Wireless sensor network data intelligent acquisition method and system

Abstract

The invention relates to the technical field of information acquisition, and discloses a wireless sensor network data intelligent acquisition method and system, and the method comprises the steps: constructing a state space; defining an action space; setting a reward function; building a deep reinforcement learning model; selecting an action to execute data acquisition according to the current state, and updating a Q value based on the new state and the obtained reward; the deep reinforcement learning method is introduced, so that the sensor nodes can autonomously learn the optimal data acquisition strategy, and the network resource utilization efficiency is remarkably improved; the system can dynamically adjust the collection behavior according to the current state, and resource waste is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of information collection, and more specifically, to a method and system for intelligently collecting data in a wireless sensor network. Background Art

[0002] Wireless sensor networks are distributed network systems composed of a large number of sensor nodes that can collaboratively sense, collect and process information within the network coverage area. With the rapid development of Internet of Things technology, wireless sensor networks have been widely used in smart agriculture, industrial monitoring, smart cities and other fields.

[0003] In practical applications, sensor nodes are usually powered by batteries and deployed in complex and changing environments. Traditional data collection methods often use fixed collection strategies, which have the following technical problems: Low resource utilization efficiency. Fixed collection strategies cannot be adjusted dynamically according to environmental changes and task requirements, which easily leads to waste of energy and storage resources. For example, high-frequency collection is maintained when the environment is stable, or important data cannot be collected in time at critical moments due to insufficient resources.

[0004] Data quality is difficult to guarantee. Due to the lack of real-time evaluation and feedback mechanism for the quality of collected data, it is difficult to balance the relationship between the frequency, accuracy and resource consumption of data collection under the condition of limited network resources, resulting in the completeness, accuracy and timeliness of collected data being difficult to meet application requirements.

[0005] Poor scenario adaptability. Different application scenarios have different requirements for data collection. For example, agricultural scenarios need to adjust the collection strategy according to the crop growth stage, and industrial monitoring scenarios need to encrypt sampling for equipment failure signs. Existing methods lack in-depth consideration of specific scenario needs and are difficult to provide accurate data collection services. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for intelligently collecting data in a wireless sensor network in order to solve the above problems.

[0007] The present invention provides a method for intelligently collecting data in a wireless sensor network, comprising: Construct the state space, including the sensor node's own state information, environment-related information, and task requirement information; Define the action space, including data collection frequency, collection accuracy, collection parameters and transmission mode; Set up a reward function to evaluate the impact of different actions on network resource utilization and data quality; Build a deep reinforcement learning model, using the deep Q network algorithm, with the input being the state space and the output being the Q value corresponding to each action in the action space; Perform training and optimization, select actions based on the current state to perform data collection, and update the Q value based on the new state and the reward obtained.

[0008] Furthermore, the sensor node's own status information includes power and storage capacity; Environmental related information includes temperature and humidity; Mission requirement information includes data collection frequency requirements and data accuracy requirements.

[0009] Further, the acquisition frequency is selected from a plurality of preset frequency values; The acquisition accuracy can be selected from different accuracy levels, including ultra-high accuracy, high accuracy, medium accuracy, and low accuracy; The acquisition parameters include the number of sampling points, sampling interval, and sampling duration; The transmission modes include real-time transmission, batch transmission, and conditional triggered transmission.

[0010] Furthermore, the reward function The calculation method is: ; in, is the current network resource utilization, is the total network resources, For the quality of the current collected data, For target data quality, , is the weight coefficient and satisfies .

[0011] Further, The calculation method is: ; in, , , is the weight coefficient of each indicator, For data integrity, the calculation method is: ; in is the number of packets successfully received, is the total number of packets that should be received; For data accuracy, the calculation method is: ; in is the measured value, is the true value; is the data timeliness, which is calculated as follows: ; in is the data transmission delay time, is the time-dependent attenuation coefficient; The calculation method is: ; in , , is the weight coefficient of each resource, and satisfies ; is the current power consumption rate, which is calculated as follows: ; in is the initial charge, is the current power; is the current storage usage, which is calculated as follows: ; in is the used storage space. is the total storage space; is the communication bandwidth occupancy rate, which is calculated as follows: ; in is the used bandwidth, is the total bandwidth.

