A method and system for calculating ship emission factors

Abstract

The present application relates to a kind of ship emission factor calculation method and system, wherein the method comprises: the exhaust of ship emission is measured, obtains ship exhaust emission data measured value;Determine the length of integral interval, the average value of measured value is converted to the length of integral interval, obtain average value sequence;Extract all SO2 peak point and CO2 peak point in average value sequence;Using dynamic time warping algorithm corrects average value sequence, establishes the matching relationship between each SO2 peak point and each CO2 peak point;From matching relationship, determine the optimal matching result as global optimal peak point;Based on global optimal peak point, calculate ship emission factor.The quality of measured data can be calculated by the system to assess the credibility of the calculation result of ship emission factor, so as to optimize the calculation of ship emission factor.Compared with prior art, the present application has the advantages of high reliability, accurate calculation and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship exhaust monitoring, and particularly relates to a ship emission factor calculation method and system. BACKGROUND

[0002] In the past decade, the global shipping industry has developed rapidly (UNC-TAD, 2017), leading to increasingly serious ship emission problems (Chen et al., 2019). Pollutant gases and particulate matters emitted by ships not only pollute the natural environment, but also affect human health (Liu et al., 2016). SO2 causes acid rain to frequently occur and damage the environment (Matthias et al., 2010), and 50% of forests in Germany and Poland and 30% of forests in Switzerland are eroded by acid rain (Mohajjan et al., 2018); PM 25 induces lung cancer and other cardio-pulmonary diseases, resulting in 2.2-3.3 million deaths worldwide each year (Sofiev et al., 2018). In 2015, the total amount of global shipping emissions was 20.1*10 6 tons of NO x , 11.5*10 6 tons of SO x , and 1.54*10 6 tons of PM (Sofiev et al., 2018). In 2017, the total amount of shipping emissions in the European Union was 2.6*10 6 tons of SO2, 7.7*10 6 tons of NO2, and 18.2*10 6 tons of CO (Jonson et al., 2020).

[0003] Accurate ship emission inventory is the data basis for analyzing the regularity of ship emissions and the scientific basis for controlling ship emissions and optimizing the supervision measures of ship emission control zones (Zhang et al., 2017; Zhang et al., 2017). The calculation of ship emission inventory mainly includes “top-down” and “bottom-up” methods. The “top-down” method is based on the fuel consumption of ships and does not consider the specific position of ships, which is suitable for calculating long-time range source inventory on a global scale. For example, Kesgin et al. calculated the CO, CO2, PM 2.5Kesgin et al. (2001); Corbett et al. (1997) developed global emission inventories by studying the fuel consumption of ships. The “bottom-up” approach is more accurate than the “top-down” approach, as it is based on the movement state and attributes of ships. In recent years, due to the rapid development of AIS, it is more convenient to obtain real-time running state of ships, so this method is widely used in the research of emission inventory. For example: Papaefthimiou et al. used a bottom-up approach based on port ship activities to calculate the emissions of NO x , SO2, PM 2.5 (Papaefthimiou et al., 2016) in 18 Greek ports by international cruise ships; Tichavska et al. obtained ship activity data through AIS and studied the relationship between ship emissions and month, ship type in Las Palmas port (Tichavska et al., 2015).

[0004] When establishing an emission inventory, ship speed, ship position, main engine power, auxiliary engine power, emission factor and other parameter information are needed. Among them, the static and dynamic information of the ship can be directly obtained through channels such as AIS data, while the emission factor needs to be measured to determine, so the accuracy of the measurement directly determines the accuracy of the emission inventory (Ekmekiolu et al., 2020; Yang et al., 2021; Toscano et al., 2021). There are two ways to measure emission factors, based on fuel and based on power. The former can be calculated by measuring the concentration of CO2 and other pollutants, and the latter needs to obtain real-time data of main engine power, auxiliary engine power, operating mode, etc. The two measurement methods can be converted to each other under the condition of known ship fuel consumption rate (Zhang et al., 2016). “Sniffing” technology is one of the methods of measuring emission factors based on fuel. It can quickly and accurately measure the ship plume. Balzani et al. used “sniffing” technology and other monitoring technologies to measure the emission factors and fuel sulfur content of ships in Rotterdam port, and compared the results of several methods. It was found that the mobile “sniffing” monitoring technology was the most convenient, and the accuracy was also the highest, with an average random error of 6% in the measurement of SO2 emission factor (Balzani et al., 2014). Beecken et al. used a small airplane equipped with “sniffing” monitoring equipment to monitor the exhaust emissions of 158 ships in the Baltic and North Seas. The average emission factor of SO2 was About 85% of the monitored ships met the sulfur content limit set by the International Maritime Organization (Beecken et al., 2014).

[0005] For example, SO2, the principle of the “sniffing” measurement technique is based on the fact that the carbon content in the ship fuel is stable at 85%-87%, and the concentration ratio of CO2 and SO2 generated by the combustion of fuel is equal to the molar ratio of carbon and sulfur in the fuel and will not change due to exhaust dilution. Therefore, measuring the concentrations of CO2 and SO2 can be used to calculate the emission factor (Huang et al., 2021).

[0006] However, various uncertainties in the calculation process can cause significant interference to the results of the emission factor, which mainly exists in three aspects: problem one, due to the inconsistent response time of different gas sensors, it is difficult to select the gas measurement value at the same time point as the calculation value of the emission factor. Therefore, the general approach is to select a period of gas measurement values for cumulative calculation to reduce the error caused by the inconsistent response time of different gas sensors. Zhou et al. proposed in a previous study that the accumulation process should be considered as integration, and then the integral result is divided by the time interval, so as to convert the gas measurement value into an average value, and on this basis, a method for selecting the global optimal peak point of the average value of the gas is found. The problem of selecting the appropriate integral interval is converted into the problem of selecting the global peak point in the set of average values of the measurement data (Zhou et al., 2020). However, the time interval of the integral interval mainly depends on experience. Problem two, on the basis of the first problem, the selection of the global optimal peak point directly determines the accuracy of the calculation of the emission factor. However, in a period of time, there will be multiple peak points of SO2 and CO2 in the gas measurement values, and how to establish a matching relationship between the peak points to determine the global optimal peak point of SO2 and CO2 still lacks in-depth research; Problem three, various environmental and equipment factors in the process of measuring ship exhaust can interfere with the gas measurement value, but there is currently no objective method to evaluate the quality of the gas measurement value.

