Model prediction-based intelligent early warning method for CO characteristics in motor home

Abstract

The invention relates to the technical field of in-car environment safety intelligent monitoring and diagnosis, in particular to a motor home interior CO characteristic intelligent early warning method based on model prediction, which comprises the following steps: step 1, collecting data and constructing a comprehensive monitoring database including real-time CO concentration data in a motor home; step 2, locking key parameters influencing the CO concentration in the vehicle based on multiple linear regression, constructing a CO concentration absolute value prediction model based on a GRU neural network algorithm, and further analyzing and calculating the CO concentration change rate; and step 3, formulating a CO concentration absolute value-change rate over-threshold early warning logic rule based on model prediction, and realizing intelligent early warning of CO characteristics in the motor home based on the constructed rule. According to the invention, the CO characteristic in the motor home can be diagnosed and early warned in time, and the CO poisoning of people in the motor home can be effectively prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and diagnosis of in-vehicle environmental safety, and particularly relates to an intelligent early warning method for CO characteristics inside a motorhome based on model prediction. Background Art

[0002] With the increasing demand for outdoor life, motorhome tourism, as a new way, has become very popular. As a mobile living space, the safety of the internal environment of a motorhome is directly related to the life and health of users. Carbon monoxide (CO) is a colorless and odorless toxic gas. In the motorhome environment, factors such as improper use of combustion equipment, poor ventilation, and incomplete fuel combustion can lead to CO accumulation, thus posing a serious threat to human health. Therefore, early warning of CO characteristics in the motorhome is an indispensable means to ensure the safety of the people inside the vehicle. How to timely diagnose the CO characteristics in the motorhome and give an early warning before the environment becomes dangerous is a difficult problem in this technical field.

[0003] Existing research shows that real-time and accurate data collection can detect abnormal situations when the CO concentration in the carriage exceeds the standard, etc., which is a key condition for subsequent timely alarm (such as patents CN201610972775.4, CN202210850025.5). There have been related technical inventions in realizing CO characteristic diagnosis and alarm (such as patents CN201820119788.1, CN202010872601.7), but these traditional CO alarm systems mostly alarm based on a fixed threshold. When the CO concentration in the vehicle reaches the set threshold, an alarm is issued, which has a certain lag, and may cause the people inside the vehicle to be exposed to a high-concentration CO environment and get hurt before the alarm is issued, and it cannot achieve the effect of avoiding danger in advance. To eliminate the lag, an alarm strategy can be designed based on the rate of change of concentration, and an early warning is issued when the rate of change of concentration is too high. This method has been applied in other environmental monitoring fields (such as patents CN202311422143.7, CN202211265837.X), but this method may have situations where although the rate of change of concentration is high, the actual concentration is still within the safe range, or the rate of change of concentration is low, but the actual concentration has already exceeded the standard, and it cannot accurately judge the degree of danger. Summary of the Invention

[0004] To solve the problems of existing technologies such as alarm lag and inability to accurately judge the degree of danger, the present invention provides an intelligent early warning method for CO characteristics inside a motorhome based on model prediction, which can timely diagnose and give an early warning of the CO characteristics in the motorhome, and helps prevent CO poisoning of the people inside the vehicle.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] Provide an intelligent early warning method for CO characteristics inside a motorhome based on model prediction, including the following steps:

[0007] S1. Collect data and construct a comprehensive monitoring database including real-time CO concentration data inside the motorhome;

[0008] S2. Based on multiple linear regression, lock the key parameters affecting the CO concentration inside the vehicle, construct a prediction model for the absolute value of CO concentration based on the GRU neural network algorithm, and further analyze and calculate the CO concentration change rate;

[0009] S3. Formulate an early warning logic rule for the "absolute value - change rate" of CO concentration based on model prediction, and realize the intelligent early warning of CO characteristics inside the motorhome based on the constructed rule.

[0010] Further, the comprehensive monitoring database is specifically composed of three parts: vehicle interior and exterior environment parameters, power parameters, and living system parameters; the vehicle interior and exterior environment parameters include: CO concentration (ppm), interior temperature (°C), interior humidity (%), exterior temperature (°C), exterior humidity (%), exterior wind speed (m / s), exterior wind pressure (Pa), ventilation volume (m 3 / h); the power parameters include: engine speed (r / min), combustion efficiency (%), fuel-air ratio (dimensionless), ignition advance angle (°), engine temperature (°C), engine throttle opening (%); the living system parameters include: real-time power of combustion equipment (kW), real-time operating power of air conditioner (kW). The vehicle interior and exterior environment parameters, power parameters, and living system parameters all come from the conventional vehicle-mounted information system, and the time intervals for sampling all the above parameters are kept consistent.

