A gas concentration prediction and real-time positioning method for a gas monitoring device

Abstract

The present application relates to a kind of gas concentration prediction and real-time positioning method for gas monitoring device, comprising the following steps: by hand-held gas monitor obtains the positioning data of a position and the gas concentration monitoring data varying with time, and positioning data and gas concentration monitoring data are transmitted to gas monitoring device main body;Gas monitoring device main body calculates position information according to positioning data, and gas monitoring device main body calculates gas concentration prediction data by prediction network model;Gas monitoring device main body judges whether the position exists risk according to gas concentration prediction data, and sends prediction result to hand-held gas monitor.The method of the present application can quickly obtain the positioning site of hand-held gas monitor, and can efficiently predict the gas concentration of positioning site, with high positioning accuracy, high prediction result accuracy, to facilitate staff to understand the site that may exist risk in time, can make action quickly, avoid safety accidents, protect life and property safety.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of gas monitoring, and particularly relates to a gas concentration prediction and real-time positioning method for a gas monitoring device. BACKGROUND

[0002] According to the safety production requirements of the state, safety monitoring devices must be equipped in some dangerous chemical production workshops, conveying tunnels and other places. The safety monitoring devices include gas monitoring devices for monitoring toxic and harmful gases. In a dangerous chemical environment, real-time monitoring of gas concentration has a positive effect on ensuring a safe production environment. In order to improve the convenience of monitoring, a mobile intelligent gas monitoring device has appeared at present, that is, the gas monitoring device includes a handheld gas monitoring instrument and a gas monitoring device main body. The handheld gas monitoring instrument and the gas monitoring device main body can be separated from each other. The handheld gas monitoring instrument can be provided with a plurality of handheld gas monitoring instruments. Each operator can carry a handheld gas monitoring instrument to monitor different areas.

[0003] However, since the gas concentration data generally has the characteristics of large time span and strong periodicity, the current gas monitoring device cannot efficiently process time series data. Therefore, the existing handheld gas monitoring instrument generally only has the function of detecting gas concentration and cannot efficiently and accurately predict the gas concentration. At present, common time series data processing methods include using autoregressive models, machine learning methods and deep learning, etc. The ARIMA model based on the autoregressive model can only predict trends and has low robustness and is easily affected by the environment, thereby making the prediction accuracy low. Moreover, the model cannot capture nonlinear relationships and cannot meet the complex actual application requirements. The machine learning method includes decision tree regression, support vector regression and ridge regression, etc. The machine learning method can collect actual data to automatically learn and update parameters and is more suitable for use in nonlinear data. However, the machine learning method is difficult to meet the long-term time series dependence of single variable time series information. Deep learning uses a neural network. The causal convolution of the convolutional neural network variant can be used for time series information modeling. The LSTM network proposed later is more suitable for time series data modeling. However, as the number of network layers increases, the model will lose long-term time series information.

[0004] In addition, the handheld gas monitoring instrument cannot be accurately positioned at present. Since the working scene of the handheld gas monitoring instrument is usually an indoor scene, the accuracy of the global satellite navigation system (GNSS) is not enough to meet the requirements of indoor positioning accuracy. Commonly used high-precision indoor positioning methods mainly include WIFI positioning and Bluetooth positioning. Although WIFI positioning has low cost and is easy to deploy, the accuracy is not high. Although the power consumption of Bluetooth positioning is relatively low, the accuracy cannot meet the requirements.

[0005] Therefore, how to realize accurate positioning of a handheld gas monitor and efficient and accurate prediction of the gas concentration at the positioning location is a problem to be solved by those skilled in the art. SUMMARY

[0006] The purpose of the present application is to provide an efficient and accurate gas concentration prediction and real-time positioning method for a gas monitoring device to solve the problems in the prior art.

