CN103249072B - Wireless sensor network abnormal data analytical approach and wireless sensor node - Google Patents

Abstract

The present invention relates to wireless sensor network technology, disclose a kind of abnormal data analytical approach and a kind of wireless sensor node.Described method comprises: S1, node i get reading x at moment t _tafter, all readings according to current window judge x _twhether be exceptional value, if so, then enter S2, otherwise enter S3; S2, node i obtain the reading of its k neighbor node at moment t, and judge x accordingly _tthe exceptional value origin cause of formation, if hardware or software error, then enter S4, otherwise enter S5; S3, node i store x _t; S4, node i abandon x _t; S5, node i are by x _tsend to base station.Described node comprises data acquisition module, abnormal data analysis module, memory module and wireless communication module.Technical scheme of the present invention can complete calculating in joint neighborhood of a point, reduces network traffics expense; Utilize the space-time relationship of data, automatically detect exceptional value and analyze the exceptional value origin cause of formation, avoid wrong report, fail to report event.

Description

Wireless sensor network abnormal data analytical approach and wireless sensor node

Technical field

The present invention relates to wireless sensor network technology field, particularly a kind of abnormal data analytical approach and a kind of wireless sensor node.

Background technology

By means of the fast development of the communication technology, chip fabrication techniques and embedding assembly technology, wireless sensor network applies to the numerous areas such as national defence, building, environment, city management just more and more.Although wireless sensor node is extensively favored due to features such as cost are low, power consumption is little, however the key factor that govern its development to be the perception data reliability of sensor node low, often containing abnormal data in data.The scene of a typical data exception is: be deployed in the sensor node in forest, and in summer, the perception data of the meridian hour is generally 30 degrees centigrade, and the data in certain moment become 100 degrees Celsius suddenly, and these 100 degrees Celsius is exactly an abnormal data.

The source of abnormal data can be divided into two kinds: one to be hardware or software error, because the cost of sensor is comparatively cheap, the numerical value of hardware may sporadicly be made mistakes, the operating system simultaneously sensor node run has relatively limited data-handling capacity, and lack effective verification scheme, may cause thus makeing mistakes when data processing; Two is there occurs event in monitored area, such as, in the example above, be probably sensor node deployment region near there occurs fire, thus cause sensor reading generation great variety.

Abnormal data analysis occupies considerable status at the Data processing of wireless sensor network, if reason can not analyze the origin cause of formation of abnormal data on the one hand, probably cause wrong report or fail to report the critical events such as fire, reducing the availability of sensor network; Reason is because wireless sensor node typically uses battery supplied energy on the other hand, if can not find fast in a kind of mode of low-power consumption and process abnormal data, can bring unnecessary traffic overhead to network, and then cuts down the life-span of sensor network.

Existing anomaly data detection algorithm, effectively cannot analyze the source of abnormal data, cause lacking the tupe of abnormal data, need to carry out computing frequently at server end, need the data that Internet Transmission is unnecessary in a large number, therefore cannot be applicable to the long-term monitoring of massive wireless sensor.

To sum up, easily there is abnormal feature in, data not high for wireless sensor network data reliability, is necessary the abnormal data analytical approach studying a kind of low-power consumption, low delay, for the Real-Time Monitoring of massive wireless sensor.

Summary of the invention

(1) technical matters to be solved

The object of the present invention is to provide the analytical approach of abnormal data in a kind of wireless sensor network and a kind of wireless sensor node, effectively cannot analyze the exceptional value origin cause of formation to solve existing abnormal deviation data examination method, easily report by mistake or fail to report critical event, and the problem that node power consumption is high, network traffics expense is large.

(2) technical scheme

In order to solve the problems of the technologies described above, the present invention proposes a kind of wireless sensor network abnormal data analytical approach, said method comprising the steps of:

The a certain node i of S1, wireless sensor network gets reading x at moment t _tafter, according to all readings of the current window that node i obtains, judge x _twhether be exceptional value, if x _tbe exceptional value, then enter step S2, otherwise enter step S3;

S2, node i obtain the reading x of its k neighbor node at moment t ₁, x ₂..., x _k, k is positive integer, and according to x ₁, x ₂..., x _kjudge x _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then enter step S4, otherwise enter step S5;

S3, node i are to x _tstore;

S4, node i abandon x _t;

S5, node i are by x _tsend to base station.

