy consumption data of the device in different past working modes (such as data collection mode, data transmission mode, sleep mode, etc.).

[0053] S102: Analyze through the long short-term memory network deep learning algorithm, and establish a power consumption prediction model with the current power, battery health data, and historical energy consumption data as inputs. The output is the predicted curve of the remaining power value of the device at different power consumption rates in a future set time period.

[0054] The adaptive moment estimation optimization algorithm is used in the training process to adjust the weights and learning rate of the power consumption prediction model. The initial learning rate is set to 0.001 and gradually decays as the number of training rounds increases. The formula is:

[0055]

[0056] where epoch is the current number of training rounds, and total _ epochs is the total number of training rounds.

[0057] S2: Determine the preliminary connection threshold of the load device based on the power consumption trend. Specifically, judge the power range of the load device based on the power consumption trend, and determine the preliminary connection threshold according to the power range, where the higher the power range, the lower the preliminary connection threshold. For example, when the device power is in the high power range (such as 80%-100%), the preliminary connection threshold is set to 5 minutes, allowing relatively frequent connection attempts; when the power is in the low power range (such as below 20%), the preliminary connection threshold is increased significantly to 60 minutes to reduce the number of connections.

[0058] S3: Analyze the current network status, and dynamically adjust the connection threshold in combination with the power consumption trend, network load, and real-time changes in device data. Specifically as follows:

[0059] In the step of analyzing the current network status, the following steps are included:

[0060] S301: Deploy network detection devices, such as network monitoring probes, at key nodes on the edge side network of the park.

[0061] S302: Based on the network detection device, real-time network load data is collected. The network monitoring probe can collect multi-dimensional network load metrics such as network bandwidth utilization (accurate to 0.1%), the number of connection requests (accurate to 1 request), data transmission delay (accurate to 1 millisecond), etc. at an interval of 100 milliseconds in real time;

[0062] S303: Summarize and analyze the network load data, and construct a network load evaluation model based on big data analysis algorithms. In this embodiment, a network load evaluation model is constructed using big data analysis technology based on support vector machine (SVM). The radial basis function (RBF) is used as the kernel function, and the kernel function parameter γ is determined by the cross-validation method, and the penalty parameter C is set to 1.0.

[0063] In the step of dynamically adjusting the connection threshold by combining the real-time changes of power consumption trends, network loads, and device data, the following steps are included: S311: Classify and evaluate the data collected by each load device in the park. For example, it can be divided into multiple different types such as key energy data (such as real-time power data of the power system, power data of energy storage devices, etc.), important device status data (such as the compressor operation frequency of large refrigeration equipment, lamp fault warning data of intelligent lighting systems, etc.), ordinary environmental data (such as PM2.5 concentration in the atmospheric environment, soil humidity), and auxiliary data (such as personnel activity monitoring data in the park, etc.).

[0064] S312: For each type of data, determine the basic importance weight of each type of data for the park operation through the analytic hierarchy process (AHP) combined with expert experience. For example, the basic importance weight of key energy data is set to 0.8, important device status data is 0.6, ordinary environmental data is 0.3, and auxiliary data is 0.1.

[0065] S313: Combine the functional attributes of each type of device and the basic information of the roles played in energy management, device operation and maintenance, or other operation links in the park to assign an initial connection priority to each load device. For example, for a device dedicated to energy transmission metering with extremely high real-time data requirements, since it provides key energy data and has a direct and significant impact on energy allocation decisions, the initial priority is set to 10; for an important device status data collection device such as a large refrigeration device that has a greater impact on the environmental comfort and energy consumption of the park, the initial priority is set to 7; for an ordinary environmental monitoring sensor that only provides general environmental information and has a low data update frequency, the initial priority is set to 4; and for auxiliary devices such as personnel activity monitoring, the initial priority is set to 1.

[0066] In some other embodiments, in addition to the above evaluation factors, the factor of data update frequency can also be combined.

[0067] S314: Monitor the power status of the load device, the network load status, and the real-time changes of the data. When the set conditions are met, re-evaluate the connection priority of the device according to the changed data. The set conditions can be that the device power changes significantly (such as the power drops by more than 10%), the network load fluctuates (such as the bandwidth utilization rate changes by more than 5% and the number of connection requests changes by more than 10 per second), or the data collected by the device is abnormal (such as the power value of the key energy data deviates from the normal range by 30%, and the key parameter change rate of the important device status data exceeds 20%, etc.).

[0068] In this embodiment, the weighted average algorithm is used to re-evaluate the connection priority, and the formula is as follows:

[0069] P new = P old + α × ΔE + β × ΔN + γ × ΔD

[0070] Where, P new is the adjusted priority, P old is the original priority, ΔE is the influence value of the power change on the priority, ΔN is the influence value of the network load change on the priority, ΔD is the influence value of the data abnormality degree on the priority, and α, β, γ are the weight coefficients corresponding to the power, network load, and data abnormality respectively, which can be set according to the actual needs of the park. For example, α = 0.3, β = 0.3, γ = 0.4.