[0007] To achieve the above purpose, the technical solution adopted by the present application is:

[0008] A gas concentration prediction and real-time positioning method for a gas monitoring device, comprising the following steps:

[0009] S1: obtaining positioning data and gas concentration monitoring data at a certain location by a handheld gas monitor, comprising:

[0010] (1) the handheld gas monitor obtains positioning data at a certain location, and transmits the positioning data to the gas monitoring device main body,

[0011] (2) the handheld gas monitor obtains time-varying gas concentration monitoring data at the location, and transmits the gas concentration monitoring data to the gas monitoring device main body;

[0012] S2: calculating position information and gas concentration prediction data by the gas monitoring device main body, comprising:

[0013] (1) the gas monitoring device main body calculates position information according to the positioning data,

[0014] (2) the gas monitoring device main body calculates gas concentration prediction data through a prediction network model;

[0015] S3: the gas monitoring device main body judges whether there is a risk at the location according to the gas concentration prediction data, and sends the prediction result to the handheld gas monitor.

[0016] Preferably, in S1(1), the handheld gas monitor obtains positioning data by measuring the distance d between the handheld gas monitor and each reference base station multiple times using a TOF ranging method n , wherein:

[0017]

[0018] In the formula:

[0019] d n is the distance between the handheld gas monitor and each reference base station,

[0020] C is the speed of light,

[0021] TTOT is the difference between the time when the handheld gas monitor sends a signal and the time when the feedback signal is received,

[0022] TTAT is the difference between the time when the reference base station receives a signal and the time when the feedback signal is sent.

[0023] Further preferably, in S2(1), the process for the gas monitoring device body to calculate the position information of the handheld gas monitor at the position is: first, pre-process the distances between the handheld gas monitor and each reference base station measured multiple times by using the median mean filtering method, then solve by using the Taylor positioning method, then filter by using the Kalman filtering method, and finally obtain the position information of the handheld gas monitor at the position.

[0024] Still further preferably, the median mean filtering method is: filter multiple d1 by using the median filtering method, sort the multiple d1 according to the numerical value, and eliminate the maximum value and the minimum value, filter the remaining d1 by using the mean filtering method, then filter multiple d2 by using the median filtering method, and so on until multiple d n Filtering is completed.

[0025] Still further preferably, the Taylor positioning method is:

[0026] The coordinates of each reference base station are known as (x n , y n , z n ), and the real position coordinates of the handheld gas monitor are set as (x, y, z), the distances d n between the handheld gas monitor and each reference base station are obtained.

[0027] d n 2 = (x n -x) 2 + (y n -y) 2 + (z n -z) 2 ,

[0028] The initial value of the estimated position coordinates is set as (x0, y0, z0), and Δx, Δy, Δz are the differences between the real position coordinates and the estimated position coordinates, so that:

[0029]

[0030] By Taylor series expansion and ignoring components of the second order and above, it can be obtained that:

[0031]

[0032] In the formula:

[0033]

[0034]

[0035] After the error amount Δx, Δy, Δz is obtained, the coordinate value is updated, and the next iteration calculation is performed,

[0036] until |Δx|+|Δy|+|Δz|<K, wherein K is a set value,

[0037] At this time (x0+Δx, y0+Δy, z0+Δz) is the real position coordinate of the handheld gas monitor.

[0038] Preferably, in S2(2), the process of calculating the gas concentration prediction data of the position in the future period by the prediction network model by the gas monitoring device body includes:

[0039] ①The gas concentration monitoring data obtained by the handheld gas monitor is time series position coded, and the gas concentration monitoring data includes n gas concentration monitoring values,

[0040] ②The time series position coded gas concentration monitoring data is input into the Autoformer prediction network model, and the Autoformer prediction network model calculates m gas concentration prediction values.

[0041] Further preferably, the time series position coding is that the gas concentration monitoring data obtained by the handheld gas monitor includes n continuous time series gas concentration values, t represents the position of the value in the time series, represents the time series position vector corresponding to the gas concentration value at t, d represents the dimension of the position vector, and i represents the i-th dimension of the position vector, is a function of generating the position vector , then:

[0042]

[0043] wherein:

[0044]

[0045]

[0046] Further preferably, the Autoformer prediction network model includes internal sequence decomposition units, autocorrelation mechanisms, and corresponding encoders and decoders, specifically:

[0047] (1) Deep decomposition architecture, in the prediction process, gradually separate the trend item and the periodic item from the hidden variable, realize progressive decomposition,

[0048] (2) Self-correlation mechanism, which contains period-based dependency discovery and time-delay information aggregation.

