CN104168641B - A kind of wireless sensor network time synchronization method based on temperature sensing - Google Patents

Abstract

The invention discloses a wireless sensor network time synchronization method based on temperature perception. The method includes the following steps: determining the sensitivity factor, determining the interval of the sensitivity factor, and updating the local time. This algorithm takes into account the impact of the node's current ambient temperature change on the node's frequency offset when estimating the frequency offset, and improves the accuracy of the frequency offset estimation. At the same time, because the algorithm mainly relies on local information in the process of time synchronization, the number of information transmissions is greatly reduced, thereby reducing energy consumption to a large extent, and reducing the accumulation of errors caused by layer-by-layer information transmission. Finally, because the algorithm is less dependent on information transmission, it solves the problem of unstable communication caused by bad weather and dynamic changes of node positions in the field environment.

Description

A Time Synchronization Method for Wireless Sensor Networks Based on Temperature Sensing

technical field

The invention relates to the technical field of wireless networks, in particular to a temperature-sensing-based time synchronization method for a wireless sensor network, which is suitable for large-scale regional monitoring wireless sensor network applications such as wild animal monitoring and earthen site monitoring.

Background technique

As an important supporting technology of wireless sensor networks, time synchronization has been widely used, such as data fusion technology, sleep scheduling technology, TOA-based positioning technology and target tracking, all of which require the nodes of the entire network to maintain time synchronization. In a large-scale sensor network, there are many network nodes, and the energy, processing power, and bandwidth of the nodes are relatively limited, and the network environment is relatively harsh. Therefore, this requires the sensor network time synchronization algorithm to have low communication overhead, low computational complexity, Good scalability and robustness.

In the process of large-scale monitoring (such as wild animals, earthen sites, etc.), different data (text data, sound data, video data, etc.) from different sensors need to be combined, and through a series of statistics and analysis, the final result is Available environmental information and speculated about possible events. In the process of merging multiple data, it is necessary to synchronize the time of each node collecting data, otherwise wrong time information will be obtained, which will eventually lead to wrong analysis results. In addition, due to the energy-constrained nature of sensor networks, nodes need to be periodically dormant to reduce energy consumption. This requires the nodes of the entire network to adjust their sleep cycle according to a specific rule, so as to ensure the correct transmission of data. However, the out-of-sync time between nodes will cause the nodes to go to sleep at the wrong time, thereby affecting the success rate of data transmission. In the prior art, in order to ensure the time synchronization between the nodes of the whole network, there are already many time synchronization strategies in the wireless sensor network:

The first category: time synchronization methods based on packet switching

This method first performs time synchronization between a pair of nodes by exchanging time stamps between nodes, and then performs layer-by-layer synchronization through the method of network layering, and finally achieves time synchronization of the entire network. There are three disadvantages in this method: 1) Since this method utilizes frequent timestamp exchange for time synchronization, it will introduce a large amount of communication overhead. In a wireless sensor network, the proportion of communication overhead in the total overhead is much higher than the overhead caused by computing overhead and data collection, so this method will cause a large loss of node energy. 2) Since the time stamp is transmitted layer by layer in the network, it will cause error accumulation, thereby affecting the time synchronization accuracy. 3) Since the sensor network uses a cheap crystal oscillator, the crystal oscillator is easily affected by the working environment such as temperature, voltage, vibration, etc., but this method does not take this into consideration.

The second category: time synchronization methods based on external periodic signals

In this method, all nodes in the entire network adjust their clock frequency according to a unified periodic signal. Such periodic signals include: wifi signals, broadcast signals, light signals emitted by sunlight, and so on. This method mainly relies on local information in the synchronization process, which greatly reduces the exchange of time stamps, reduces energy consumption, and reduces error accumulation. The defects of this method are as follows: 1) There are certain restrictions on the environment, and this type of method is not applicable to the field environment where various signals cannot reach. Moreover, the method of synchronizing according to fluorescent lamps requires that the sensor network must work in an indoor environment. 2) WIFI signals and broadcast signals require additional hardware equipment to receive, which not only increases the economic cost, but also requires high energy consumption support, which is not suitable for large-scale deployment. 3) This method also does not take into account the impact of the working environment on cheap crystal oscillators.