Optionally, step S1 specifically comprises:

S101, window size is set as w, front w-1 reading of the current window that node i obtains is designated as x successively _t-w+1, x _t-w+2..., x _t-1, calculate x _t-w+1, x _t-w+2..., x _t-1, x _taverage

S102, according to sequence x _t-w+1, x _t-w+2..., x _t-1, x _tand average structure Cumulate Sum sequence S={s _t-w, s _t-w+1, s _t-w+2..., s _t, wherein,

$\left\{\begin{array}{cc}& s_{t-w}=s_t=0\\ s_j=s_{j-1}+(x_j-\stackrel{\‾}{x}),& j=t-w+1,t-w+2,\·\·\·,t-1\end{array}\right.$

Find out the maximal value s in sequence S _maxwith minimum value s _min, and calculate both difference s _diff=s _max-s _min;

S103, judge s _maxwhether equal s _t, if so, then enter step S104, otherwise enter step S3;

S104, by sequence x _t-w+1, x _t-w+2..., x _t-1, x _tupset at random, obtain new sequence , then repeat step S102, thus obtain new difference

S105, repetition step S104, until obtain N number of new difference , n is positive integer;

S106, to calculate ... in compare s _diffthe number M of little difference, then calculates x _tthe degree of confidence ∈ of exceptional value, wherein,

$\∈=\frac{M}{N}*100\%,$

Judge whether ∈ is greater than default threshold value ∈ _t, if so, then judge x _tbe exceptional value, and enter step S2, otherwise enter step S3.

Optionally, step S2 specifically comprises:

Radius is set as d, percentage threshold is set as q, according to x ₁, x ₂..., x _kand x _tjudge whether following formula is set up:

$\frac{\left|\right\{j\left|\right|x_j-x_t|<d,j\&Element;[1,k\left]\right\}|}{k}\&le;q,$

If formula is set up, then judge x _tthe exceptional value origin cause of formation be hardware or software error, and enter step S4, otherwise enter step S5.

Optionally, in step S2, node i obtains its k neighbor node and specifically comprises at the reading of moment t:

Node i is by wireless signal to its k neighbor node broadcast CROSS_CHECK order, and the reading of each comfortable moment t, after receiving CROSS_CHECK order, is sent to node i by each neighbor node.

Optionally, step S5 comprises further:

Node i notice k neighbor node sends its reading at moment t to base station.

The present invention proposes a kind of wireless sensor node simultaneously, described node comprises data acquisition module, abnormal data analysis module, memory module and wireless communication module, described abnormal data analysis module is connected with described data acquisition module, described memory module and described wireless communication module respectively, wherein:

Described data acquisition module is for obtaining the reading x of node at moment t _t, and by reading x _tsend described abnormal data analysis module to;

Described abnormal data analysis module is for receiving the reading x from described data acquisition module _t, and the abnormal data analytical approach described in utilizing, judge x _twhether be exceptional value, if x _tnot exceptional value, then by x _tsend described memory module to, if x _tbe exceptional value, then judge x further _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then direct by x _tabandon, otherwise by x _tsend described wireless communication module to;

Described memory module is for receiving and storing the reading from described abnormal data analysis module;

Described wireless communication module is for the reading that receives from described abnormal data analysis module and send it to base station.

Optionally, described abnormal data analysis module is based on TinyOS system.

(3) beneficial effect

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

1, described abnormal data analytical approach just can complete calculating in the neighborhood of wireless sensor node, can not bring unnecessary network traffics expense;

2, technical scheme of the present invention make use of the space-time relationship of data, automatically can detect exceptional value and analyze the exceptional value origin cause of formation, and the abnormal data of the elimination non-event origin cause of formation, can not cause wrong report, fail to report event.