[0049] Further preferably, in the prediction process, the trend term and the period term are separated from the latent variable step by step, realizing gradual decomposition, specifically:

[0050] ① Sequence decomposition unit, which is based on the moving average idea, smooths the time series, and separates the period term and the trend term:

[0051]

[0052]

[0053] In the formula: is the latent variable to be decomposed, and are the trend term and the period term, respectively, and the above formula is denoted as

[0054] ② Encoder and decoder,

[0055] Encoder: through the sequence decomposition unit, the period term can be separated and Based on this periodicity, a self-correlation mechanism is designed to aggregate similar sub-processes of different periods:

[0056]

[0057]

[0058] Decoder: the trend term and the period term are predicted respectively, for the period term, the self-correlation mechanism is used to mine dependencies based on the periodicity of the sequence, and to aggregate sub-sequences with similar processes, for the trend term, the cumulative method is used to extract trend information from the predicted latent variable step by step:

[0059]

[0060]

[0061]

[0062]

[0063] Preferably, in S3, a threshold N is set in the gas monitoring device body, if the gas concentration prediction data is greater than N, the position is at risk, the gas monitoring device body sends an alarm signal and the position information of the position at risk to the handheld gas monitor, otherwise the positioning and gas concentration prediction of the next position are continued.

[0064] Due to the above technical scheme, the present application has the following advantages compared with the prior art:

[0065] The present application obtains positioning data and gas concentration monitoring data through the handheld gas monitor and calculates the position information and gas concentration prediction data through the gas monitoring device main body, which can quickly obtain the positioning location of the handheld gas monitor, efficiently predict the gas concentration of the positioning location, has high positioning accuracy and high prediction result accuracy, helps the staff to timely understand the possible risk location, quickly make actions, avoid safety accidents, and protect life and property safety. BRIEF DESCRIPTION OF DRAWINGS

[0066] ATTACHED Figure 1 The gas concentration prediction and real-time positioning method of the present embodiment is shown in the overall flowchart.

[0067] ATTACHED Figure 2 The present embodiment gives a gas monitoring device, but is not limited to the gas monitoring device, which mainly includes a handheld gas monitor 1 and a gas monitoring device main body 2.

[0068] In the above drawings: 1, handheld gas monitor; 2, gas monitoring device main body; 3, reference base station. DETAILED DESCRIPTION

[0069] The technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0070] A gas concentration prediction and real-time positioning method, which is mainly applied to a gas monitoring device, such as Figure 2 The present embodiment gives a gas monitoring device, but is not limited to the gas monitoring device, which mainly includes a handheld gas monitor 1 and a gas monitoring device main body 2. The handheld gas monitor 1 and the gas monitoring device main body 2 can be separated from each other and can communicate with each other. The handheld gas monitor 1 can be provided in multiple numbers, and multiple handheld gas monitors 1 can be used simultaneously. A reference base station 3 is provided in the working area of the handheld gas monitor 1. The reference base station 3 is provided in at least three numbers. The handheld gas monitor 1 and the gas monitoring device main body 2 can communicate with the reference base station 3.

[0071] The steps of gas concentration prediction and real-time positioning will be described in detail below, as shown in Figure 1 The steps include:

[0072] S1:

[0073] The positioning data and gas concentration monitoring data of a certain position are acquired by the handheld gas monitor 1, including:

[0074] (1) The handheld gas monitor 1 acquires the positioning data thereof at a certain position, specifically, the distance d between the handheld gas monitor 1 and each reference base station 3 is measured multiple times by using the TOF ranging method n , wherein:

[0075]

[0076] In the formula:

[0077] d n is the distance between the handheld gas monitor 1 and each reference base station 3,

[0078] C is the speed of light,

[0079] TTOT is the difference between the time when the handheld gas monitor 1 sends a signal and the time when the feedback signal is received,

[0080] TTAT is the difference between the time when the reference base station 3 receives a signal and the time when the feedback signal is sent.

[0081] Subsequently, the handheld gas monitor 1 transmits the positioning data to the gas monitoring device main body 2;

[0082] (2) The handheld gas monitor 1 acquires the time-varying gas concentration monitoring data at the position, and transmits the gas concentration monitoring data to the gas monitoring device main body 2.