[0011] Further, all parameters use the minimum sampling period ΔT of CO concentration as the time interval, and the maximum limit of the minimum sampling period ΔT of CO concentration depends on the vehicle compartment volume, and the specific value requirements are as follows:

[0012]

[0013] In the formula: ΔT max is the maximum limit of the minimum sampling period ΔT (s), and V is the vehicle compartment volume (m 3 ).

[0014] Further, based on multiple linear regression, lock the key parameters affecting the CO concentration inside the vehicle. After preprocessing the parameters of the comprehensive monitoring database, use the multiple linear regression equation to lock the key parameters affecting the CO concentration in the comprehensive monitoring database:

[0015] Y C =β0 + β1x1 + β2x2 + β3x3 + …… + β j x j +ε;

[0016] Where: Y C is the dependent variable: CO concentration, x j is the independent variable: the key parameter affecting the CO concentration in the vehicle, β j is the regression coefficient of the independent variable, β0 and ε are the intercept and the random error term respectively. When the coefficient of determination of the current j-th parameter is reached, it can be considered that the locked key parameters are sufficient to explain the CO concentration in the carriage.

[0017] Furthermore, a prediction model for the absolute value of CO concentration based on the GRU neural network algorithm is constructed, and the rate of change of CO concentration is further analyzed and calculated, including the following steps:

[0018] S1. The input layer of the prediction model for the absolute value of CO concentration constructed based on the GRU neural network algorithm is the key parameters that lock the influence of the CO concentration in the vehicle, and the output layer is the predicted absolute value of the CO concentration in the carriage after a sampling period.

[0019] S2. Based on the predicted absolute value of CO concentration obtained from the GRU model, the rate of change of CO concentration is further calculated by the following rules:

[0020]

[0021] Where: T i and T i+1 are the end times of the i-th and i+1-th sampling periods respectively, t is the time counted from 0 within each T i , is the predicted absolute value of CO concentration (ppm) at time t, is the rate of change of the predicted absolute value of CO concentration (ppm / s) from time T i to time T i+1 , is the measured value of CO concentration at time T i , is the predicted absolute value of CO concentration (ppm) at time T i+1 .

[0022] Furthermore, based on the "absolute value - rate of change" over-threshold warning logic rules of CO concentration predicted by the model, they are determined by the over-threshold warning logic rules of CO concentration absolute value and the over-threshold warning logic rules of CO concentration rate of change respectively. The over-threshold warning logic rules of CO concentration absolute value are as follows:

[0023]

[0024] Where: ⊙ represents normal, represents warning, represents low-risk alarm, Represents a high-risk alarm, C cr Is the absolute threshold value of CO concentration, and the recommended value is 25 ppm.

[0025] The warning logic rule for the CO concentration change rate exceeding the threshold is as follows:

[0026]

[0027] In the formula: Is the CO concentration change rate threshold (ppm / s), Is the predicted absolute value threshold of CO concentration approaching (ppm), and the recommended value is (0.85 - 0.95)·C cr , τ is the time exceeding the threshold (s), and its value is related to the volume inside the RV, and is specifically determined by the following formula:

[0028]

[0029] Furthermore, to achieve intelligent warning of the CO characteristics inside the RV based on the constructed rules, its application path is: when m1∧m2 = ⊙, it is a normal signal; when It is a warning signal; when It is a low-risk alarm signal; when It is a high-risk alarm signal.

[0030] Advantages of the present invention:

[0031] Compared with the prior art, the intelligent warning method for the CO characteristics inside the RV based on model prediction proposed by the present invention can establish a prediction model according to various factors inside and outside the RV, predict the CO concentration and its change rate in advance, accurately judge the degree of danger when the CO concentration is about to change abnormally, and give an early warning, providing more time for the people inside the vehicle to take measures. Compared with the traditional CO alarm system based on fixed threshold alarm, the present invention can timely diagnose and warn the CO characteristics inside the RV, and can effectively prevent people inside the vehicle from CO poisoning.