[0007] Some studies on calculating emission factors by other methods also mentioned that uncertainty factors would interfere with the calculation results. Betha et al. compared the particulate matter, black carbon and NOx generated by ultra-low sulfur diesel and hydrogen-derived renewable diesel by monitoring the exhaust gas of the research ship. The study mentioned that there was an unavoidable time delay between the monitoring instruments, which needed to be calibrated by minimizing the variability of the scanning electromobility spectrometer and the lag correlation of the single-particle soot photometer to achieve synchronous measurement. In the study of Villa et al. on analyzing the particulate emission characteristics of ships using unmanned aerial vehicles, the peak point sequence after correction was fitted by Gaussian curve, and the maximum peak point in the fitting result was taken as the data basis for calculating the particle number emission factor. This method has good effect on measurement data with clear peak trend, but for measurement data with poor synchronization and complex changes (such as the 100-meter cross-section mentioned in the literature), it cannot give accurate calculation results. In general, there are few literatures on the uncertainty of emission factor calculation, and simple processing methods are often used to select peak points from measurement data for emission factor calculation. Affected by the measurement environment, there are a large number of fluctuations, and measurement data with unclear peak trend are not uncommon, so it is necessary to study the selection of peak points for calculating emission factors in complex data. SUMMARY

[0008] The purpose of the present application is to overcome the defects of the prior art and provide a ship emission factor calculation method and system.

[0009] The purpose of the present application can be achieved by the following technical solutions:

[0010] A ship emission factor calculation method, comprising the following steps:

[0011] Measuring the exhaust gas emitted by the ship to obtain ship exhaust gas emission data measurement values;

[0012] Determining the integral interval length, converting the measurement values into average values of the integral interval length, and obtaining an average value sequence;

[0013] Extracting all SO2 peak points and CO2 peak points in the average value sequence;

[0014] Using dynamic time warping algorithm to correct the average value sequence and establishing a matching relationship between each SO2 peak point and each CO2 peak point;

[0015] Determining the optimal matching result from the matching relationship as the global optimal peak point;

[0016] Calculating the ship emission factor based on the global optimal peak point.

[0017] Further, the determination of the integral interval comprises the following steps:

[0018] S1, if a complete measurement process has n measurement values, use a sliding window algorithm to traverse all measurement values, and the window size is the length of the candidate integral interval, and the window moves one measurement value each time;

[0019] S2, if a measurement value point is greater than the left and right adjacent measurement value points, define the measurement value point as a peak value point; calculate the number of peak value points in each window, and calculate the peak value density of the window, and the peak value density calculation formula of the window is:

[0020]

[0021] Wherein density(j) represents the peak value density of the jth window, count peak(j) The number of peak value points in the jth window is count data(j) The total number of measurement value points in the jth window is count

[0022] S3, calculate the peak value density standard deviation of all windows in the measurement process:

[0023]

[0024] Wherein average density The average of the peak value density of all windows in the measurement process is average

[0025] S4, if there are N complete measurement processes, execute steps S1 to S3 for each measurement process to obtain N peak value density standard deviations, and calculate the average of the N peak value density standard deviations;

[0026] S5, use multiple candidate integral interval lengths to execute steps S1 to S4, compare the peak value density standard deviations of each candidate integral interval length, and select the candidate integral interval length with the largest peak value density standard deviation as the optimal integral interval length.

[0027] Further, based on the optimal integral interval length, the SO2 measurement value sequence and the CO2 measurement value sequence are converted into SO2 average value sequence and CO2 average value sequence respectively.

[0028] Further, the matching relationship between each SO2 peak value point and each CO2 peak value point includes the following steps:

[0029] The SO2 average value sequence and the CO2 average value sequence are normalized to obtain a normalized SO2 sequence and a normalized CO2 sequence;

[0030] The distance matrix is constructed by improving the Manhattan distance;

[0031] Finding a regular path in the distance matrix, starting from point (0, 0) and matching each point in the normalized SO2 sequence and the normalized CO2 sequence one by one;

[0032] Each point on the regular path corresponds to an SO2 peak point and a CO2 peak point, and there is a matching relationship between the two points, thereby obtaining the matching relationship between the SO2 peak points and the CO2 peak points.

[0033] Further, the SO2 average value sequence and the CO2 average value sequence are normalized to 0-1 respectively to obtain a normalized SO2 sequence Q and a normalized CO2 sequence C, and their expressions are respectively:

[0034]

[0035]

[0036] Wherein, SO2data(i) represents the i-th value in the SO2 sequence, SO2data(min) represents the minimum value in the SO2 sequence, SO2data(max) represents the maximum value in the SO2 sequence, CO2data(j) represents the j-th value in the CO2 sequence, CO2data(min) represents the minimum value in the CO2 sequence, and CO2data(max) represents the maximum value in the CO2 sequence.

[0037] Further, the improved Manhattan distance expression is:

[0038]

[0039] Wherein, Q i represents the i-th value in the normalized sequence Q, and C j represents the j-th value in the normalized sequence C.

[0040] Further, if the number of SO2 peak points is k and the number of CO2 peak points is m, then a k*m distance matrix A is constructed by the improved Manhattan distance, and the element (i, j) in A represents the distance d(Q i , C j ); i j

[0041] Finding a regular path in the distance matrix A, starting from point (0, 0) and matching each point in the normalized SO2 sequence and the normalized CO2 sequence one by one, and accumulating the distance, when reaching the terminal point (k, m), the result of the accumulated distance is the similarity of the normalized SO2 sequence and the normalized CO2 sequence; performing recursion can obtain:

[0042] ​​

[0043] wherein ψ(i, j) represents the shortest distance corresponding to the distance matrix A.