Contents of the invention

The sensor network time synchronization method working in a large-scale field environment is significantly different from the method in a normal environment. Aiming at the current situation that the existing synchronization method cannot be applied to a large-scale network, this invention proposes a wireless sensor network based on temperature perception The time synchronization method enables the synchronization process to meet the requirements of high precision and low energy consumption in large-scale environments in the field.

In order to realize above-mentioned task, the technical scheme that the present invention adopts is:

A method for synchronizing time in a wireless sensor network based on temperature perception, comprising the following steps:

Note that R is the reference node in the wireless sensor network, and N is any sensor node except the reference node. After the network is initialized, the node N repeatedly executes the following cycle, which includes steps 1 to 3:

Step 1, determine the sensitivity factor

Step S10, the node N sends a time synchronization request packet to the reference node R;

Step S11, after receiving the time synchronization request data packet, node R returns four response data packets to node N in turn: M ₀ , M ₁ , M ₂ , M ₃ , each data packet records the time when the data packet is sent The local time of node R is time(R) ₀ ~time(R) ₃ respectively; the interval between M ₀ and M ₁ , M ₂ and M ₃ is 1s; the interval between M ₁ and M ₂ is 10min;

Step S12, while receiving the data packets M ₀ -M ₃ , the node N records its own local time time ₀ -time ₃ and the current ambient temperature temp ₀ -temp ₃ of the node N;

Step S13, node N calculates its frequency offset skew ₁ and skew ₃ at time ₁ and time ₃ :

Step S14, the node N calculates the current sensitivity factor TSF value according to the frequency offset and temperature information:

In formula 2, Temp is the standard temperature, the value is 25°C;

Step 2, determine the interval of sensitivity factor

Step S20, the node N obtains the ambient temperature T ₁ at the moment, and the ambient temperature of the node N at this moment in the last cycle is T _pre , then the temperature change rate DT of the node N is:

In formula 3, d _pre is the value of the sensitivity factor interval d obtained in step S22 of the previous cycle;

Step S21, node N calculates the current cumulative error value error:

Step S22, the node N sets the sensitivity factor interval d, the method is:

In formula 5, μ=150～900μs, λ=0.6～1.4℃, d _std =20min;

In step S23, the node N sets the current frequency offset value skew to skew ₃ , and after Δt, go to step S31, 100s<Δt<10000s;

Step 3, local time update

In step S30, the node N obtains the ambient temperature T ₂ where it is located at the moment, and calculates the current frequency offset of the node according to the sensitivity factor TSF calculated in step S14:

skew＝TSF·(T-Temp) ² (Formula 6)

In the above formula, T represents time;

Step S31, node N calculates the current phase offset:

In Formula 7, skew _pre is the current frequency offset value obtained in step S23 or step S30 in the previous cycle, and offset _pre is the current phase offset value calculated in step S31 in the previous cycle;

Step S32, if the current phase offset of node N satisfies:

Then node N updates its own local time, and the updated local time clock is:

clock＝clock _pre +offset (Formula 9)

In Equation 8 and Equation 9, ε is the local clock cycle, and clock _pre is the local time before the update; after the update is completed, node N clears offset _pre to zero;

Step S33, the node N checks the timer, if the duration d has not expired, go to step S30 after sleeping for Δt time; otherwise, complete this cycle, go to step S10 to start the next cycle.

Compared with the prior art, the present invention has the following technical advantages:

1. Reduced energy consumption;

First of all, since the nodes mainly rely on local information to estimate the clock frequency offset and update the local clock during the synchronization process, the communication overhead is greatly reduced.

Secondly, because the nodes are time-synchronized according to their ambient temperature information, and the acquisition of temperature does not require additional hardware devices, only temperature sensors are needed, thus reducing the energy consumption of signal reception.

2. Improved synchronization accuracy;

First of all, the node takes into account the influence of the working environment (temperature) on the crystal oscillator of the node during the synchronization process, and compensates for this influence, so that the frequency offset change caused by temperature can be avoided, thereby reducing the accumulation of clock phase offset.