[0071] For example, the original priority of a common environmental monitoring sensor is 4, but it suddenly detects that the PM2.5 concentration in the atmospheric environment exceeds the normal range by 50% (abnormality determination), and at the same time the network load is at a low level (bandwidth utilization rate is less than 30% and the number of connection requests is less than 5 per second), its priority will be increased. Assuming that there is no obvious change in the power, ΔE = 0, the low network load makes ΔN = 0.5 (calculated according to the set relationship between network load and priority influence), and the data abnormality degree of 50% makes ΔD = 1 (calculated according to the set relationship between data abnormality and priority influence), then P new = 4 + 0.3 × 0 + 0.3 × 0.5 + 0.4 × 1 = 4.55. After rounding, the priority will be increased to 5 to transmit the abnormal data to the park management system in time.

[0072] S315: Dynamically adjust the connection threshold of the load device with the adjusted priority based on the preset corresponding relationship between the priority and the connection threshold adjustment strategy; where the adjustment strategy is that when the priority increases, the connection threshold decreases, and vice versa.

[0073] Embodiment 2

[0074] Based on the content of the above-mentioned First Embodiment, the management method of this embodiment further includes step S4: further adjusting the connection priority and connection threshold, which specifically includes the following steps:

[0075] S41: Regarding each load device as an agent, the action space includes adjusting the connection priority and connection threshold, and the state space includes the device power state, network load state, data importance, and the connection decisions of surrounding agents;

[0076] S42: Using the Q-Learning algorithm, after the agent selects an action according to the current state, it obtains an immediate reward, where the reward function comprehensively considers the timeliness of data transmission, energy consumption savings, and the impact on the overall network stability;

[0077] For example, if the agent successfully transmits important data in a timely manner with low energy consumption and without affecting network stability, it obtains a higher reward; conversely, if high energy consumption or network congestion is caused by the connection, it obtains a negative reward. The agent updates the Q value according to the reward and the next state, and the formula is:

[0078]

[0079] where α is the learning rate (set to 0.1 - 0.3), γ is the discount factor (set to 0.9 - 0.99), s is the current state, a is the current action, r is the immediate reward, and s′ is the next state.

[0080] S43: After multiple rounds of iteration, each agent determines the final connection priority and connection threshold. For example, when a low-priority device (initial priority is 3) detects abnormal energy data (the abnormal determination is that the data value deviates from the normal range by more than 50%) and the network load is within an acceptable range (bandwidth utilization rate is less than 50% and the number of connection requests per second is less than 30), its connection priority will instantly increase to 7, and the connection threshold will correspondingly decrease to 0.5 times the original, and it will be connected first to transmit data. Conversely, if a high-priority device (initial priority is 9) is in a low battery state (battery level is lower than 30%) and the network load is too high (bandwidth utilization rate exceeds 80% and the number of connection requests per second exceeds 80), its priority will be appropriately reduced to 6, and the connection threshold will be increased to 2 times the original, and the connection will be postponed until the conditions improve.

[0081] Third Embodiment

[0082] Based on the content of the above-mentioned First Embodiment or Second Embodiment, the embodiment of the present invention further includes step S5 Feedback and Optimization, which specifically includes the following steps:

[0083] S51: After the device completes the connection, collect the actual energy consumption data (accurate to 0.001 watt-hours) and network response data (including connection establishment time accurate to 1 millisecond, data transmission success rate accurate to 0.1%, etc.) during this connection process;

[0084] S52: Feed the actual energy consumption data and network response data back to the power consumption prediction model and the network load assessment model respectively, and use the stochastic gradient descent algorithm to update the model parameters:

[0085] S53: Establish a loss function L(θ), and calculate the gradient of the loss function L(θ) with respect to the model parameter θ Among them, for the power consumption prediction model, the loss function can adopt the mean squared error formula:

[0086]

[0087] where n is the number of samples, y i is the actual remaining power value, is the predicted remaining power value, and θ is the model parameter: For the network load assessment model, an appropriate loss function can be constructed according to the prediction error of the network load index. For example, the cross-entropy loss function is used to evaluate the accuracy of network load status classification. For the power consumption prediction model of the LSTM model, the gradient is calculated step by step according to the chain rule: For the SVM network load assessment model, the gradient is calculated based on its mathematical principle and the form of the loss function.

[0088] S54: Update the model parameters according to the gradient, and the update formula is where η is the learning rate, initially set to 0.01. During the update process, in order to prevent gradient explosion, the gradient clipping technique is adopted to limit the norm of the gradient within a threshold, for example, the threshold is set to 5.