[0044] Further, the optimal matching result is determined from the matching relationship as the global optimal peak point, that is, the K-means clustering algorithm is used to cluster the difference of the normalized average values of the matching relationship, including the following steps:

[0045] A1, the number of initial cluster centers is 2, that is, the sample set is divided into two categories, one is normal change, and the other is abnormal change;

[0046] A2, the difference of the normalized average values of the matching relationship is recorded as D, and two data points in D are randomly selected as the initial cluster centers of two clusters;

[0047] A3, the similarity of each sample point to the two cluster centers is calculated, and the sample point is divided into the cluster corresponding to the cluster center with the largest similarity;

[0048] A4, according to the existing samples in the cluster, the cluster center of each cluster is recalculated;

[0049] A5, steps A3 and A4 are iteratively executed until the cluster center no longer changes;

[0050] A6, the K-means clustering is performed on D for multiple times, and if D can be stably divided into two categories, it is indicated that the threshold is found;

[0051] A7, if the normalized difference of SO2 and CO2 in the matching relationship is greater than the threshold, it is indicated that the matching relationship belongs to abnormal change, and the matching relationship is removed;

[0052] A8, the matching result containing the maximum average value of SO2 is searched from the remaining matching relationship, and the matching result containing the maximum average value of CO2 is selected from the matching result containing the maximum average value of SO2 as the global optimal peak point.

[0053] Further, the method further includes a measurement data evaluation step, that is, an evaluation index is proposed, the data quality of the ship exhaust emission measurement data is evaluated, and the reliability of the calculation result of the ship emission factor is evaluated, so as to optimize the calculation of the ship emission factor.

[0054] Further, the data quality of the ship exhaust emission measurement data is evaluated, that is, the uncertainty of the evaluation index is quantitatively evaluated by using a numerical method based on bootstrap sampling, so as to determine the confidence interval, including the following steps:

[0055] The measurement data is sampled multiple times, the mean value of each sampling is calculated, and a set composed of multiple mean values is obtained;

[0056] If the evaluation index value is in the 95% confidence interval corresponding to the evaluation index, it means that the evaluation index determines the measurement data to be high-quality data, and the evaluation result is recorded as 1. If the evaluation index value is outside the 95% confidence interval, it means that the measurement data is determined to be low-quality data by the evaluation index, and the evaluation result is recorded as 0.

[0057] The quality of the measurement data is labeled. If the peak trend is obvious and the SO2 and CO2 average value sequences have high synchronicity, it is considered to be high-quality data, and the quality label is 1. If the change trend of the SO2 and CO2 average value sequences is quite different and changes sharply, it is considered to be low-quality data, and the quality label is 0.

[0058] The evaluation accuracy of each evaluation index is calculated in combination with the evaluation results of the evaluation indexes and the quality labels. The evaluation index set with higher accuracy is selected to jointly evaluate the quality of the measurement data.

[0059] The distance between the evaluation index value of the measurement data and the overall mean of the center position of the confidence interval is calculated, and the ratio of the distance to the one-sided length of the confidence interval is obtained. The closer the ratio is to 0, the better the quality of the measurement data. The closer the ratio is to 1, the worse the quality of the measurement data. The result of the ratio greater than 1 is re-assigned as 1.

[0060] In the joint evaluation, if the calculated ratio of all evaluation indexes is 1, it is judged that the quality of the measurement data is poor. Otherwise, the mean of all ratios less than 1 is calculated. The closer the mean is to 0, the better the quality of the measurement data is considered by the evaluation index set.

[0061] Based on the evaluation results of the measurement data, the calculation of the ship emission factor is optimized.

[0062] A ship emission factor calculation system includes a data measurement module, a calculation module, and an optimization module.

[0063] The data measurement module is used to measure the exhaust gas emitted by the ship to obtain the ship exhaust emission data measurement value.

[0064] The calculation module calculates the ship emission factor based on the ship exhaust emission data measurement value, including the following steps:

[0065] The length of the integral interval is determined, the measurement value is converted into the average value of the integral interval length, and the average value sequence is obtained.

[0066] All SO2 peak points and CO2 peak points in the average value sequence are extracted.

[0067] The dynamic time warping algorithm is used to correct the average value sequence to establish the matching relationship between each SO2 peak point and each CO2 peak point.

[0068] determine an optimal matching result from the matching relationship as a global optimal peak point;

[0069] calculate a ship emission factor based on the global optimal peak point;

[0070] The optimization module is used to propose evaluation indexes, evaluate the data quality of the ship exhaust emission measurement data, evaluate the reliability of the ship emission factor calculation result, and optimize the calculation of the ship emission factor.

[0071] Compared with the prior art, the present application has the following beneficial effects:

[0072] 1. The present application proposes the concept of peak density standard deviation to measure the ability of a integral interval to represent data variation characteristics. By calculating the peak density standard deviation of each candidate integral interval length, the candidate integral interval length with the maximum peak density standard deviation calculation result is selected as the optimal integral interval length. The greater the value of the peak density standard deviation, the more obvious the data variation characteristics after preprocessing with the integral interval, and the clearer the peak trend, so that the global optimal peak point is more accurate. The present application overcomes the situation of relying on experience to determine the integral interval length in the prior art, and makes the calculation result of the ship emission factor more reliable.

[0073] 2. The present application applies the DTW algorithm to the screening of peak points, can obtain the similarity of CO2 and SO2 sequences and the matching relationship between peak points in the SO2 and CO2 average value sequence, and in combination with the optimal peak point selection standard, can select suitable gas measurement values for the calculation of the emission factor, so that the calculation result of the ship emission factor is more accurate.

[0074] 3. The present application defines multiple evaluation indexes to evaluate the ship exhaust emission measurement data, evaluate the reliability of the calculation result of the ship emission factor, and optimize the calculation of the ship emission factor, which makes up for the blank in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0075] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0076] The present application will be described in detail below in combination with the drawings and specific embodiments. The present embodiment is implemented on the basis of the technical solution of the present application, and gives a detailed implementation manner and specific operation process, but the protection scope of the present application is not limited to the following embodiments.

[0077] Air pollution caused by ship emissions has attracted more and more attention, and the establishment of emission inventory is an important means to evaluate the impact of ship emissions on the environment and human beings. Emission factor is an important parameter in the process of compiling ship emission inventory, and there is a certain uncertainty in estimating the emission factor based on the "sniffing" method to measure ship exhaust.