Secondly, since the time synchronization method basically does not require layer-by-layer transmission of time stamps, the accumulation of synchronization errors is reduced.

3. Improved robustness

Also, since nodes mainly rely on local information in the process of time synchronization, rather than the exchange of time stamps, this method has lower requirements on communication conditions. When the communication equipment of the node is abnormal and cannot work, or the position of the network node changes dynamically and cannot maintain communication with the reference node, the node can perform time synchronization according to the local voltage information, thus improving the robustness.

Description of drawings

Fig. 1 is a flow chart of the inventive method;

Figure 2 is a schematic diagram of the "temperature-frequency offset" relationship;

Fig. 3 is a schematic diagram of the frequency offset phase offset estimation process and the temperature sensitivity factor TSF estimation process cycle relationship diagram;

Fig. 4 is a schematic diagram of simulated temperature variation in the simulation experiment;

Fig. 5 is the influence of the precision control parameter μ on the re-λ synchronization interval and the number of time stamp exchanges and the experimental results of determining the optimal value of μ;

Fig. 6 is the influence of the temperature adjustment factor λ on the TSF estimation interval and the number of time stamp exchanges and the experimental results of determining the optimal value of λ;

Fig. 7 is a comparison experiment result diagram of frequency offset estimation accuracy of dynamic period TSFB algorithm, fixed period TSFB algorithm and EACS algorithm;

Fig. 8(a) is a comparison experiment result of energy consumption between TSFB algorithm and FTSP algorithm under different precision control parameter μ values;

Figure 8(b) is a diagram of the energy consumption comparison experiment results of the TSFB algorithm and the FTSP algorithm under different values of the temperature adjustment factor λ;

Figure 9 is a diagram of the experimental results of the robustness comparison between the TSFB algorithm and the FTSP algorithm;

detailed description

In order to ensure the smooth progress of the dormant scheduling process and the correctness of the data processing results in the data fusion stage in large-scale monitoring such as wildlife monitoring and earthen site monitoring, the applicant needs to establish a time synchronization strategy with high precision and low energy consumption. However, there are two problems in the large-scale field environment compared with the ordinary network environment: 1) the large network scale leads to a rapid increase in the energy consumption of node communication and the cumulative error of time synchronization; 2) the dynamic change of the environment leads to unstable communication between nodes. Stable, unable to maintain continuous normal communication.

Aiming at the fact that the existing time synchronization methods are not suitable for large-scale field environments, the present invention proposes a time synchronization method based on temperature-aware sensor networks, which utilizes the high correlation between node clock frequency deviation and temperature (as shown in Figure 2) Time synchronization is performed so that the time synchronization algorithm can still achieve high precision, low energy consumption and high robustness under such special network conditions.

The present invention needs to periodically estimate and update the sensitivity factor TSF of the sensor nodes during the time synchronization process. The sensitivity factor TSF here indicates the sensitivity of the node to the temperature change of its environment. During the interval between two TSF estimates, the node estimates and compensates the frequency offset according to its ambient temperature and the current TSF estimate. The frequency offset here refers to the frequency offset of each node relative to the same reference node. At the same time, the node can adjust the TSF estimation interval for the node according to the accumulation of estimation errors and temperature changes, so as to achieve a balance between energy consumption and estimation accuracy.

One, the detailed steps of the inventive method

The present invention proposes a time synchronization method for a wireless sensor network based on temperature perception. During the synchronization process, the method continuously corrects the clock frequency offset of the node according to the local temperature value and the sensitivity factor TSF, and compensates the clock phase offset. At the same time, The sensitivity factor TSF is estimated and updated periodically according to the accumulation of errors and the temperature change. As shown in Figure 1, the method includes the following steps:

A method for time synchronization of a wireless sensor network based on temperature perception, which utilizes the relationship between the temperature of the wireless sensor and the frequency offset to perform time synchronization, comprising the following steps:

Note that R is the reference node in the wireless sensor network, and N is any sensor node except the reference node. After the network is initialized, the node N repeats the following cycle, that is, all nodes are repeatedly executed according to the cycle; the time synchronization process is accompanied by As long as the network life cycle is not over, the time synchronization process will continue periodically; the cycle process is shown in Figure 3, and this cycle includes steps 1 to 3:

Step 1, determine the sensitivity factor

In order to enable the nodes to compensate the time frequency offset according to the current temperature in time during the synchronization process, the sensitivity factor TSF needs to be estimated and updated periodically during the time synchronization process of the wireless sensor network nodes. The estimation of TSF mainly depends on the variation of node temperature and the corresponding variation of frequency offset during the estimation process. In order to obtain its frequency offset information, a node needs to exchange time stamps with the reference node R; the reference node here is a standard node, which is determined before deployment, and can be an ordinary node or a base station; the nodes of the entire network refer to The time of the node is time-synchronized for the label: Here, take any node N in the sensor network except the standard node as an example:

Step S10, the node N sends a time synchronization request packet to the reference node R, informing the reference node that a synchronization process is required;

Step S11, after receiving the time synchronization request data packet, node R returns four response data packets to node N in turn: M ₀ , M ₁ , M ₂ , M ₃ , each data packet records the time when the data packet is sent The local time of node R is time(R) ₀ ~time(R) ₃ respectively; (for example, M ₀ contains the time when node R sends M ₀ : time(R) ₀ ) M ₀ and M ₁ , M ₂ and The interval between M ₃ is 1s; the interval between M ₁ and M ₂ is 10min;

Step S12, in order to obtain the time difference between the node's own time and the reference node R, the node N records its own local time time ₀ to time ₃ and the current environment of the node N when it receives the data packets M ₀ to M ₃ Temperature temp ₀ ~ temp ₃ (that is, a data packet corresponds to a time value and a temperature value, such as when receiving a data packet M ₀ , record the local time time ₀ and the ambient temperature temp ₀ at this time); the temperature information here is determined by the temperature Obtained by the sensor, mainly used for the measurement of the temperature variation within the sensitivity factor estimation cycle;

Step S13, node N calculates its frequency offset skew ₁ and skew ₃ at time ₁ and time ₃ according to the information in the data packets M ₀ to M ₃ and its own local time information time ₀ to time ₃ , the calculation method is as follows:

Step S14, the node N calculates its current sensitivity factor TSF value according to the frequency offset and temperature information, and the calculation formula is:

In formula 2, since the acquisition time of temp ₀ and temp ₁ , temp ₂ and temp ₃ are relatively close (interval 1s), the sensitivity factor TSF is estimated by using the average temperature Temp _a and Temp _b during this period, namely

Temp is the standard temperature, the value is 25°C;

Step 2, determine the interval of sensitivity factor

Since the change of frequency offset of nodes will be different with the change rate of temperature, when the temperature changes significantly, the change of frequency offset of nodes will also be more obvious. At this time, in order to ensure the accuracy of synchronization, it is necessary to shorten the sensitivity factor TSF The period is estimated, so that a high synchronization accuracy can be guaranteed. When the temperature is relatively stable, the frequency offset of the node is almost unchanged. At this time, in order to reduce the communication overhead, it is necessary to extend the estimation period of the sensitivity factor TSF. Therefore, the determination of the sensitivity factor interval needs to depend on the rate of change of temperature and the accumulation of errors:

Step S20, calculating the temperature change rate DT: the node N obtains its current ambient temperature T ₁ , and the ambient temperature of the node N at this moment in the last cycle is T _pre , then the temperature change rate DT of the node N is:

In Equation 3, d _pre is the value of the sensitivity factor interval d obtained in step S22 of the previous cycle; Note: Since this solution is repeatedly executed after network initialization, the parameter calculation of one cycle depends on the previous cycle When calculating the temperature change rate DT in formula 3, the ambient temperature and the sensitivity factor interval of the previous cycle are required. That is, if the current ambient temperature in this cycle is T ₁ , then T ₁ will be used as T _pre when calculating DT parameters in the next cycle, the sensitivity factor interval d of this cycle will be used as d _pre in the next cycle, and other parameters will be based on By analogy; when the parameters are calculated for the first time after the network is initialized and the parameter values of the previous cycle are needed, a parameter value of the previous cycle is randomly set for the current parameter calculation, and the parameters of the previous cycle can be continuously corrected in subsequent cycles.