[0089] S55: After each round of parameter update, use the validation set data to evaluate the updated model to obtain evaluation metrics, monitor metrics such as the loss value and accuracy of the model on the validation set, and determine whether the evaluation metrics meet the preset conditions after several consecutive rounds of updates:

[0090] S56: If not, then reduce the learning rate and continue training until the evaluation metrics meet the preset conditions.

[0091] Preferably, it further includes steps of feedback and optimization of priorities, specifically: after the load device completes connection and data transmission, collect the data transmission effect after this connection (such as whether the data is processed in time, the degree of change in the park operation decision after processing, etc., quantified by the expert scoring method, with a full score of 10 points), the impact on the park operation decision, and adjust the weight of the data base importance. Use the gradient descent algorithm to optimize the weight and priority. For example, if a certain type of data is found to have a small impact on the park operation decision after multiple connection transmissions (the average data processing effect score is less than 3 points), its basic importance weight will be appropriately reduced by 10%-20%, thereby affecting the connection priority setting of related devices. If a certain device type has little impact on the park operation decision after 8 out of 10 past connections, its initial priority setting will be reduced by 2-3 levels.

[0092] Embodiment 4

[0093] Based on the methods of the above-mentioned Embodiment 1 to Embodiment 3, an intelligent connection management system for edge devices in a park is provided in an embodiment of the present invention. The management system includes a power consumption monitoring module, a network load monitoring module, and a device connection decision module.

[0094] The power consumption monitoring module is used to predict the power consumption trend of each load device within a set future time period;

[0095] The network load monitoring module is used to analyze the current network condition and evaluate the real-time change of network load:

[0096] The device connection decision module is used to determine the preliminary connection threshold of the load device according to the power consumption trend, and is also used to dynamically adjust the connection threshold in combination with the power consumption trend, network load, and real-time change of device data. Among them, the device connection decision module includes a device data processing unit, a priority evaluation unit, a connection decision unit, and a decision optimization unit.

[0097] The device data processing unit is used to classify and evaluate the data collected by the load device, and determine the basic importance weight of each type of data for the park operation;

[0098] The priority evaluation unit is used to assign an initial connection priority to each load device according to the device data processing unit, and is also used to re-evaluate the connection priority of the device according to the change data when the set conditions are met according to the power condition, network load state, and real-time change of the data of the load device;

[0099] The connection decision unit is used to dynamically adjust the connection threshold of the load device after priority adjustment according to the corresponding relationship between the preset priority and the connection threshold adjustment strategy; where the adjustment strategy is that when the priority increases, the connection threshold decreases, and vice versa;

[0100] The decision optimization unit is used to further optimize the connection priorities and connection thresholds determined by the priority evaluation unit and the connection decision unit. The optimization steps of the decision optimization unit include: regarding each load device as an agent, where the action space includes adjusting the connection priorities and connection thresholds, and the state space includes the device power state, network load state, data importance, and the connection decisions of surrounding agents; using the Q-Learning algorithm, after the agent selects an action according to the current state, it obtains an immediate reward, where the reward function comprehensively considers the timeliness of data transmission, energy consumption savings, and the impact on the overall network stability; after multiple rounds of iteration, each agent determines the final connection priorities and connection thresholds.

[0101] Furthermore, this embodiment further includes a feedback and optimization module, which is used to optimize the models of the power consumption prediction model and the network load evaluation model, as well as to optimize the judgment of the initial priorities.

[0102] It should be noted that for the specific implementation processes of the functions and roles of each module unit in this system, please refer to the corresponding steps in the methods of the above-mentioned Embodiments 1, 2, and 3, which will not be elaborated here.

[0103] The above are only preferred embodiments of the present invention and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or equivalent changes and modifications within the scope of the technical solution of the present invention. However, as long as the content does not depart from the technical solution of the present invention, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for intelligent connection management of campus edge devices, characterized in that: The management method comprises the following steps: Predict the power consumption trend of each load device within a set time period in the future; Determining a preliminary connection threshold of the load device based on the power consumption trend; Analyze the current network status and dynamically adjust the connection threshold based on power consumption trends, network load, and real-time changes in device data.

2. The method for intelligent connection management of campus edge devices according to claim 1, characterized in that: The step of predicting the power consumption trend of each load device within a future set time period includes: Obtain the current power and battery health data of the load device in real time, and retrieve the historical energy consumption data of the load device; Through the long short-term memory network deep learning algorithm, analysis is performed to establish a power consumption prediction model with current power, battery health data and historical energy consumption data as input. The training process uses an adaptive moment estimation optimization algorithm to adjust the weight and learning rate of the power consumption prediction model.