[0078] Taking the calculation of sulfur dioxide emission factor as an example, the present application proposes the concept of "peak density standard deviation" to determine the optimal integral interval length of the measured value in the "sniffing method", and converts the SO2 and CO2 measured value sequence into an average value sequence with the interval length. Then, the improved Manhattan distance is used to represent the position relationship of the peak points in the SO2 and CO2 average value sequence. Then, the dynamic time warping algorithm is used to find the corresponding relationship of the peak points in the SO2 and CO2 average value sequence, and the alternative peak point set is constructed. Finally, after removing the outliers in the set, the maximum value in the SO2 and CO2 average value sequence is selected as the global optimal peak point to calculate the emission factor. At the same time, in order to evaluate the credibility of the calculation result of the emission factor, 16 evaluation indexes reflecting the characteristics of the measured data are selected, 10000 times of self-exhibition sampling is calculated for each evaluation index to obtain the 95% confidence interval, and the evaluation result of the evaluation index for the quality of the measured data is obtained according to the data quality label (1 represents good quality, 0 represents poor quality), and the index with high positive rate for the quality evaluation of the measured data is selected. From 2019 to 2021, 148 groups of "SO2+CO2" measurement data were collected by using the unmanned aerial vehicle "sniffing" monitoring system in the Shanghai Waigaoqiao port area, and the method proposed in the present application was used for calculation and verification. The results show that: for this data set, 12s is the most appropriate integral length, the algorithm can automatically calculate the emission factor, and the global optimal peak point of 129 groups of data is consistent with the manual screening result, with a correct rate of 87.16%. The correct rate of the joint evaluation result of sample entropy (SO2), information entropy (SO2), skewness (CO2) and interquartile range (SO2) is 71%, the distance between the calculated value of the measured data and the overall mean of the center position of the confidence interval is calculated, and the ratio of the distance to the one-sided length of the confidence interval (the ratio is greater than 1, and the ratio is revalued as 1), if the calculation ratio of the four indexes is 1, it indicates that the quality of the measured data is poor, otherwise, the mean value of all ratios less than 1 is calculated, the closer the mean value to 0, the better the quality of the measured data, and the data quality is quantified in this way.

[0079] The present application has carried on the thorough research analysis to the three problems existing in the prior art proposed in the background art. For problem one, the concept of peak density standard deviation is proposed to measure the ability of an integral interval to represent data change characteristics. The larger the value of the peak density standard deviation, the more obvious the change characteristics of the data after preprocessing by the integral interval, and the more clear the peak trend, so that the selection of the global optimal peak point is more accurate.

[0080] For problem two, in order to select the optimal peak point from a large number of peak points, it is necessary to establish a matching relationship between the average value sequences of the two gases and judge the rationality of all matching relationships. The DTW algorithm is applied to the selection of peak points, which can obtain the similarity of CO2 and SO2 sequences and the matching relationship between the peak points in the average value sequences of SO2 and CO2, and combined with the optimal peak point selection standard proposed by Zhou et al. (Zhou et al., 2019), the appropriate gas measurement value can be selected for the calculation of the emission factor.

[0081] Among them, the DTW algorithm is proposed by Itakura (Itakura, 1975), which is used to measure the similarity of two time series. Dmytrów et al. evaluated the similarity of energy commodity prices and daily COVID-19 case time series through the DTW algorithm (Dmytrów et al., 2021). Li et al. improved the DTW algorithm to compare the correspondence between ship trajectory time series, which improved the performance of trajectory modeling (Li et al., 2020).

[0082] For problem three, 16 evaluation indexes that can be used to reflect data quality are proposed, which are divided into two cases for single gas measurement value, one is to use SO2 measurement value for evaluation, and the other is to use CO2 measurement value for evaluation. The 95% confidence interval of the evaluation index is calculated through bootstrap sampling for the obtained measurement data, and the evaluation result of the evaluation index for the measurement data quality is given accordingly, and the correct rate of the index for data quality evaluation can be obtained combined with the quality label of the measurement data. Select the evaluation index with higher correct rate to form a set for joint evaluation, which further improves the correct rate of gas measurement data evaluation. Finally, the ship exhaust data measured at Shanghai Waigaoqiao Port is tested using the method proposed in the present application, which verifies that the method can find the appropriate global optimal peak point to calculate the emission factor with high correct rate, and the accuracy of joint evaluation of measurement data quality.

[0083] The application provides a ship emission factor calculation system, which comprises a data measurement module, a calculation module and an optimization module; the data measurement module is used for measuring tail gas emitted by a ship to obtain a ship tail gas emission data measurement value; the calculation module calculates a ship emission factor based on the ship tail gas emission data measurement value; and the optimization module is used for evaluating data quality of the ship tail gas emission measurement data, evaluating the reliability of a ship emission factor calculation result, and optimizing the calculation of the ship emission factor.

[0084] Based on the system, a ship emission factor calculation method can be realized, as shown in the method, the method comprises the following steps: Figure 1

[0085] Determining an integral interval length, converting the measurement value into an average value of the integral interval length to obtain an average value sequence;

[0086] Extracting all SO2 peak points and CO2 peak points in the average value sequence;

[0087] Using a dynamic time warping algorithm to correct the average value sequence to establish a matching relationship between each SO2 peak point and each CO2 peak point;

[0088] Determining an optimal matching result from the matching relationship as a global optimal peak point;

[0089] Calculating a ship emission factor based on the global optimal peak point;

[0090] An evaluation index is proposed to evaluate the data quality of the ship tail gas emission measurement data, evaluate the reliability of a ship emission factor calculation result, and optimize the calculation of the ship emission factor.