Step S21, node N calculates the current cumulative error value error:

In Formula 4, node N estimates the current error based on the mean value of the difference between time ₃ and time(R) ₃ and the difference between time ₂ and time(R) ₂ between the last two timestamp exchange results with the reference node;

Step S22, node N sets the sensitivity factor interval d according to the results of step S21 and step S20, and starts timing from this moment. The calculation formula of d is as follows:

In Formula 5, μ and λ are the error control factor and temperature adjustment factor respectively, and the values are: μ=150～900μs, λ=0.6～1.4℃; according to the results of Experiment 1 and Experiment 2, when μ=300us, λ= At 1°C, this method has the best performance; d _std is the standard interval, usually d _std = 20min;

It can be seen from Equation 5 that when the ambient temperature changes around the node or the synchronization error of the node increases, the TSF estimation interval will be shortened accordingly to ensure the synchronization accuracy; otherwise, the interval will be increased to reduce the communication overhead;

Step 23, for the convenience of the subsequent phase offset estimation process, the node N sets the current frequency offset value skew to skew ₃ , and then transfers to step S31 after the duration of Δt, where Δt is the time estimation interval; Δt satisfies 100s<Δt<10000s, that is, the network setting , as long as a Δt is selected within this range, the requirements of this scheme are met;

Step 3, local time update

After the node N updates its temperature sensitivity factor TSF and the sensitivity factor interval d, it turns to the local time update process. The clock frequency offset estimation in this process mainly depends on the temperature sensitivity factor TSF obtained in step 1 and the node's current time. Ambient temperature, the duration of the time update process is d.

Step S30, in order to estimate the clock frequency offset skew (due to the different time, the skew in this step and the skew in step S23 are not the same value, but two values in different situations), node N first obtains its ambient temperature at the moment T ₂ , and then estimate the current frequency offset of the node according to the sensitivity factor TSF calculated in step S14 of this cycle:

skew＝TSF·(T-Temp) ² (Formula 6)

In the above formula, T represents time;

Step S31, node N calculates the current phase offset:

In formula 7, skew _pre is the current frequency offset value obtained in step S23 or step S30 of the previous cycle, that is, if jumping from step S23 to step S31, the value of skew _pre is the frequency offset value skew in step S23 of the previous cycle; If it is executed from step S30 to step S31, the value of skew _pre is the frequency offset value skew in step S30 of the previous cycle; offset _pre is the current phase offset value calculated in step S31 of the previous cycle;

Step S32, if the current phase offset of node N satisfies Formula 8, it is considered that the phase offset of node N is too large and will affect network applications. At this time, node N needs to update its own local time, otherwise no update:

In Equation 8, ε is the local clock period, and the value of ε can be appropriately increased when the synchronization accuracy is not strictly required; the updated local time clock is:

clock＝clock _pre +offset (Formula 9)

In Formula 9, clock _pre is the local time before the update; after the update is completed, node N clears offset _pre to zero; the relationship between frequency offset and phase offset cycle is shown in Figure 3;

In step S33, node N checks the timer. If the time period d has not expired since the timing of step S22, then go to step S30 after sleeping for Δt time; otherwise, complete this cycle and go to step S10 to start the next cycle.

Two, the determination of each relevant parameter in the inventive method:

Experiment 1: The influence of the error control factor μ on the packet exchange interval and the number of data packet exchanges and the determination of the optimal value of μ:

Step 1: Initialize the simulation experiment scene

The applicant simulated 12,000 pieces of temperature data based on the temperature change in the real field scene, representing the temperature value of each second within 200 minutes. The temperature and pressure change curve obtained according to the data is shown in Figure 4. To demonstrate the synchronous performance of the method under varying temperature conditions, the trajectory mainly alternates between two temperature regimes: a temperature flat and a temperature ramp regime. The temperature range is 25°C to 43°C, and according to the relationship between temperature and frequency offset, the corresponding 12,000 pieces of frequency offset data vary from 20ppm to 60ppm.