[0032] The intelligent warning method for the CO characteristics inside the RV based on model prediction proposed by the present invention is applicable to towed RVs and self-propelled RVs, and is applicable to driving and parking rest scenarios, etc. It has a wide range of applications, can give an early warning in a timely manner, effectively guarantee the life safety of the people inside the vehicle, and improve the safety of RV use. Brief Description of the Drawings

[0033] Figure 1 Is the flow block diagram of the diagnosis and warning method of the present invention.

[0034] Figure 2 Is the comprehensive monitoring database at any moment in the implementation case of the present invention.

[0035] Figure 3This is the evaluation result of the prediction model in the embodiment of the present invention.

[0036] Figure 4 This is the real-time CO characteristics and warning signal characteristics within a certain 20 minutes in the embodiment of the present invention. Detailed implementation manners

[0037] To clearly illustrate the technical features of this solution, the following is an elaboration of this solution through specific implementation manners.

[0038] This embodiment relies on a certain towed caravan. The size of the carriage is 4090mm×2100×2060mm, and its structural schematic diagram is as shown in the appendix Figure 2 as follows. The specific implementation process is as follows:

[0039] An intelligent warning method for the internal CO characteristics of a caravan based on model prediction includes the following steps:

[0040] Step 1: Collect data and construct a comprehensive monitoring database including real-time CO concentration data inside the caravan;

[0041] The caravan in this embodiment is equipped with a conventional vehicle-mounted information system for real-time storage of caravan information. The data collected for constructing a comprehensive monitoring database including real-time CO concentration data inside the caravan specifically consists of three parts: vehicle interior and exterior environment parameters, power parameters, and living system parameters; Vehicle interior and exterior environment parameters include: CO concentration (ppm), interior temperature of the carriage (°C), interior humidity of the carriage (%), exterior temperature of the carriage (°C), exterior humidity of the carriage (%), exterior wind speed (m / s), exterior wind pressure of the carriage (Pa), ventilation volume (m 3 / h); Power parameters include: engine speed (r / min), combustion efficiency (%), fuel-air ratio (dimensionless), ignition advance angle (°), engine temperature (°C), engine throttle opening (%) ; Living system parameters include: real-time power of the combustion equipment (kW), real-time operating power of the air conditioner (kW). The vehicle interior and exterior environment parameters, power parameters, and living system parameters all come from the conventional vehicle-mounted information system. The time intervals of sampling for all the above parameters are kept consistent, and all parameters use the minimum sampling period ΔT of the CO concentration as the time interval. The maximum limit of the minimum sampling period ΔT of the CO concentration depends on the volume of the carriage. The specific value requirements are as follows:

[0042]

[0043] In this embodiment, the size of the carriage is 4090mm×2100×2060mm, and the volume of the carriage V = 17.7m 3 , so the sampling periods of the vehicle interior and exterior environment parameters, power parameters, and living system parameters are all ΔT = 60s. The comprehensive monitoring database at any certain moment is as shown in Figure 3 the figure.

[0044] Step 2: Based on multiple linear regression, lock the key parameters affecting the CO concentration inside the vehicle, construct an absolute value prediction model of CO concentration based on the GRU neural network algorithm, and further analyze and calculate the CO concentration change rate.

[0045] First, collect the parameter data within 7 consecutive days in the comprehensive monitoring database in the conventional in-vehicle information system, with a time interval of 60 s. Use the Matlab software to supplement the missing values and outliers in the data with the average value. After standardizing using the z-score method, use the multiple linear regression equation to lock the key parameters affecting the CO concentration inside the vehicle in the comprehensive monitoring database.

[0046] Y C = β0 + β1x1 + β2x2 + β3x3 + …… + β j x j + ε;

[0047] In the formula: Y C is the dependent variable: CO concentration, x j is the independent variable: the key parameters affecting the CO concentration inside the vehicle, β j is the regression coefficient of the independent variable, and β0 and ε are the intercept and random error amount respectively. When the determination coefficient of the current j-th parameter is satisfied, it can be considered that the locked key parameters are sufficient to explain the CO concentration inside the carriage. The multiple linear regression equation fitted in this implementation case is as follows

[0048] Y C = 1.00266 + 2.00026x1 + 1.50021x2 - 1.00059x3 + 0.99682x4 - 0.801213x5 + 0.49913x6 + 0.29513x7;