Accompanying drawing explanation

Fig. 1 is the basic flow sheet of the wireless sensor network abnormal data analytical approach that the present invention proposes.

Fig. 2 is the principle schematic judging the exceptional value origin cause of formation in a kind of embodiment of the present invention.

Fig. 3 is the fundamental block diagram of the wireless sensor node that the present invention proposes.

Embodiment

Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.

As shown in Figure 1, the wireless sensor network abnormal data analytical approach that the present invention proposes, comprises the following steps:

The a certain node i of S1, wireless sensor network gets reading x at moment t _tafter, according to all readings of the current window that node i obtains, judge x _twhether be exceptional value, if x _tbe exceptional value, then enter step S2, otherwise enter step S3;

S2, node i obtain the reading x of its k neighbor node at moment t ₁, x ₂..., x _k, k is positive integer, and according to x ₁, x ₂..., x _kjudge x _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then enter step S4, otherwise enter step S5;

S3, node i are to x _tstore;

S4, node i abandon x _t;

S5, node i are by x _tsend to base station.

Described method may be summarized to be three phases: first stage is anomaly data detection, and wireless sensor node detects exceptional value, i.e. step S1 according to the perception data of self; Second stage is that abnormal data analyzes agreement, and the node be in same neighborhood carries out collaborative work according to this agreement, analyzes abnormal source, i.e. step S2; Three phases is dealing of abnormal data, and sensor node processes exceptional value according to the feedback of surroundings nodes, i.e. step S3-S5.

Below the embodiment of step S1 and S2 is described in detail.

The perception data of wireless sensor network has temporal relevance usually, if a reading is exceptional value, so certainly exists larger difference between itself and reading before.Step S1 mainly utilizes the temporal associativity of data to detect.

Cycle reads perception data to sensor node at regular intervals, and note node i is x at the reading of time t _t, setting window size is w, and so node needs to judge relative to front w-1 reading x _t-w+1, x _t-w+2, x _t-w+3..., x _t-1, x _twhether be an exceptional value.

According to Cumulate Sum algorithm, first calculate the average of data in current window:

$\stackrel{\&OverBar;}{x}=\frac{{\mathrm{\&Sigma;}}_{j=1}^w{x}_t-w+j}{w},$

Then according to reading sequence x _t-w+1, x _t-w+2..., x _t-1, x _tand average structure Cumulate Sum sequence S={s _t-w, s _t-w+1, s _t-w+2..., s _t, wherein,

$\left\{\begin{array}{cc}& s_{t-w}=s_t=0\\ s_j=s_{j-1}+(x_j-\stackrel{\&OverBar;}{x}),& j=t-w+1,t-w+2,\&CenterDot;\&CenterDot;\&CenterDot;,t-1\end{array}\right.$

It should be noted that due to

$s_t=\underset{j=1}{\overset{w}{\mathrm{\&Sigma;}}}{x}_{t-w+j}-w\stackrel{\&OverBar;}{x}$

So s _t=0.

Then the maximal value s in sequence S is found out _maxwith minimum value s _min, and calculate both difference s _diff=s _max-s _min.

Judge s _maxwhether equal s _tif, s _maxbe not equal to s _t, this detection algorithm directly completes, and judges x _tnot an exceptional value, directly proceed to step S3.

If s _max=s _t, then following calculating is continued: by sequence x _t-w+1, x _t-w+2..., x _t-1, x _tupset at random, obtain new reading sequence , use the same method structure Cumulate Sum sequence S ¹, and obtain new difference

Repeat above upset at random and calculate process N time of maximin difference in Cumulate Sum sequence, N is positive integer, s more initial respectively _diffwith newly-generated N number of difference value , size, if there is M to compare s in N number of value _difflittle, then the reading x of initial current sensor _tthe degree of confidence being exceptional value at current window is

$\&Element;=\frac{M}{N}*100\%,$

If degree of confidence ∈ is greater than default threshold value ∈ _t, then x is judged _tbe an exceptional value, then proceed to step S2, otherwise judge x _tnot an exceptional value, directly proceed to step S3.