[0012] Furthermore, the Q value update method is: ; in, is the learning rate, is the discount factor, To perform actions After the reward, Indicates in status All possible actions The maximum value of the corresponding Q value, Indicates in status Take action The Q value at that time.

[0013] Furthermore, it also includes: Add crop growth stage information to the state space; Add actions to the action space to adjust the combination of acquisition parameters for different growth stages; Add a quantitative value evaluation of the support level for crop growth to the reward function.

[0014] Furthermore, it also includes: Add equipment operation status information and equipment fault history data information library index in the state space; Add actions in the action space to adjust the key data collected according to the different fault precursor tendencies of the equipment; Add a quantitative value evaluation of the accuracy of equipment failure prediction in the reward function.

[0015] Furthermore, it also includes: Add traffic flow related information and traffic event information in the state space; Add actions to the action space to adjust the collection area and focus for different traffic events; Add a quantitative value evaluation of the traffic incident response effect to the reward function.

[0016] The present invention provides a wireless sensor network data intelligent collection system, which is used for storing computer-readable instructions. When the computer-readable instructions are read, the aforementioned wireless sensor network data intelligent collection method can be executed.

[0017] The beneficial effects of the present invention are as follows: the present invention enables sensor nodes to autonomously learn the optimal data collection strategy by introducing a deep reinforcement learning method, thereby significantly improving the efficiency of network resource utilization; the system can dynamically adjust the collection behavior according to the current state to avoid resource waste; A multi-dimensional reward function was designed, which incorporated network resource utilization and data quality into the optimization objectives, achieving a dynamic balance between resource efficiency and data quality. By evaluating the integrity, accuracy, and timeliness of collected data in real time, the quality of data collection was ensured to meet application requirements.

[0018] Corresponding technical solutions are proposed for different scenarios such as smart agriculture, industrial equipment monitoring, and smart transportation to enhance the system's scenario adaptability. In agricultural scenarios, information on crop growth stages is considered to implement accurate data collection strategies. In industrial monitoring scenarios, the collection focus is dynamically adjusted according to the equipment's operating status to improve fault prediction accuracy. In traffic monitoring scenarios, traffic incidents are responded to quickly, and the collection area and sampling frequency are optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a method for intelligent data collection of wireless sensor networks of the present invention; Figure 2 is the initial state data of the sensor node of the present invention; Figure 3 is an example of the action space options of the present invention; Figure 4It is the performance index of the system of the present invention after running for 24 hours; Figure 5 These are examples of optimal acquisition strategies for different scenarios of the present invention. DETAILED DESCRIPTION

[0020] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0021] A wireless sensor network data intelligent collection method and system, including the following embodiments: Example 1 A wireless sensor network data intelligent collection method, such as Figure 1 As shown, the following steps are included: Step 101: State space construction. Obtain the sensor node's own state information, environment-related information, and task requirement information, integrate the code to construct the state space Among them, the sensor node's own status information includes power , Storage Capacity etc.; environmental related information includes temperature ,humidity Environmental parameters such as the task requirements include data collection frequency , Data accuracy requirements Etc. State Space It is expressed as:

[0022] The specific implementation method is as follows: Sensor node status information collection: read the current power value through the power management module , normalized to the interval [0, 1]; obtain the remaining storage capacity through the storage management module , normalized to the interval [0, 1]; Environmental related information collection: temperature sensor collects ambient temperature , converted to standard unit ℃, the humidity sensor collects relative humidity , expressed as a percentage; Obtaining task requirement information: Read the required data collection frequency from the task configuration file ; Read data accuracy requirements from the task configuration file ; State space encoding: All information is standardized and mapped to a unified interval [0, 1]; one-hot encoding is used to process discrete variables; the encoded information of each dimension is concatenated to form a state vector.