[0049] Assuming that there is no systematic deviation in the error, set ε = 0. The key parameters represented by x1, x2, x3, x4, x5, x6, and x7 are ventilation volume, engine speed, combustion efficiency, fuel-air ratio, real-time power of the combustion device, real-time operating power of the air conditioner, and wind speed outside the carriage respectively. Calculate the determination coefficient to get:

[0050] Secondly, use the locked key parameters to construct an absolute value prediction model of CO concentration based on the GRU neural network algorithm. The input layer of the absolute value prediction model of CO concentration constructed based on the GRU neural network algorithm is the locked key parameters affecting the CO concentration inside the vehicle, and the output layer is the predicted absolute value of the CO concentration inside the carriage after a sampling period. In this implementation case, the data is divided into a training set, a validation set, and a test set according to 7:1:2. Select the Adam optimizer to search for the optimal model parameters. The evaluation results of the prediction model are as Figure 4as shown

[0051] Finally, based on the absolute value of the predicted CO concentration obtained from the GRU model, the CO concentration change rate is further calculated by the following rules:

[0052]

[0053]

[0054] where: T i and T i+1 are the end times of the i-th and (i + 1)-th sampling periods respectively, t is the time counted from 0 within each T i During is the absolute value of the predicted CO concentration at time t (ppm), is from T i time to T i+1 the change rate of the absolute value of the predicted CO concentration at time (ppm / s), is the measured value of the CO concentration at time T i , is the absolute value of the predicted CO concentration at time T i+1 (ppm). In this implementation case, the real-time CO concentration, the absolute value of the prediction, and their change characteristics within any 5 minutes are as Figure 4 shown. Taking the time period from T3 to T4 as an example, then at this time

[0055] Step 3: Formulate the "absolute value - change rate" over-threshold warning logic rule for the CO concentration based on model prediction, and implement the intelligent warning of the CO characteristics inside the RV based on the constructed rule.

[0056] Select the data from time T3 to T4 within the above time period of this implementation case to illustrate the "absolute value - change rate" over-threshold warning logic rule for the CO concentration based on model prediction. The described logic rule is determined by the over-threshold warning logic rule of the CO concentration absolute value and the over-threshold warning logic rule of the CO concentration change rate respectively. The over-threshold warning logic rule of the CO concentration absolute value is as follows:

[0057]

[0058] where: ⊙ represents normal, represents warning, represents low-risk alarm, represents high-risk alarm, C cr is the absolute value threshold of the CO concentration, and the recommended value is 25 ppm. Within the time period from T3 to T4 and The values are all less than 25 ppm, so m1 = ⊙.

[0059] The early warning logic rule for the CO concentration change rate exceeding the threshold is as follows:

[0060]

[0061] In the formula: is the CO concentration change rate threshold (ppm / s), is the absolute value of the predicted CO concentration approaching the threshold (ppm), and the recommended value is (0.85 - 0.95)·C cr , τ is the time exceeding the threshold (s), and its value is related to the volume inside the RV and is specifically determined by the following formula:

[0062]

[0063] In this embodiment, τ = 10 s, then At this time K 50 = 0.023 ppm / s, then m2 = ⊙.

[0064] The intelligent early warning of the CO characteristics inside the RV based on the constructed rules is realized, and its application path is: when m1 ∧ m2 = ⊙, it is a normal signal; when it is an early warning signal; when it is a low-risk alarm signal; when it is a high-risk alarm signal. During the selected time from T3 to T4, m1 ∧ m2 = ⊙, and a normal signal is output.

[0065] Of course, the above description is not limited to the above examples. The technical features not described in the present invention can be implemented by existing technologies and will not be elaborated here; the above embodiments and drawings are only used to illustrate the technical solutions of the present invention and are not a limitation to the present invention. The present invention has been described in detail with reference to the preferred embodiments. Those of ordinary skill in the art should understand that any changes, modifications, additions, or substitutions made by those of ordinary skill in the art within the scope of the essence of the present invention do not depart from the purpose of the present invention and should also fall within the scope of the claims of the present invention.