3. The method for intelligent connection management of campus edge devices according to claim 1, characterized in that: The step of determining the preliminary connection threshold of the load device includes: Determining the power range of the load device based on the power consumption trend; The preliminary connection threshold is determined based on the power interval, wherein a higher power interval means a lower preliminary connection threshold.

4. The method for intelligent connection management of campus edge devices according to claim 1, characterized in that: The management method further comprises the following steps: Deploy network detection equipment at key nodes of the campus edge network; Collect network load data in real time based on network detection equipment; The network load data is summarized and analyzed, and a network load evaluation model is constructed based on a big data analysis algorithm.

5. The method for intelligent connection management of campus edge devices according to claim 1, characterized in that: The step of dynamically adjusting the connection threshold includes: Classify and evaluate the data collected by each load device in the park, determine the basic importance weight of each type of data to the park operation, and assign initial connection priority to each load device based on the basic information of the role played by each type of equipment in the park; Monitor the power status of load equipment, network load status and real-time changes in data, and re-evaluate the connection priority of the device based on the changed data when the set conditions are met; Based on the correspondence between the preset priority and the connection threshold adjustment strategy, the connection threshold of the load device after the priority adjustment is dynamically adjusted; wherein the adjustment strategy is that the connection threshold decreases when the priority increases, and vice versa.

6. The method for intelligent connection management of campus edge devices according to claim 5, characterized in that: The management method further comprises the following steps: After the load equipment completes the connection and transmits data, the data transmission effect after this connection and the impact on the park operation decision are collected, and the importance weight of the data basis is adjusted.

7. The method for intelligent connection management of campus edge devices according to claim 4, characterized in that: The management method further comprises the following steps: After the device completes the connection, the actual energy consumption data and / or network response data during the connection are collected; Feeding back the actual energy consumption data and / or network response data to the power consumption prediction model and / or the network load assessment model, and updating the model parameters using a stochastic gradient descent algorithm; Establish the loss function L(θ) and calculate the gradient of the loss function L(θ) with respect to the model parameter θ Update the model parameters according to the gradient, and the update formula is: Where η is the learning rate; After each round of parameter update, the updated model is evaluated using the validation set data to obtain evaluation indicators, and it is determined whether the evaluation indicators meet the preset conditions after several consecutive rounds of updates; If not, the learning rate is reduced to continue training until the evaluation index meets the preset conditions.

8. The method for intelligent connection management of campus edge devices according to any one of claims 1 to 6, characterized in that: The management method further comprises the following steps: Each load device is regarded as an intelligent agent. The action space includes adjusting the connection priority and connection threshold. The state space includes the device power status, network load status, data importance, and the connection decision of surrounding intelligent agents. Using the Q-Learning algorithm, the agent receives an immediate reward after selecting an action based on the current state. The reward function comprehensively considers the timeliness of data transmission, energy saving, and the impact on the overall stability of the network. After multiple rounds of iterations, each agent determines the final connection priority and connection threshold.

9. A smart connection management system for edge devices in a park, characterized in that: The management system comprises: The power monitoring module is used to predict the power consumption trend of each load device within a set time period in the future; The network load monitoring module is used to analyze the current network status and evaluate the real-time changes in network load; The device connection decision module is used to determine the preliminary connection threshold of the load device according to the single quantity consumption trend, and is also used to dynamically adjust the connection threshold in combination with the power consumption trend, network load and real-time changes of device data.

10. The intelligent connection management system for campus edge devices according to claim 9, characterized in that: The device connection decision module includes: Equipment data processing unit, used to classify and evaluate the data collected by load equipment and determine the basic importance weight of each type of data to the park operation; A priority evaluation unit, used to assign an initial connection priority to each load device according to the device data processing unit, and to re-evaluate the connection priority of the device according to the change data when the set conditions are met according to the power status of the load device, the network load status and the real-time change of the data; The connection decision unit is used to dynamically adjust the connection threshold of the load device after priority adjustment according to the corresponding relationship between the preset priority and the connection threshold adjustment strategy; wherein the adjustment strategy is that the connection threshold decreases when the priority increases, and vice versa.

11. The intelligent connection management system for campus edge devices according to claim 9, characterized in that: The device connection decision module also includes: A decision optimization unit is used to further optimize the connection priority and connection threshold determined by the priority evaluation unit and the connection decision unit; the optimization steps of the decision optimization unit include: treating each load device as an intelligent agent, the action space includes adjusting the connection priority and connection threshold, and the state space includes the device power status, network load status, data importance and connection decisions of surrounding intelligent agents; using the Q-Learning algorithm, the intelligent agent obtains immediate rewards after selecting an action according to the current state, wherein the reward function comprehensively considers the timeliness of data transmission, energy saving and the impact on the overall stability of the network; after multiple rounds of iterations, each intelligent agent determines the final connection priority and connection threshold.

Images