[0023] Step 102: Action space definition. Define the action space corresponding to the data collection strategy , including adjusting the acquisition frequency , change the acquisition accuracy , adjust the acquisition parameters , Select the transfer mode Action Space It is expressed as:

[0024] in, Indicates Various acquisition frequency options, Indicates There are 2 options for acquisition accuracy. Indicates Collection parameter options, Indicates There are multiple transmission mode options; sensor nodes can select appropriate frequency, accuracy, parameters and transmission mode combinations from the action space according to the current state to perform data collection. This multi-dimensional action space setting enables sensor nodes to adjust data collection strategies more flexibly to meet the needs of different application scenarios; for example: Collection frequency You can choose from multiple preset frequency values, such as collecting data every 1 minute. , collected every 5 minutes , collected every 15 minutes , collected every 30 minutes wait; Collection accuracy You can choose from different accuracy levels, such as ultra-high accuracy , high precision , medium precision , low precision wait; Acquisition parameters Including sampling points , sampling interval , Sampling duration wait; Transmission Mode Including real-time transmission , Batch Transfer , Conditional Trigger Transmission wait.

[0025] Step 103: Reward function calculation. Calculate the reward function , used to evaluate the impact of different actions on network resource utilization and data quality. Reward function Defined as:

[0026] in, is the current network resource utilization, is the total network resources, For the quality of the current collected data, For target data quality, , is the weight coefficient and satisfies .

[0027] The specific implementation method is as follows: Calculate the current power consumption rate :

[0028] in is the initial charge, is the current power; Calculate current storage usage :

[0029] in is the used storage space. is the total storage space; Calculate communication bandwidth utilization :

[0030] in is the used bandwidth, is the total bandwidth; Comprehensively calculate the current network resource utilization:

[0031] in , , is the weight coefficient of each resource, and satisfies ; Data quality assessment: Calculating Data Integrity :

[0032] in is the number of packets successfully received, is the total number of packets that should be received; Calculate data accuracy :

[0033] in is the measured value, is the true value; Calculate data timeliness :

[0034] in is the data transmission delay time (unit: seconds), is the time-dependent attenuation coefficient; Comprehensively calculate the quality of the current collected data:

[0035] in , , is the weight coefficient of each indicator, and satisfies ; Step 104: Build a deep reinforcement learning model. Select the Deep Q Network (DQN) algorithm, build a neural network, and input the state space , the output is the action space The Q value corresponding to each action in , .

[0036] Step 105: Training and optimization. During the operation of the wireless sensor network, the sensor nodes , from the action space Select Action Perform data collection. According to the new state after executing the action and rewards received , according to the Q value update rule of the DQN algorithm, the training optimization is performed. The update rule is:

[0037] in, is the learning rate (value range [0, 1]), is the discount factor (value range [0, 1]), Indicates in status All possible actions The maximum value of the corresponding Q value, Indicates in status Take action The Q value at that time.

[0038] Through the above steps, the sensor nodes in the wireless sensor network can autonomously adjust the data collection strategy according to environmental changes and task requirements, thereby improving network resource utilization and data quality.

[0039] The following is an example of actual application data in a factory environment monitoring scenario: from Figure 2-Figure 5 It can be seen that the intelligent acquisition strategy of the present invention has the following advantages over the traditional fixed strategy: 1. Energy efficiency is improved by 14.3%, effectively extending the service life of sensor nodes; 2. Data quality (completeness, accuracy) improved by more than 6% on average; 3. Network resource utilization is more reasonable, with load and storage utilization rates reduced by 23.5% and 26.7% respectively; 4. Ability to adaptively adjust collection strategies according to different scenarios to achieve a dynamic balance between resource utilization and data quality.

[0040] Example 2 On the basis of Example 1, the influence of crop growth stage on data collection is also considered to implement accurate data collection strategy to support agricultural precision management. The following steps are included: Step 201: state vector construction. Based on the state space of Example 1, crop growth stage information is added , which can be divided into sowing period, seedling period, growth period, maturity period, etc. Agricultural scene state vector It is expressed as:

[0041] in, is the light intensity, The nutrient content of soil.