The perception data of wireless sensor network has relevance spatially usually.For example, if breaking out of fire makes temperature sensor, reading raises suddenly, and sensor reading so is around inevitable also all can be raised, accordingly if only have the reading generation cataclysm of a sensor, this reading is likely and is caused by software or hardware error, instead of to be caused by event.Step S2 mainly utilizes the spatial correlation of data to analyze.

At moment t, if sensor node i detects reading x _tbe an exceptional value, then this node broadcasts CROSS_CHECK order towards periphery by wireless signal, and node around, after receiving this order, all sends its current reading to node i, is denoted as x ₁, x ₂..., x _k, the wherein number of the neighbor node of k representation node i.

Given radius d and percentage threshold q, judges whether following formula is set up:

$\frac{\left|\right\{j\left|\right|x_j-x_t|<d,j\&Element;[1,k\left]\right\}|}{k}\&le;q,$

If formula is set up, then judge x _tthe exceptional value origin cause of formation be hardware or software error, namely this reading is the exception caused by hardware and software failure, otherwise, judge x _tit is the exceptional value caused by event.Judge the concrete principle of the exceptional value origin cause of formation as shown in Figure 2.

For the application such as fire monitoring, gas leakage monitoring, background server only pays close attention to the event such as fire, gas leakage, if the value of temperature, gas density is in a rational scope, then sensor node is without the need to sending this data to base station, only need to carry out this locality to store, in order to inquiry in the future.

Therefore, if x _tbe not exceptional value, node is stored by the flash memory carried, without the need to sending it back base station by network;

If x _tbe exceptional value, but be judged as and caused by hardware and software failure, then abandon this numerical value, neither need to send to base station, also not need to store;

If x _tbe exceptional value, and be judged as and caused by event, then send this numerical value to base station, and notice neighbor node around sends its current perception data to base station.

The present invention proposes a kind of wireless sensor node simultaneously, as shown in Figure 3, described node comprises data acquisition module, abnormal data analysis module, memory module and wireless communication module, described abnormal data analysis module is connected with described data acquisition module, described memory module and described wireless communication module respectively, wherein:

Described data acquisition module is for obtaining the reading x of node at moment t _t, and by reading x _tsend described abnormal data analysis module to;

Described abnormal data analysis module is for receiving the reading x from described data acquisition module _t, and the abnormal data analytical approach described in utilizing, judge x _twhether be exceptional value, if x _tnot exceptional value, then by x _tsend described memory module to, if x _tbe exceptional value, then judge x further _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then direct by x _tabandon, otherwise by x _tsend described wireless communication module to;

Described memory module is for receiving and storing the reading from described abnormal data analysis module;

Described wireless communication module is for the reading that receives from described abnormal data analysis module and send it to base station.

The technical scheme that the present invention proposes can realize based on TinyOS2.1.1 system.In fact, enforcement of the present invention does not also rely on specific operating system or hardware, programming language etc., can run in any configuration.

First described abnormal data analytical approach programming language is realized, and become independently abnormal data analysis module OutlierAnalysis;

Sensor node is entered in said procedure burning, after burning completes, disposes sensor network;

Sensor node often collects a reading, reading is imported into OutlierAnalysis module and analyzes;

According to analysis result, sensor node is selected to store data, abandon data, or sends data to base station.

In a word, the technical scheme adopting the present invention to propose can improve the quality of perception data, reduces the energy expense of sensor node, prolong network lifetime, and catches deployment region event in real time.

The above is only the preferred embodiment of the present invention; it should be pointed out that for the person of ordinary skill of the art, under the prerequisite not departing from the technology of the present invention principle; can also make some improvement and replacement, these improve and replace and also should be considered as protection scope of the present invention.