[0091] The basic method principle of the ship emission factor calculation is a sniffing method based on three assumptions: first, the carbon content in fuel oil of different ships is similar, about 87%; second, almost all of the carbon and sulfur elements in the fuel oil of the ship are converted into carbon dioxide and sulfur dioxide after combustion, and the remaining sulfur and carbon oxides only account for a very small part and can be ignored; and third, the ratio of CO2 to SO2 does not change when the tail gas generated after the combustion of the ship fuel oil is diluted in the air (Hu et al., 2018). Accordingly, by measuring the concentrations of SO2 and CO2 within a period of time, the ratio of the two is calculated after the concentrations are added respectively, and then the SO2 emission factor can be calculated, and the calculation formula is:

[0092]

[0093] Wherein m(·) represents mass, M(C) is the relative atomic mass of carbon, M(SO2) is the relative molecular mass of sulfur dioxide, and ∑[·] represents the summation of the concentration of the measured gas within a period of time. ​

[0094] Considering that the sensors cannot achieve complete synchronization, it is possible to adopt the method of selecting a period of time for measurement value accumulation, rather than a single measurement value, that is, by integrating the gas measurement value in a period of time, the calculation result of the emission factor is more stable. In the present application, we adopt the method of converting the gas measurement value into an average value to calculate the SO2 emission factor, as shown in formula 2:

[0095] ∫(SO 2，peak 2)dt[ppb] / t 2，bkg

[0096]

[0097] wherein SO 2，peak 2 is the peak value of SO2 in the data measurement value, SO 2,bkg 2 is the background value of SO2 in the data measurement value, t is the integral interval length, ∫(·)dt is the integral calculation function with the integral interval t, and AVG(·) is the function for calculating the average measurement value in t seconds.

[0098] Through the above conversion, the problem of selecting the integral interval of the "sniffing" method for the gas is converted into the problem of selecting the integral interval and the peak point. In the research of Zhou et al., the integral interval is an empirical value of 10 s, and then by observing and analyzing the change trend of the peak point in the average value sequence, a suitable global optimal peak point is selected for the calculation of the emission factor. Through the above method, a relatively accurate result can be calculated. However, the selected integral interval length is an empirical value, and lacks theoretical demonstration. In addition, when there are many peak points in a period of time, there is uncertainty in how to pair SO2 and CO2.

[0099] In order to eliminate the influence of the inconsistent response time of SO2 and CO2 sensors, it is necessary to find the matching relationship between the peak points in the average value sequences of SO2 and CO2, so as to facilitate the subsequent screening of the SO2 and CO2 global optimal peak points with corresponding relationship. According to the distance relationship between the peak points, the present application uses the dynamic time warping algorithm (DTW) to establish the matching relationship between the peak points of SO2 and CO2. The purpose of the dynamic time warping algorithm is to find the difference between each data point in the target time sequence and the standard time sequence, and to calculate the minimum value of the accumulated difference, and to find the corresponding path, which is used in the present application for matching the peak points of SO2 and CO2. The target time sequence is denoted as X=(x1, x2,...,x n ), and the standard time sequence is denoted as Y=(y1, y2,...,y m ), which respectively correspond to the results of extracting the peak points in the average value sequences of SO2 and CO2. f represents the distance between the point on the target time sequence and the corresponding point on the standard time sequence in the ideal state without deviation:​

[0100] d(i,j) = f(x i , y j ) (3)

[0101] Calculate the distance between data points in X, Y and their corresponding points to get the distance matrix D:

[0102] D ij = d(i,j) (4)

[0103] The shortest distance is to use the cost matrix D cost Iteratively calculate the dynamic programming path distance between the target sequence and the standard sequence, and the shortest distance path is the matching relationship between the data points on the two sequences.

[0104] D cost (i,j) = D(i,j) + min(D cost (i-1), D cost (j-1), D cost (i-1,j-1)) (5)

[0105] The shortest distance matrix corresponding to the path between the peak points of CO2 and SO2 is obtained by the DTW algorithm, that is, the matching relationship between the peak points in the average value sequence of SO2 and CO2 is obtained, and the effect of correcting the deviation sequence is achieved.

[0106] High-accuracy calculation of emission factors requires selecting measurement values of stable air flow within a certain period of time and eliminating the influence of various uncertainty factors as much as possible. In the present application, a method for selecting the optimal integral interval length is first determined, then the DTW algorithm is used to find the matching relationship between the peak points on the average value sequences of CO2 and SO2, and the globally optimal peak points are selected from them. The specific process is divided into the following four steps:

[0107] 1. Integral interval length selection and peak point extraction. The concept of "peak density standard deviation" is proposed to analyze the distribution and trend of peak points in the measurement value sequence. The larger the value of the peak density standard deviation, the more stable the calculation of the emission factor will be. The measurement value is converted into the average value of the integral interval length, and the global peak points in the average value sequence are extracted;

[0108] 2. DTW-based matching. The average value sequence is corrected using the dynamic time warping algorithm based on the improved Manhattan distance to obtain the matching relationship between the peak points of SO2 and CO2;

[0109] 3. Screening matching relationship. Establish matching relationship between peak points in several obtained measurement data, and perform k-means clustering on all matching relationships as samples. If it can be stably divided into two categories, it means that the dividing line between the two categories can distinguish normal change and abnormal change, so as to find the threshold of concentration change. Then, the results in step 2 are screened in combination with the time span threshold to eliminate abnormal matching results. In the remaining matching results, find the matching results containing the maximum average value of SO2, and screen the matching results containing the maximum average value of CO2 from all the found matching results as the global optimal peak point.

[0110] In this embodiment, the optimization module of the system is used to evaluate the reliability of the measurement results. This embodiment proposes 16 evaluation indexes that can represent the data quality. For several obtained measurement data, the 95% confidence interval of each index is calculated by self-display sampling. If the index value is in the confidence interval, it is recorded as 1, otherwise, it is recorded as 0. In combination with the quality label of the measurement data, some indexes with strong representation ability are selected from the 16 indexes to form an index set for joint evaluation. The distance between the index value of the measurement data and the overall mean of the center position of the confidence interval is calculated, and the ratio of the distance to the one-sided length of the confidence interval is obtained (the ratio greater than 1 is revalued as 1). In joint evaluation, if the calculated ratio of all indexes is 1, it is judged that the measurement data quality is poor; otherwise, the mean of all ratios less than 1 is calculated, and the closer the mean is to 0, the better the quality of the measurement data.