[0042] Step 202: Action set definition. Based on the action space of Example 1, add actions to adjust the collection parameter combination for different growth stages, such as focusing on collecting temperature and humidity related data during the seedling stage, and focusing on collecting light and soil nutrient data during the growth period. Agricultural scene action set It is expressed as:

[0043] in, Indicates the acquisition parameter combinations corresponding to different growth stages.

[0044] Step 203: Setting the reward function. Based on the reward function of Example 1, add the quantitative value of the support degree of crop growth and the quantitative value of the support degree for crop growth stages under ideal conditions Agricultural scene reward value Defined as:

[0045] in, , , is the weight coefficient and satisfies .

[0046] The specific implementation method is as follows: Quantification of the degree of support for crop growth: Calculate temperature and humidity suitability :

[0047] in, is the measured temperature value, is the optimum temperature value, is the permissible temperature range, is the measured humidity value, is the optimum humidity value, is the permissible range of humidity; Calculate light suitability :

[0048] in, is the measured light intensity, For the optimal light intensity, is the permissible range of light intensity; Calculate soil nutrient suitability :

[0049] in, To measure the nutrient content, For the optimal nutrient content, is the permissible range of nutrient content; Comprehensive calculation :

[0050] in , , is the weight of each indicator, and satisfies ; set up: Set the ideal temperature and humidity range according to crop variety and growth stage; Set the ideal light intensity range according to crop variety and growth stage; Set ideal soil nutrient ranges based on crop variety and growth stage; Weight coefficient setting: Dynamic adjustment according to crop growth stage , , ; Increase during key growth periods Weights to ensure growing conditions; Increase appropriately when resources are tight weights to conserve resources; Step 204: Deep reinforcement learning model building: As in Example 1, the DQN algorithm is selected to build a neural network.

[0051] Step 205: Training and optimization: Similar to Example 1, training and optimization are performed according to the Q value update rule of the DQN algorithm.

[0052] Through the above steps, the sensor nodes in the smart agricultural wireless sensor network can collect data more accurately according to the growth stage of crops, improve network resource utilization and data quality, and provide more effective data support for the refined management of crops.

[0053] Example 3 On the basis of Example 1, the special requirements of equipment failure prediction are also considered to realize a data collection strategy that dynamically adapts to the change of equipment from normal operation to the occurrence of failure signs. The following steps are included: Step 301: state vector construction. Based on the state space of Example 1, add equipment operation state information and equipment fault history data information library index. Equipment operation state information includes equipment speed , Temperature of key components , Equipment operation time Etc.; Equipment failure history data information database Contains various data features of previous similar equipment failures. Industrial scenario state vector It is expressed as:

[0054] in, is the vibration intensity, Represents the index of related data in the equipment failure history data information library.

[0055] Step 302: Action set definition. Based on the action space of Example 1, add actions to adjust the key data collected according to the different fault prediction tendencies of the equipment, such as focusing on collecting vibration intensity and key component temperature data when the equipment speed is abnormal. Industrial scenario action set It is expressed as:

[0056] in, It represents the key data collection combination for different fault prediction tendencies.

[0057] Step 303: Reward function setting. Based on the reward function of Example 1, add a quantitative value of the accuracy of equipment failure prediction Quantitative value of equipment failure prediction accuracy under ideal conditions . Industrial scene reward value Defined as:

[0058] in, , , is the weight coefficient and satisfies .