Step 2: Take the precision control parameter μ=150, 300, 450, 750, 900 (unit: μs). Under each value of the control parameter μ, the 12,000 pieces of frequency offset data are synchronized according to the time synchronization steps described above, and a total of five experiments are performed. Each experiment records the number of time stamp exchanges during the synchronization process, and the interval time between two time stamp exchanges, and observes the influence of μ on these two parameters.

Step 3: Analyze and process experimental data

Figure 5 shows the variation trend of the temperature sensitivity factor TSF estimation interval and the number of time stamp exchanges with the precision control parameter μ. It can be seen that when the parameter μ is changed from 150 μs to 300 μs, the number of timestamp exchanges is significantly reduced, and at the same time, the average TSF estimation interval is significantly increased. However, after the parameter μ changes to 300μs, the changes of these two parameters tend to be gentle. Therefore, in order to ensure the accuracy of time synchronization while reducing energy consumption, this method has the best performance when the accuracy control parameter μ=300 μs.

Experiment 2: The influence of the temperature coordination factor λ on the packet exchange interval and the number of data packet exchanges and the determination of the optimal value of λ:

Step 1: Initialize the simulation experiment scene

The applicant simulated 12,000 pieces of temperature data according to the temperature changes in the real field scene, representing the temperature value of each second within 200 minutes. The temperature and pressure variation curve obtained according to the data is shown in Fig. 4 . To demonstrate the synchronous performance of the method in the case of temperature changes, the trajectory mainly alternates between two temperature regimes: a temperature flat and a temperature ramp regime. The temperature range is 25°C to 43°C, and according to the relationship between temperature and frequency offset, the corresponding 12,000 pieces of frequency offset data vary from 20ppm to 60ppm.

Step 2: Take temperature coordination factors λ=0.6, 0.8, 1.0, 1.2, 1.4 (unit: °C). Under each value of the coordination factor λ, the 12,000 pieces of frequency offset data are synchronized according to the time synchronization steps described above, and a total of five experiments are performed. Each experiment records the number of time stamp exchanges during the synchronization process, and the interval between each time stamp exchange, and observes the influence of λ on these two parameters;

Step 3: Analyze and process experimental data

Figure 6 shows the variation trend of the temperature sensitivity factor TSF estimation interval and the times of timestamp exchange with the precision control parameter μ. It can be seen that when the parameter λ changes from 0.8°C to 1.0°C, the number of time stamp exchanges decreases significantly, and at the same time, the average TSF estimation interval increases significantly; and after the parameter λ changes to 1.0°C, the changes in these two parameters leveled off. Therefore, in order to ensure the accuracy of time synchronization while reducing energy consumption, this method has the best performance when the accuracy control parameter λ=1.0°C.

Three, the method performance test of the present invention and the comparative experiment with other algorithms

Below we verify the performance of the synchronization method of the present invention and its advantages over other methods through a set of experiments. The experiments mainly compare the performance of the following four algorithms:

(1) the inventive method (recorded as TSFB method);

(2) The situation when this method uses a fixed period (denoted as a fixed period TSFB);

(3) FTSP algorithm: This algorithm is a time synchronization algorithm based on timestamp exchange. This method first performs time synchronization between a pair of nodes by exchanging time stamps between nodes, and then performs layer-by-layer synchronization through the method of network layering, and finally achieves time synchronization of the entire network. And this method does not consider the influence of the working environment of the node on its frequency offset.

(4) EACS algorithm: This algorithm also uses temperature for time synchronization, but this method is to obtain the relationship between node frequency offset and temperature before node deployment, and store the relationship in the node in the form of a table. In the process of synchronization, this method assumes that the relationship between frequency offset and temperature will not change with time and temperature, which does not conform to the actual frequency offset change mode, so it will cause synchronization errors.