[0111] The principle and method of selecting the integral interval length will be described in detail below. The definition of the peak point is explained as necessary. How to use the DTw algorithm to match the peak point is introduced, so as to correct the SO2 and CO2 average value sequence. A distance calculation method based on Manhattan distance is also proposed, which can represent the distance relationship between the peak points in the average value sequence. The calculation method of the threshold for eliminating abnormal matching relationship is described in detail. The specific method of screening some indexes from the 16 evaluation indexes to form an index set for joint evaluation is described.

[0112] When the "sniffer" device monitors the ship exhaust, the response time of the SO2 and CO2 sensor cannot be completely synchronized, resulting in a deviation between the time series of CO2 and SO2 measurement values. Using a suitable integral interval length (set as t) to convert the measurement value per second into the average value in t seconds can reduce the deviation between the time series to a certain extent.

[0113] However, the setting of t value is often taken as an empirical value without theoretical basis, and there is great uncertainty. If the selected t value is too small, there will be too many fine peaks in the overall measurement data set, or a peak is composed of multiple data points, which causes great difficulty in peak selection, resulting in great instability of the calculation result of the emission factor. If the selected t value is too large, the fluctuation of the entire data set is relatively flat, which cannot present a representative peak trend, which will hinder the selection of the peak and have a greater impact on the calculation of the emission factor.

[0114] The present application proposes a method for determining the optimal integral interval from the perspective of data mining, thereby reducing the uncertainty of selecting an empirical value artificially. In order to select a suitable integral interval for preprocessing, a plurality of candidate intervals need to be analyzed, because a peak point needs to be composed of at least three data points, and a too large interval will excessively smooth the data change trend, so only the corresponding relationship between the data change trend and the interval needs to be found, and therefore the selection range of the length of the set integral interval is 3s to 30s. The sliding window algorithm is used to traverse the measurement values, the window size is the length of the candidate integral interval, the window moves one measurement value point each time, the ratio of the number of peak points in each window to the total number of data points in the window is calculated, which is the peak density of the window, and the peak density standard deviation of the length of the candidate integral interval is calculated to reflect the change trend of the peak density in the measurement values. The peak density standard deviations of each candidate integral interval length are compared, and the greater the value is, the more the integral interval length can reflect the fluctuation of the peak density in the measurement values. The specific steps of the method are as follows:

[0115] 1. Assuming that a complete measurement process has n measurement values, the sliding window algorithm is used to traverse all the measurement values. The window size is the length of the selected integral interval, and the window moves one measurement value point each time;

[0116] 2. If a measurement value point is greater than the left and right adjacent measurement value points, the measurement value point is defined as a peak point. The number of peak points in each window is calculated, and the peak density of the window is calculated by formula 6. Wherein density(j) represents the peak density of the jth window, count peak(j) represents the number of peak points in the jth window, count data(j) represents the total number of measurement value points in the jth window;

[0117]

[0118] 3. The peak density standard deviation of all windows in the measurement process is calculated;

[0119]

[0120] 4. Assuming there are N complete measurement processes, steps 1-3 are performed to obtain N peak density standard deviations, and the mean is calculated;

[0121] 5. Steps 1-4 are performed using alternative integral interval lengths of 3s-30s, and the integral interval length with the largest result is the optimal integral interval length.

[0122] Using the optimal integral interval length, the SO2 and CO2 measurement value sequences are converted into SO2 and CO2 average value sequences according to the description in formula 2, and all peak points in the sequences are extracted for subsequent matching.

[0123] In ship plume, the concentration of SO2 is generally 0-10 ppm, and the concentration of CO2 is generally 300-10000 ppm. In order to use the DTW algorithm for matching, the SO2 and CO2 average value sequences need to be normalized to 0-1 to obtain new SO2 and CO2 sequences (denoted as Q and C respectively):

[0124]

[0125]

[0126] Generally, the DTW algorithm is based on the Euclidean distance matrix between two sequences to calculate the similarity. When calculating the Euclidean distance, only the size of one-dimensional value is considered, but when selecting the global optimal peak point, not only the relative size of the concentration needs to be considered, but also the time span (the response time deviation of different sensors is generally a few seconds). The matching between two points with large time span is abnormal matching, so a two-dimensional relative position relationship is used to represent the distance between two points. Manhattan distance takes into account the values on both coordinate axes, and is more suitable for the selection of global optimal peak points than Euclidean distance. Since the average value has been normalized, the time span needs to be converted to a value between 0 and 1, so it is replaced by the ratio of the time span to the total sequence length:

[0127]

[0128] The Manhattan distance calculated by formula 10 is |Q i -C j | is too large, so a compensation coefficient is added to balance the size of the two dimensions:

[0129]

[0130] If the number of SO2 peak points is k and the number of CO2 peak points is m, then a k*m distance matrix A is constructed by the improved Manhattan distance, where the element (i, j) of A represents the Manhattan distance between Q i and Cj Distance d(Q) i C j );d(Q i C j The smaller the value, the better. i and C j The higher the similarity, the better. Find a regular path in matrix A that starts at point (0, 0) and ends at point (k, m) with the minimum superposition distance. The regular path trajectory is mainly constrained by three aspects: monotonicity, continuity, and boundary conditions.

[0131] 1. The monotonicity constraint ensures that regular paths can only extend in a specified direction. For a point (a, b) on the path, there is a next point (a′, b′) on the path, which must satisfy 0 ≤ (a′-a) and 0 ≤ (b′-b).

[0132] 2. The continuity constraint ensures that any point in sequence Q and sequence C can be mapped to a regular path, and can only be aligned adjacently, not across points. For a point (a, b) on the path, there is a next point (a′, b′) on the path, which must satisfy (a′-a)≤1 and (b′-b)≤1.

[0133] 3. Boundary constraints require that the first and last points on the regular path must be (0, 0) and (k, m).

[0134] Starting from point (0, 0), match each point in sequences Q and C one by one, accumulating the distances. When the endpoint (k, m) is reached, the final accumulated distance is the final distance metric between sequences Q and C, which is the similarity between Q and C. Executing the recursion yields:

[0135]

[0136] Each point on the regularized path corresponds to an SO2 peak point and a CO2 peak point, and there is a matching relationship between the two points. However, some of the peak points extracted from the average value sequence are due to minor fluctuations caused by external environmental factors during the measurement process. These fluctuations can affect the direction of the regularized path when participating in the matching process. Therefore, not all matching results are reasonable, and reasonable matching results need to be selected.