[0059] The specific implementation is as follows: Quantification of fault prediction accuracy: Calculate equipment operating parameter deviation :

[0060] in, For the current parameter values, is the normal parameter value, is the parameter allowable range, For the parameter weights, is the total number of parameters; Calculate the fault feature matching degree :

[0061] in, For the current Fault characteristics, is the feature in the historical fault database, is the total number of features, represents the number of features in the historical fault database, is the feature matching function used to calculate the feature The degree of match with historical data; Comprehensive calculation :

[0062] in , is the weight coefficient and satisfies ; set up: Set the normal range of parameters according to the equipment type and operating conditions; Establish a comprehensive fault feature database; Weight coefficient setting: Dynamic adjustment based on equipment operating status , , ; Increase when signs of failure are obvious weights to improve prediction accuracy; Increase appropriately when resources are tight weights to conserve resources; Step 304: Deep reinforcement learning model building: As in Example 1, the DQN algorithm is selected to build a neural network.

[0063] Step 305: Training and optimization. Similar to Example 1, training and optimization are performed according to the Q value update rule of the DQN algorithm.

[0064] Through the above steps, the wireless sensor network for industrial equipment monitoring can better collect data reflecting the potential failure risks of equipment, dynamically adjust the data collection strategy to adapt to the change of equipment from normal to failure signs, and improve the accuracy of equipment failure prediction.

[0065] Example 4 On the basis of Example 1, the impact of traffic events on data collection strategies is also considered to achieve rapid response and data collection that adapts to special traffic conditions. The following steps are included: Step 401: constructing a state vector. Based on the state space of Example 1, traffic flow related information and traffic event information are added. Traffic flow related information includes vehicle flow , vehicle speed , vehicle density Traffic incident information Including the location and type of events such as traffic accidents and road construction. Traffic scene state vector It is expressed as:

[0066] in, is the weather condition (sunny, rainy, etc.).

[0067] Step 402: Action set definition. Based on the action space of Example 1, add actions to adjust the collection area and focus for different traffic events. For example, when a traffic accident occurs, focus on collecting data such as traffic flow and speed within a certain range around the accident. Traffic scene action set It is expressed as:

[0068] in, Indicates the collection area and focus adjustment combination for different traffic events.

[0069] Step 403: Reward function setting. Based on the reward function of Example 1, a traffic incident response effect quantification value is added. and the quantitative value of the effectiveness of traffic incident response under ideal conditions Traffic scene reward value Defined as:

[0070] in, , , is the weight coefficient and satisfies .

[0071] The specific implementation method is as follows: Quantification of traffic incident response effects: Calculate traffic flow rate of change :

[0072] in, is the current traffic volume, Normal traffic flow, vertical lines Indicates taking the absolute value; Calculate the average vehicle speed change rate :

[0073] in, is the current average vehicle speed, is the normal average speed, vertical line Indicates taking the absolute value; Calculating incident response time :

[0074] in, Time to take countermeasures, The time when the event occurred; Comprehensive calculation :

[0075] in, , , is the weight coefficient and satisfies The maximum allowed response time set up: Set traffic flow and speed change thresholds based on road grade and traffic event type Set the maximum allowed response time ; Weight coefficient setting: Dynamic adjustment based on the severity of traffic incidents , , Increase in traffic incidents Weighting to improve response effectiveness Increase appropriately when resources are tight Weights to save resources Step 404: Deep reinforcement learning model building: Similar to Example 1, the DQN algorithm is selected to build a neural network.

[0076] Step 405: Training and optimization: Similar to Example 1, training and optimization are performed according to the Q value update rule of the DQN algorithm.

[0077] Through the above steps, the smart city traffic flow monitoring wireless sensor network can quickly adjust the data collection strategy when a traffic incident occurs, obtain key information more effectively, and provide strong data support for the traffic management department to timely guide traffic and handle traffic incidents.

[0078] The wireless sensor network data intelligent collection method provided by the present invention enables sensor nodes to autonomously adjust data collection strategies according to environmental changes and task requirements by introducing deep reinforcement learning, thereby improving network resource utilization and data quality. At the same time, corresponding technical solutions are proposed for different application scenarios such as smart agriculture, industrial equipment monitoring, and smart city traffic flow monitoring, which fully consider the special needs of each scenario and achieve accurate optimization of data collection strategies. The present invention can be widely used in various types of wireless sensor networks and has significant technical advancement and practical value.