Claims (6)

1. a wireless sensor network abnormal data analytical approach, is characterized in that, said method comprising the steps of:

The a certain node i of S1, wireless sensor network gets reading x at moment t _tafter, according to all readings of the current window that node i obtains, judge x _twhether be exceptional value, if x _tbe exceptional value, then enter step S2, otherwise enter step S3;

S2, node i obtain the reading x of its k neighbor node at moment t ₁, x ₂..., x _k, k is positive integer, and according to x ₁, x ₂..., x _kjudge x _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then enter step S4, otherwise enter step S5;

S3, node i are to x _tstore;

S4, node i abandon x _t;

S5, node i are by x _tsend to base station;

Step S1 specifically comprises:

S101, window size is set as w, front w-1 reading of the current window that node i obtains is designated as x successively _t-w+1, x _t-w+2..., x _t-1, calculate x _t-w+1, x _t-w+2..., x _t-1, x _taverage

S102, according to sequence x _t-w+1, x _t-w+2..., x _t-1, x _tand average structure Cumulate Sum sequence S={s _t-w, s _t-w+1, s _t-w+2..., s _t, wherein,

$\left\{\begin{array}{cc}& s_{t-w}=s_t=0\\ s_j=s_{j-1}+(x_j-\stackrel{\&OverBar;}{x}),& j=t-w+1,t-w+2,...,t-1\end{array}\right.$

Find out the maximal value s in sequence S _maxwith minimum value s _min, and calculate both difference s _diff=s _max-s _min;

S103, judge s _maxwhether equal s _t, if so, then enter step S104, otherwise enter step S3;

S104, by sequence x _t-w+1, x _t-w+2..., x _t-1, x _tupset at random, obtain new sequence then repeat step S102, thus obtain new difference

S105, repetition step S104, until obtain N number of new difference n is positive integer;

S106, to calculate in compare s _diffthe number M of little difference, then calculates x _tthe degree of confidence ∈ of exceptional value, wherein,

$\&Element;=\frac{M}{N}*100\%,$

Judge whether ∈ is greater than default threshold value ∈ _t, if so, then judge x _tbe exceptional value, and enter step S2, otherwise enter step S3.

2. wireless sensor network abnormal data analytical approach according to claim 1, it is characterized in that, step S2 specifically comprises:

Radius is set as d, percentage threshold is set as q, according to x ₁, x ₂..., x _kand x _tjudge whether following formula is set up:

$\frac{\left|\right\{j\left|\right|x_j-x_t|<d,j\&Element;[1,k\left]\right\}|}{k}\&le;q,$

If formula is set up, then judge x _tthe exceptional value origin cause of formation be hardware or software error, and enter step S4, otherwise enter step S5.

3. wireless sensor network abnormal data analytical approach according to claim 1, is characterized in that, in step S2, node i obtains its k neighbor node and specifically comprises at the reading of moment t:

Node i is by wireless signal to its k neighbor node broadcast CROSS_CHECK order, and the reading of each comfortable moment t, after receiving CROSS_CHECK order, is sent to node i by each neighbor node.

4. wireless sensor network abnormal data analytical approach according to claim 1, it is characterized in that, step S5 comprises further:

Node i notice k neighbor node sends its reading at moment t to base station.

5. a wireless sensor node, it is characterized in that, described node comprises data acquisition module, abnormal data analysis module, memory module and wireless communication module, described abnormal data analysis module is connected with described data acquisition module, described memory module and described wireless communication module respectively, wherein:

Described data acquisition module is for obtaining the reading x of node at moment t _t, and by reading x _tsend described abnormal data analysis module to;

Described abnormal data analysis module is for receiving the reading x from described data acquisition module _t, and utilize the method according to any one of claim 1-4, judge x _twhether be exceptional value, if x _tnot exceptional value, then by x _tsend described memory module to, if x _tbe exceptional value, then judge x further _tthe exceptional value origin cause of formation, if x _tthe exceptional value origin cause of formation be hardware or software error, then direct by x _tabandon, otherwise by x _tsend described wireless communication module to;

Described memory module is for receiving and storing the reading from described abnormal data analysis module;

Described wireless communication module is for the reading that receives from described abnormal data analysis module and send it to base station.

6. wireless sensor node according to claim 5, is characterized in that, described abnormal data analysis module is based on TinyOS system.