The experiment mainly proves the advantages of the present invention from the following aspects:

①Frequency offset estimation accuracy; ②Energy consumption (ie, resynchronization interval); ③Algorithm robustness;

Simulation network initialization:

The applicant simulated 12,000 pieces of temperature data according to the temperature changes in the real field scene, representing the temperature value of each second within 200 minutes. The temperature and pressure change curve obtained according to the data is shown in Fig. 4 . To demonstrate the synchronous performance of the method under varying temperature conditions, the trajectory mainly alternates between two temperature regimes: a temperature flat and a temperature ramp regime. The temperature range is 225°C to 43°C, and according to the relationship between temperature and frequency offset, the corresponding 12,000 pieces of frequency offset data vary from 20ppm to 60ppm.

a. Evaluation of Algorithm Frequency Offset Estimation Accuracy

Simulation experiment process:

This experiment mainly compares the frequency offset estimation accuracy of dynamic period TSFB algorithm, fixed period TSFB algorithm and EACS algorithm. In this experiment, the TSFB algorithm with a dynamic period is first simulated. In order to verify the performance of the algorithm under different periods, we take the standard TSF estimated period length d _std as 1500s, 1800s, 2000s, 2300s and 2500s at the beginning of the experiment. After the end, the average packet exchange intervals are: 1050s, 1190s, 1380s, 1450s and 1950s. Then, set the temperature sensitivity factor TSF estimation interval of the EACS algorithm and the TSF estimation interval of the fixed-period TSFB algorithm according to the obtained average packet exchange interval, and repeat the experiment. Finally, the frequency offset estimation results of the three methods are compared with the real frequency offset data generated in the initialization process, and the maximum and average values of the three frequency offset estimation errors are calculated.

Experimental results:

As shown in Figure 7, it is obvious that by comparing the TSFB algorithm with fixed and dynamic intervals with the EACS algorithm, it can be seen that since the TSFB algorithm can dynamically estimate the sensitivity between temperature and frequency offset TSF, it can avoid environmental bands. The frequency offset estimation error caused by the current "temperature-frequency offset" correspondence relationship does not match the prior knowledge, thereby improving the synchronization accuracy. In addition, by comparing the dynamic and fixed TSFB algorithms, it can be seen that the introduction of the dynamic period adjustment mechanism enables the dynamic period TSFB algorithm to correct the TSF value in time when the ambient temperature of the node changes. Compared with the fixed period TSFB algorithm The frequency offset estimation error can be reduced to a certain extent.

b. Algorithm energy consumption evaluation

Simulation experiment process:

This experiment mainly compares the algorithm energy consumption of TSFB algorithm and FTSP algorithm.

First take the precision control parameter μ = 150, 300, 450, 750, 900 (unit: μs). For the TSFB algorithm, under each value of the control parameter μ, the 12,000 pieces of frequency offset data are synchronized according to the time synchronization steps described above, and a total of five experiments are performed. Each experiment records the time between every two timestamp exchanges during the synchronization process. As for the FTSP algorithm, the resynchronization period of the algorithm is fixed (150s).

Secondly, take the temperature coordination factor λ=0.6, 0.8, 1.0, 1.2, 1.4 (unit: ℃). For the TSFB algorithm, under each value of the coordination factor λ, the 12,000 pieces of frequency offset data are synchronized according to the time synchronization steps described above, and a total of five experiments are performed. Each experiment records the time between every two timestamp exchanges during the synchronization process. As for the FTSP algorithm, the resynchronization period of the algorithm is fixed (150s).

Experimental results:

In Fig. 8(a) and Fig. 8(b), it is obvious that for the TSFB algorithm, the interval between two time stamp exchanges increases with the increase of the precision control parameter μ and the degree coordination factor λ. However, even when the values of the error control factor μ and the temperature coordination factor λ are very low (μ=150μs, λ=0.6℃), the nodes can still keep the average interval between two timestamp exchanges greater than 900s, which is far from Much higher than FTSP's 150s. The exchange of time stamps will bring huge communication overhead, and the communication overhead accounts for a large proportion of the total energy consumption of the sensor network. Therefore, it is proved that the TSFB algorithm can reduce energy consumption under the premise of ensuring accuracy. It is very suitable for wireless sensor networks with limited energy.

c. Algorithm robustness evaluation

Simulation experiment process:

The experimental process simulates the situation that the node information transmission cannot be carried out under the harsh environment in the wild. For the TSFB algorithm, the 12,000 pieces of frequency offset data are still synchronized according to the previously described synchronization steps, but starting from the 1000th s, the previously described step 1, that is, the TSF estimation process (since this process depends on timestamp exchange) . The FTSP algorithm also cannot exchange timestamps.