[0137] The points passed by the regular path found by the DTW algorithm in the distance matrix do not all have reasonable matching relationships. Zhou et al. proposed several standards to be followed when selecting the global optimal peak point in previous studies: first, eliminate peak points with drastic changes, because these abnormal changes are caused by uncertainties of the sensor, the monitored gas, the monitored gas content in the atmosphere, etc., which will affect the selection of the global optimal peak point; second, eliminate peak points with a time span exceeding 20 s.

[0138] In a completely ideal case, the average value sequences of SO2 and CO2 should be in a completely synchronized state. When the monitoring equipment slowly approaches the ship plume, the average values of SO2 and CO2 will be in an increasing trend, and when it is away from the plume, it will be in a decreasing trend. Since there is no interference from external environmental factors and sensor factors in the ideal state, the difference between the average values of SO2 and CO2 remains stable at any time point and does not change greatly. However, in actual measurement, the complex external environment may cause a large difference in the average value difference at different time points, which cannot be used as a global optimal peak point. Therefore, a threshold is needed to distinguish between normal changes and abnormal changes, and to eliminate matching results with large average value differences. In the present invention, the k-means clustering algorithm is used to cluster the difference (denoted as D) between the normalized average values of all matching results, in order to obtain a threshold that distinguishes between "normal changes" and "abnormal changes". The specific process is as follows:

[0139] 1. The number of initial cluster centers k is 2, i.e. the sample set is divided into two categories, one for "normal changes" and the other for "abnormal changes";

[0140] 2. Randomly select two data points in D as the initial cluster centers of the two clusters;

[0141] 3. Calculate the similarity of each sample point to the two cluster centers, and divide the sample point into the cluster corresponding to the cluster center with the largest similarity;

[0142] 4. Recalculate the cluster center of each cluster according to the existing samples in the cluster;

[0143] 5. Iteratively execute steps 3 and 4 until the cluster center no longer changes.

[0144] Perform multiple k-means clustering on D. If D can be stably divided into two categories, it means that the threshold has been found. If the normalized difference between SO2 and CO2 in the matching result is greater than the threshold, it means that the matching result belongs to abnormal changes and should be eliminated.

[0145] According to the two conditions, reasonable matching results are screened from the matching results, the matching results containing the maximum average value of SO2 are searched in the reserved matching results, and the matching results containing the maximum average value of CO2 are screened from all the found matching results as the global optimal peak point.

[0146] The above method can realize automatic calculation of emission factors. However, there is no suitable evaluation method for the quality of measurement data. Therefore, the present application proposes a method for evaluating the quality of measurement data, which can give a quality evaluation result according to the index calculation result of the measurement data. The present embodiment proposes 16 evaluation indexes, which are sample entropy, information entropy, third and fourth quartiles, standard deviation, skewness, peak density standard deviation, permutation entropy, fuzzy entropy, approximate entropy, mutual information, first quartile, kurtosis, DTW shortest distance, interquartile range, coefficient of variation, and peak point number ratio. Among them, the three evaluation indexes of DTW shortest distance, peak point number ratio and mutual information are obtained by calculating SO2 and CO2 measurement data. The remaining evaluation indexes can be obtained by calculating SO2 or CO2 only. In order to ensure the accuracy of the evaluation result, the evaluation accuracy rate calculated by using SO2 and CO2 measurement data for these indexes needs to be verified respectively, and the one with higher accuracy rate is selected to participate in the calculation of the index. The role of each evaluation index is shown in Table 1.

[0147] Table 1. 16 evaluation indexes and their roles in evaluating data quality

[0148]

[0149] The uncertainty of the evaluation index can be quantitatively evaluated by using a numerical method based on self-unfolding sampling, so as to determine the confidence interval thereof. Therefore, 10000 times of self-unfolding sampling is performed on the existing measurement data, the mean value of each sampling is calculated, and a set composed of a plurality of mean values is obtained. According to the set, the confidence interval of the evaluation index can be calculated. If the index value is in the 95% confidence interval corresponding to the index, it is indicated that the measurement data is determined as high-quality data by the index, and is recorded as 1. If the index value is out of the 95% confidence interval, it is indicated that the measurement data is determined as low-quality data by the index, and is recorded as 0. Meanwhile, the quality of the measurement data is labeled. If the peak trend is obvious, and the average value sequences of SO2 and CO2 have high synchronism, the measurement data is considered as high-quality data, and the quality label is 1. If the change trend of the two sequences is quite different and the frequent change is obvious, the measurement data is considered as low-quality data, and the quality label is 0. The evaluation accuracy of each index is calculated in combination with the evaluation result of the index and the quality label. A certain number of indexes with high accuracy are selected to form an index set for joint evaluation of the data quality. The distance between the index value of the measurement data and the overall mean value of the center position of the confidence interval is calculated, and the ratio of the distance to the length of the single side of the confidence interval is obtained. The closer the ratio is to 0, the better the quality of the measurement data is. Conversely, the closer the ratio is to 1, the worse the quality of the measurement data is. When the numerator in the ratio is greater than the denominator, the ratio is greater than 1. Therefore, the result of the ratio greater than 1 is also re-assigned as 1. In the joint evaluation, if the calculated ratios of all indexes are 1, it is judged that the quality of the measurement data is poor. Otherwise, the mean value of all ratios less than 1 is calculated. The closer the mean value is to 0, the better the quality of the measurement data is considered by the index set. Finally, according to the evaluation result of the measurement data, the measurement data can be selected to be re-collected or part of the measurement data can be removed for calculation of the emission factor, so as to obtain more accurate calculation result.

[0150] The preferred embodiments of the present application are described in detail above. It should be understood that those skilled in the art can make many modifications and changes without creative work according to the concept of the present application. Therefore, any technical solution obtained by logical analysis, reasoning or limited experiment on the basis of the prior art according to the concept of the present application shall be within the protection scope defined by the claims.