[0083] S2:

[0084] The position information and gas concentration prediction data are calculated by the gas monitoring device main body 2, including:

[0085] (1) The gas monitoring device main body 2 calculates the position information according to the positioning data, specifically, first, the distances d between the handheld gas monitor 1 and each reference base station 3 measured multiple times are preprocessed by using the median mean filtering method, then solved by using the Taylor positioning method, and then filtered by using the Kalman filtering method, to minimize the error influence of random interference, and finally the position information of the handheld gas monitor 1 at the position is obtained, wherein:

[0086] ① The median mean filtering method is: multiple d1 are filtered by using the median filtering method, multiple d1 are sorted according to the numerical value, and the maximum value and the minimum value are removed, the remaining d1 are filtered by using the mean filtering method, then multiple d2 are filtered by using the median filtering method, and so on, until multiple d n are filtered.

[0087] The Taylor positioning method is: the coordinates of each reference base station 3 are known as (x n , y n , z n ), the real position coordinates of the handheld gas monitor are (x, y, z), the distance d n between the handheld gas monitor 1 and each reference base station is known, and the following can be obtained:

[0088] d n 2 = (x n -x) 2 +(y n -y) 2 +(z n -z) 2 ,

[0089] The initial value of the estimated position coordinates is (x0, y0, z0), Δx, Δy, and Δz are the differences between the real position coordinates and the estimated position coordinates, and the following is obtained:

[0090]

[0091] The Taylor series expansion is obtained by ignoring components of the second order and above:

[0092]

[0093] In the formula:

[0094]

[0095]

[0096] After the error amount Δx, Δy, and Δz are obtained, the coordinate values are updated, and the next iteration calculation is performed,

[0097] until |Δx|+|Δy|+|Δz|<K, where K is a set value,

[0098] At this time, (x0+Δx, y0+Δy, z0+Δz) is the real position coordinates of the handheld gas monitor 1.

[0099] In the embodiment, the initial value (x0, y0, z0) of the estimated position coordinates is the coordinate of the reference base station 3 closest to the handheld gas monitor 1.

[0100] The Kalman filtering method is: the state variable is n-dimensional, the state at time k is x k , the m-dimensional measurement at time k is z k , the state equation is x k = φ k-1 x k-1 +Γk-1 w k-1 , the measurement equation is z k = H k x k + V k , where φ k-1 is the state transition matrix, Γ k-1 is the state noise coefficient matrix, w k-1 is the state noise, H k is the measurement matrix, V k is the measurement noise, which is generally zero-mean white noise,

[0101] According to the uniform acceleration motion model, we have:

[0102]

[0103]

[0104] where s(k) is the one-dimensional position information of the moving target, is the velocity of the moving target,

[0105] According to the Kalman filtering principle, we can obtain the equation set:

[0106]

[0107] Since the sampling frequency of the system is high, the moving target can be considered to move at a constant speed within one sampling period, and thus the above formula can be rewritten as:

[0108]

[0109] Considering that the state model in the two-dimensional case includes the position and velocity in the horizontal and vertical directions, the state equation can be defined as:

[0110]

[0111] where x(k) and are the position and velocity in the x direction, respectively, y(k) and are the position and velocity in the y direction, respectively,

[0112] The system equation can be expressed as:

[0113]

[0114] The corresponding measurement error equation and state equation are:

[0115]

[0116]

[0117] wherein V k and are the residual vectors of the observation L k and the state prediction .

[0118] (2) The gas monitoring device body 2 calculates the gas concentration prediction data through the prediction network model, and the process specifically includes:

[0119] ① Time sequence position coding is performed on the gas concentration monitoring data acquired by the handheld gas monitor 1, and the gas concentration monitoring data includes n gas concentration monitoring values. The time sequence position coding specifically includes that the gas concentration monitoring data acquired by the handheld gas monitor 1 includes n continuous time sequence gas concentration values, t represents the position of the value in the time sequence, represents the time sequence position vector corresponding to the gas concentration value at t moment, d represents the dimension of the position vector, and i represents the i-th dimension of the position vector, is a function of generating the position vector , then:

[0120]

[0121] wherein:

[0122]

[0123]