Experimental results:

As shown in Figure 9, at the beginning of the experiment (ie 0-1000s), the gap between the TSFB algorithm and the FTSP algorithm is not very obvious. However, as the ambient temperature begins to change, the frequency offset also changes continuously with the temperature. At this time, the FTSP algorithm can only estimate and compensate the time phase offset by relying on the initial frequency offset estimation value, so the error continues to accumulate. It can also be seen from Figure 9 that the ambient temperature of the node has been rising continuously since 3000s, which makes the frequency offset change rate of the node increase, and the FTSP algorithm cannot capture this change at this time. Therefore, the FTSP's increasing frequency offset estimation error eventually leads to the continuous accumulation of its synchronization errors. Compared with the FTSP algorithm, the TSFB algorithm can still maintain a relatively low synchronization error in the case of communication failure. This is because the communication failure only caused the TSF to be unable to be updated normally, and the node can update the frequency offset according to the latest TSF estimation result and the temperature information collected by the sensor node, which greatly reduces the frequency offset caused by the voltage change. Estimate error. Therefore, compared with the FTSP algorithm, the TSFB algorithm has higher robustness.

Claims (1)

1. A kind of 1. wireless sensor network time synchronization method based on temperature sensing, it is characterised in that：

   Remember that R is the reference mode in wireless sensor network, N is any one sensor node in addition to reference mode, net After network initialization, node N repeats the following cycle, and the cycle includes step 1 to step 3：

   Step 1, Sensitivity Factor determine

   Step S10, node N send time synchronized request data package to reference mode R；

   Step S11, node R return to four reply data bags successively after time of receipt (T of R) synchronization request packet, to node N：M₀, M₁,M₂,M₃, the local zone time of moment node R, respectively time (R) when record sends the packet in each packet₀~ time(R)₃；M₀With M₁、M₂With M₃Interval time is 1s；M₁With M₂Interval time is 10min；

   Step S12, node N are receiving packet M₀~M₃While, record the local zone time time of oneself₀~time₃And The environment temperature temp that node N is presently in₀~temp₃；

   Step S13, node N is to it in time₁And time₃Frequency deviation skew₁And skew₃Calculated：

   Step S14, node N are calculated current Sensitivity Factor TSF values according to frequency deviation and temperature information：

   In formula 2,Temp is normal temperature, value 25 ℃；

   Step 2, Sensitivity Factor interval determine

   Step S20, node N obtain local environment temperature T this moment₁, the node N upper cycles, the moment local environment temperature was T_pre, Then node N rate of temperature change DT is：

   In formula 3, d_preFor the upper cycle step S22 Sensitivity Factor interval d obtained value；

   Step S21, node N calculate current accumulated error value error：

   Step S22, node N are set to Sensitivity Factor interval d, and method is：

   In formula 5, the μ s of μ=150~900, λ=0.6~1.4 DEG C, d_std=20min；

   It is skew that step S23, node N, which set current frequency offset value skew,₃, step S31,100s ＜ Δ t ＜ are transferred to after Δ t durations 10000s；

   Step 3, local zone time renewal

   Step S30, node N obtain its local environment temperature T this moment₂, the Sensitivity Factor TSF calculated according to step S14 is to node Current frequency offset is calculated：

   Skew=TSF (T-Temp)²(formula 6)

   In above formula, T represents the time；

   Step S31, node N calculate current skew：

   In formula 7, skew_preThe current frequency offset value obtained for upper cycle step S23 or step S30, offset_preFor upper one week The current phase bias that phase step S31 is calculated；

   Step S32, if node N current skew meets：

   Then node N is updated to the own local time, and the local zone time clock after renewal is：

   Clock=clock_pre+ offset (formula 9)

   In formula 8 and formula 9, ε is local clock cycles, clock_preFor the local zone time before renewal；After renewal, section Point N is by offset_preReset；

   Step S33, node N check timer, if duration d is not timed out, step S30 is gone to after dormancy Δ t durations；Otherwise, it is complete In the cost cycle, it is transferred to step S10 and starts to perform next cycle.