Claims

1. An intelligent early warning method for CO characteristics inside a motorhome based on model prediction, characterized in that: It includes the following steps: Step 1: Collect data and construct a comprehensive monitoring database including real-time CO concentration data inside the RV; Step 2: Based on multiple linear regression, lock the key parameters affecting the CO concentration inside the vehicle, construct a prediction model for the absolute value of CO concentration based on the GRU neural network algorithm, and further analyze and calculate the CO concentration change rate; Step 3: Formulate an early warning logic rule for the "absolute value - change rate" of CO concentration exceeding the threshold based on model prediction, and realize intelligent early warning of the CO characteristics inside the RV based on the constructed rule.

2. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 1, characterized in that: In Step 1, the comprehensive monitoring database includes vehicle interior and exterior environment parameters, power parameters, and living system parameters; the vehicle interior and exterior environment parameters include: CO concentration, interior temperature, interior humidity, exterior temperature, exterior humidity, exterior wind speed, exterior wind pressure, and ventilation volume; the power parameters include: engine speed, combustion efficiency, fuel-air ratio, ignition advance angle, engine temperature, engine throttle opening; the living system parameters include: real-time power of the combustion equipment and real-time operating power of the air conditioner; the time intervals for sampling the above parameters are kept consistent.

3. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 2, characterized in that: All parameters are at time intervals of the minimum sampling period ΔT of CO concentration, and the maximum limit of the minimum sampling period ΔT of CO concentration depends on the vehicle compartment volume, and the specific value requirements are as follows: where: ΔT max is the maximum limit value of the minimum sampling period ΔT, and V is the volume of the carriage.

4. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 1, wherein: In Step 2, based on multiple linear regression, lock the key parameters affecting the CO concentration inside the vehicle. After preprocessing the comprehensive monitoring database parameters, use the multiple linear regression equation to lock the key parameters in the comprehensive monitoring database that affect the CO concentration inside the vehicle: Y C = β0 + β1x1 + β2x2 + β3x3 + …… + β j x j + ε; where: Y C is the dependent variable: CO concentration, x j is the independent variable: the key parameters affecting the CO concentration in the vehicle, β j is the regression coefficient of the independent variable, β0 and ε are the intercept and the random error term respectively; the determination coefficient of the current j - th parameter When, it can be considered that the locked key parameters are sufficient to explain the CO concentration in the vehicle compartment.

5. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 1, characterized in that: In Step 2, construct a prediction model for the absolute value of CO concentration based on the GRU neural network algorithm, and further analyze and calculate the CO concentration change rate, including the following steps: S1: The input layer of the prediction model for the absolute value of CO concentration constructed based on the GRU neural network algorithm is the key parameters affecting the CO concentration inside the vehicle, and the output layer is the predicted absolute value of CO concentration inside the vehicle after one sampling period; S2: Based on the predicted absolute value of CO concentration obtained from the GRU model, the CO concentration change rate is further calculated by the following rules: Where: T i and T i+1 are the end times of the i-th and (i + 1)-th sampling periods respectively, t is the time counted from 0 within each T i , is the absolute predicted value of CO concentration at time t, is the change rate of the absolute predicted value of CO concentration from time T i to time T i+1 , is the measured value of CO concentration at time T i , is the absolute predicted value of CO concentration at time T i+1 .

6. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 1, characterized in that: In Step 3, the early warning logic rule for the "absolute value - change rate" of CO concentration exceeding the threshold based on model prediction is determined by the early warning logic rule for the absolute value of CO concentration exceeding the threshold and the early warning logic rule for the CO concentration change rate exceeding the threshold respectively. The early warning logic rule for the absolute value of CO concentration exceeding the threshold is as follows: Where: ⊙ represents normal, represents early warning, represents low-risk alarm, represents high-risk alarm, C cr is the absolute value threshold of CO concentration, with a value of 25 ppm; The early warning logic rule for the CO concentration change rate exceeding the threshold is as follows: In the formula: is the threshold value of the CO concentration change rate, is the threshold value close to the absolute predicted value of the CO concentration, and the recommended value is (0.85 - 0.95)·C cr , τ is the time above the threshold, and its value is related to the volume inside the RV and is specifically determined by the following formula:

7. The intelligent early warning method for the internal CO characteristics of a motorhome based on model prediction according to claim 1, characterized in that: In step 3, intelligent early warning of the CO characteristics inside the RV based on the constructed rules is realized, and its application path is as follows: when m1∧m2 = ⊙, it is a normal signal; when it is an early warning signal; when it is a low-risk alarm signal; when it is a high-risk alarm signal.

Images