[0124] ② The time sequence position coded gas concentration monitoring data is input into the Autoformer prediction network model, and the Autoformer prediction network model calculates m gas concentration prediction values, wherein the Autoformer prediction network model includes internal sequence decomposition units, autocorrelation mechanisms and corresponding encoders and decoders, and specifically includes:

[0125] A. Deep decomposition architecture

[0126] In the prediction process, the trend item and the periodic item are gradually separated from the latent variable to realize progressive decomposition, and specifically includes:

[0127] 1. Sequence decomposition unit: The sequence decomposition unit is based on the idea of moving average, smooths the time sequence, and separates the periodic item and the trend item:

[0128]

[0129]

[0130] wherein: is the latent variable to be decomposed, and Respectively, the trend item and the periodic item, the above formula is recorded as:

[0131]

[0132] 2. Encoder and decoder:

[0133] Encoder: through the sequence decomposition unit, the periodic item can be separated And Based on this periodicity, the autocorrelation mechanism is designed to aggregate similar sub-processes of different periods:

[0134]

[0135]

[0136] Decoder: predict the trend item and the periodic item respectively, for the periodic item, use the autocorrelation mechanism to mine the dependence based on the periodicity of the sequence, and aggregate the sub-sequences with similar processes, for the trend item, use the cumulative method to gradually extract the trend information from the predicted hidden variables:

[0137]

[0138]

[0139]

[0140]

[0141] B, autocorrelation mechanism

[0142] The autocorrelation mechanism includes period-based dependence discovery and delay information aggregation.

[0143] S3: The gas monitoring device body 2 judges whether there is a risk at the position according to the gas concentration prediction data, and sends the prediction result to the handheld gas monitor 1, specifically: the gas monitoring device body 2 is provided with a threshold N, if the gas concentration prediction data is greater than N, there is a risk at the position, the gas monitoring device body 2 sends an alarm signal and the position information of the position with risk to the handheld gas monitor 1, otherwise continue to position and gas concentration prediction of the next position. In this embodiment: the threshold N is 10 -5 .

[0144] The above embodiments are only for illustrating the technical concept and characteristics of the present application, the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made in accordance with the spirit and essence of the present application shall be covered within the protection scope of the present application.