[0079] In at least one embodiment of the present invention, a wireless sensor network data intelligent collection system is provided, which is used to store computer-readable instructions. When the computer-readable instructions are read, the aforementioned wireless sensor network data intelligent collection method can be executed.

[0080] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A wireless sensor network data intelligent collection method, characterized in that: include: Construct the state space, including the sensor node's own state information, environment-related information, and task requirement information; Define the action space, including data collection frequency, collection accuracy, collection parameters and transmission mode; Set up a reward function to evaluate the impact of different actions on network resource utilization and data quality; Build a deep reinforcement learning model, using the deep Q network algorithm, with the input being the state space and the output being the Q value corresponding to each action in the action space; Perform training and optimization, select actions based on the current state to perform data collection, and update the Q value based on the new state and the reward obtained.

2. A wireless sensor network data intelligent collection method according to claim 1, characterized in that: The sensor node’s own status information includes power and storage capacity; Environmental related information includes temperature and humidity; Mission requirement information includes data collection frequency requirements and data accuracy requirements.

3. The method for intelligently collecting data in a wireless sensor network according to claim 1, characterized in that: The acquisition frequency is selected from multiple preset frequency values; The acquisition accuracy can be selected from different accuracy levels, including ultra-high accuracy, high accuracy, medium accuracy, and low accuracy; The acquisition parameters include the number of sampling points, sampling interval, and sampling duration; The transmission modes include real-time transmission, batch transmission, and conditional triggered transmission.

4. The method for intelligently collecting data in a wireless sensor network according to claim 1, characterized in that: Reward Function The calculation method is: ； in, is the current network resource utilization, is the total network resources, The quality of the current collected data is For target data quality, , is the weight coefficient and satisfies .

5. A wireless sensor network data intelligent collection method according to claim 4, characterized in that: The calculation method is: ； in, , , is the weight coefficient of each indicator, For data integrity, the calculation method is: ； in is the number of packets successfully received, is the total number of packets that should be received; For data accuracy, the calculation method is: ； in is the measured value, is the true value; is the data timeliness, which is calculated as follows: ； in is the data transmission delay time, is the time-dependent attenuation coefficient; The calculation method is: ； in , , is the weight coefficient of each resource, and satisfies ; is the current power consumption rate, which is calculated as follows: ； in is the initial charge, is the current power; is the current storage usage, which is calculated as follows: ； in is the used storage space. is the total storage space; is the communication bandwidth occupancy rate, which is calculated as follows: ； in is the used bandwidth, is the total bandwidth.

6. A wireless sensor network data intelligent collection method according to claim 1, characterized in that: The Q value update method is: ； in, is the learning rate, is the discount factor, To perform actions After the reward, Indicates in status All possible actions The maximum value of the corresponding Q value, Indicates in status Take action The Q value at that time.

7. A wireless sensor network data intelligent collection method according to claim 1, characterized in that: Also includes: Add crop growth stage information to the state space; Add actions to the action space to adjust the combination of acquisition parameters for different growth stages; Add a quantitative value evaluation of the support level for crop growth to the reward function.

8. The method for intelligently collecting data in a wireless sensor network according to claim 1, characterized in that: Also includes: Add equipment operation status information and equipment fault history data information library index in the state space; Add actions in the action space to adjust the key data collected according to the different fault precursor tendencies of the equipment; Add a quantitative value evaluation of the accuracy of equipment failure prediction in the reward function.

9. The method for intelligently collecting data in a wireless sensor network according to claim 1, characterized in that: Also includes: Add traffic flow related information and traffic event information in the state space; Add actions to the action space to adjust the collection area and focus for different traffic events; Add a quantitative value evaluation of the traffic incident response effect to the reward function.

10. A wireless sensor network data intelligent collection system, characterized in that: It is used to store computer-readable instructions, and when the computer-readable instructions are read, a method for intelligently collecting data in a wireless sensor network as described in any one of claims 1 to 9 can be executed.

Images