Claims

1. A gas concentration prediction and real-time positioning method for a gas monitoring device, characterized by: Comprise the following steps: S1: by handheld gas monitor obtains the positioning data and gas concentration monitoring data of a position, including: (1) The handheld gas monitor obtains positioning data thereof at a position and transmits the positioning data to the gas monitoring device main body. The handheld gas monitor obtains the positioning data by using a TOF ranging method to measure the distance d between the handheld gas monitor and each reference base station multiple times n wherein: In the formula: d n for the distance between the handheld gas monitor and each reference base station, C is the speed of light, TTOT is the difference between the time when the handheld gas monitor sends a signal and the time when the feedback signal is received, TTAT is the difference between the time when the reference base station receives a signal and the time when the feedback signal is sent; (2) The handheld gas monitor obtains the time-varying gas concentration monitoring data at the position, and transmits the gas concentration monitoring data to the gas monitoring device main body; S2: The gas monitoring device main body calculates the position information and the gas concentration prediction data, including: (1) The gas monitoring device body calculates the position information according to the positioning data, first adopts the median mean filtering method to pretreat the distance between the handheld gas monitor and each reference base station measured multiple times, then adopts the Taylor positioning method to solve, then adopts the Kalman filtering method to filter, and finally obtains the position information of the handheld gas monitor at the position, the median mean filtering method is: adopting the median filtering method to filter multiple d1, sorting multiple d1 according to the numerical value, and eliminating the maximum value and the minimum value, and then adopting the mean filtering method to filter the remaining d1, then adopting the median filtering method to filter multiple d2, and so on, until multiple d n Filtering is completed; The Kalman filtering method is: let the state variable be n-dimensional, the state at time k be x k , the m-dimensional measurement at time k be z k , the state equation be x k = φ k-1 x k-1 + Γ k-1 w k-1 , the measurement equation be z k = H k x k + V k , wherein: φ k-1 is a state transition matrix, Γ k-1 is a state noise coefficient matrix, w k-1 is state noise, H k is a measurement matrix, V k is measurement noise, generally zero-mean white noise, According to the uniform acceleration motion model: where s(k) is one-dimensional position information of the moving target, is the velocity of the moving target, According to the Kalman filter principle, the equation group can be obtained: Due to the high sampling frequency of the system, the moving target can be approximately considered to move at a constant speed within a sampling period, so the above formula can be rewritten as: Considering that the state model in the two-dimensional case includes the horizontal and vertical positions and speeds, the state equation can be defined as: where x(k) and are the position and velocity in the x direction, respectively, and y(k) and are the position and velocity in the y direction, respectively, and the system equation can be expressed as: The corresponding measurement error equation and state equation are respectively: where V k and are the residual vectors of the observation L k and state prediction respectively; (2) The gas monitoring device main body calculates the gas concentration prediction data through the prediction network model, and the process of calculating the gas concentration prediction data of the position in the future period of time through the prediction network model by the gas monitoring device main body includes: ①The gas concentration monitoring data obtained by the handheld gas monitor is time-series position coded, the gas concentration monitoring data includes n gas concentration monitoring values, the time-series position coding is: the gas concentration monitoring data obtained by the handheld gas monitor includes n continuous time-series gas concentration values, t represents the position of the value in the time series, represents the time-series position vector corresponding to the gas concentration value at t moment, d represents the dimension of the position vector, i represents the i dimension of the position vector, f: N→R d is a function of generating the position vector , then: Wherein: ② Input the time-series position encoded gas concentration monitoring data into the Autoformer prediction network model, and the Autoformer prediction network model calculates m gas concentration prediction values, the Autoformer prediction network model includes internal sequence decomposition units, autocorrelation mechanisms, and corresponding encoders and decoders, specifically: (1) Deep decomposition architecture, in the prediction process, gradually separate the trend item and the periodic item from the hidden variable, realize progressive decomposition, (2) Autocorrelation mechanism, the autocorrelation mechanism includes period-based dependency discovery and time delay information aggregation, In the prediction process, gradually separate the trend item and the periodic item from the hidden variable, realize progressive decomposition, specifically: ① Sequence decomposition unit, the sequence decomposition unit is based on the sliding average idea, smooths the time series, separates the periodic item and the trend item: where: is the latent variable to be decomposed, and are the trend and periodic terms, respectively, and the above equation is written as: ② Encoder and decoder, Encoder: By sequence decomposition unit, periodic terms can be separated and Based on this periodicity, a self-correlation mechanism is designed to aggregate similar sub-processes of different periods: Decoder: predict the trend item and the periodic item respectively, for the periodic item, use the autocorrelation mechanism to mine the dependencies based on the periodicity of the sequence, and aggregate sub-sequences with similar processes, for the trend item, use the cumulative method to extract trend information from the predicted hidden variable step by step: S3: The gas monitoring device main body judges whether the position is risky according to the gas concentration prediction data, and sends the prediction result to the handheld gas monitor. 2.The gas concentration prediction and real-time positioning method for a gas monitoring device according to claim 1, wherein: The Taylor positioning method is: The coordinates of each reference base station are known as (x n ,y n ,z n ), the real position coordinates of the handheld gas monitor are (x, y, z), and the distance d n between the handheld gas monitor and each reference base station, so that the following can be obtained: d n 2 = (x n - x) 2 + (y n - y) 2 + (z n - z) 2 , Let the initial value of the estimated position coordinates be (x0, y0, z0), and Δx, Δy, Δz be the difference between the true position coordinates and the estimated position coordinates, then: From the Taylor series expansion and ignoring the components of the second order and above, we have: In the formula: After the error amount Δx, Δy, Δz is found out, the coordinate value is updated, and the next iteration calculation is performed, Until |Δx|+|Δy|+|Δz|<K, where K is a set value, At this time (x0+Δx, y0+Δy, z0+Δz) is the real position coordinates of the handheld gas monitor.

3. The gas concentration prediction and real-time positioning method for a gas monitoring device according to claim 1, characterized in that: In S3, a threshold value N is set in the gas monitoring device main body, if the gas concentration prediction data is greater than N, the position is at risk, the gas monitoring device main body sends an alarm signal and the position information of the position at risk to the handheld gas monitor, otherwise the positioning and gas concentration prediction of the next position are continued.

Images