Claims

1. A method of calculating a ship emission factor, characterized by, The method comprises the following steps: Measuring the exhaust gas emitted by the ship to obtain ship exhaust emission data measurement values; Determining the integral interval length, converting the measurement values into average values of the integral interval length, and obtaining an average value sequence; Extracting all SO2 peak points and CO2 peak points in the average value sequence; Using a dynamic time warping algorithm to correct the average value sequence and establishing a matching relationship between each SO2 peak point and each CO2 peak point; Determining an optimal matching result from the matching relationship as a globally optimal peak point; Calculating a ship emission factor based on the globally optimal peak point; The determination of the integral interval comprises the following steps: S1. If there are n measurement values in a complete measurement process, a sliding window algorithm is used to traverse all the measurement values, and the window size is a candidate integral interval length, and the window is moved by one measurement value each time; S2. If a measurement value point is greater than the adjacent measurement value points on the left and right, the measurement value point is defined as a peak point. The number of peak points in each window is calculated, and the peak density of the window is calculated. The peak density calculation formula of the window is: wherein denotes the peak density of the jth window, denotes the number of peak points of the jth window, denotes the total number of measured points of the jth window; S3. The peak density standard deviation of all windows in the measurement process is calculated: wherein is the average of the peak density of all windows during the segment of the measurement process; S4. If there are N complete measurement processes, steps S1 to S3 are performed for each measurement process to obtain N peak density standard deviations, and the average of the N peak density standard deviations is calculated; S5. Steps S1 to S4 are performed using a plurality of candidate integral interval lengths, the peak density standard deviations of each candidate integral interval length are compared, and the candidate integral interval length with the largest peak density standard deviation calculation result is selected as the optimal integral interval length.

2. A method of calculating an emission factor for a ship as claimed in claim 1, characterised in that, The matching relationship between each SO2 peak point and each CO2 peak point comprises the following steps: Based on the optimal integral interval length, the SO2 measurement value sequence and the CO2 measurement value sequence are converted into an SO2 average value sequence and a CO2 average value sequence, respectively; The SO2 average value sequence and the CO2 average value sequence are normalized to obtain a normalized SO2 sequence and a normalized CO2 sequence; A distance matrix is constructed by using an improved Manhattan distance; In the distance matrix, a warping path is found, and each point in the normalized SO2 sequence and the normalized CO2 sequence is matched one by one starting from the point (0, 0); Each point on the warping path corresponds to an SO2 peak point and a CO2 peak point, and a matching relationship exists between the two points, thereby obtaining the matching relationship between the SO2 peak points and the CO2 peak points.

3. A method of calculating an emission factor for a ship as claimed in claim 2, characterised in that, The SO2 average value sequence and the CO2 average value sequence are normalized to 0-1 respectively to obtain a normalized SO2 sequence and a normalized CO2 sequence The expressions are respectively:​ wherein represents the value at the i th position in the SO2 sequence, represents the minimum value in the SO2 sequence, represents the maximum value in the SO2 sequence, represents the value at the j th position in the CO2 sequence, represents the minimum value in the CO2 sequence, represents the maximum value in the CO2 sequence.

4. A method of calculating an emission factor for a ship as claimed in claim 3, characterised in that, The expression of the improved Manhattan distance is: wherein denotes the normalized sequence at the i th value, denotes the normalized sequence at the j th value.

5. A method of calculating an emission factor for a ship as claimed in claim 4, characterised in that, If the number of SO2 peak points is The number of CO2 peak points is Then, a modified Manhattan distance is constructed. The distance matrix A, where the element (i,j) in A represents the distance matrix A. and distance ; In the distance matrix A, a warping path is found, and each point in the normalized SO2 sequence and the normalized CO2 sequence is matched one by one starting from the point (0, 0), and the distance is accumulated. When the end point (k, m) is reached, the accumulated distance is the similarity of the normalized SO2 sequence and the normalized CO2 sequence. Recursion is performed to obtain: wherein denotes the shortest distance corresponding to the distance matrix A.

6. A method of calculating an emission factor for a ship as claimed in claim 5, characterised in that, From the matching relationship, an optimal matching result is determined as a globally optimal peak point, that is, a K-means clustering algorithm is used to cluster the difference between the normalized average values of all matching relationships, comprising the following steps: A1, the number of initial cluster centers is 2, that is, the sample set is divided into two categories, one is normal change, and the other is abnormal change; A2, the difference between the normalized average values of the whole matching relationship is recorded as D, and two data points in D are randomly selected as the initial cluster centers of the two clusters; A3, the similarity of each sample point to the two cluster centers is calculated, and the sample point is divided into the cluster corresponding to the cluster center with the largest similarity; A4, according to the existing samples in the cluster, the cluster center of each cluster is recalculated; A5, iteratively execute steps A3 and A4 until the cluster center no longer changes; A6, multiple K-means clustering is performed on D, and if D can be stably divided into two categories, it is considered that the threshold is found; A7, if the normalized difference between SO2 and CO2 in the matching relationship is greater than the threshold, it is considered that the matching relationship belongs to abnormal change, and the matching relationship is removed; A8, find the matching result containing the maximum average value of SO2 from the remaining matching relationship, and select the matching result containing the maximum average value of CO2 from the matching result containing the maximum average value of SO2 as the global optimal peak point.

7. A method of calculating an emission factor for a ship as claimed in claim 1, characterised in that, It also includes a measurement data evaluation step, that is, an evaluation index is proposed to evaluate the data quality of the ship exhaust emission measurement data to evaluate the reliability of the calculation result of the ship emission factor, so as to optimize the calculation of the ship emission factor.

8. A method of calculating an emission factor for a ship as claimed in claim 7, characterised in that, The data quality of the ship exhaust emission measurement data is evaluated, that is, the uncertainty of the evaluation index is quantitatively evaluated by using a numerical method based on self-unfolding sampling, so as to determine its confidence interval, including the following steps: Multiple self-unfolding sampling is performed on the measurement data, the mean value of each sampling is calculated, and a set composed of multiple mean values is obtained; If the evaluation index value is in the 95% confidence interval corresponding to the evaluation index, it is considered that the measurement data is high-quality data, and the evaluation result is recorded as 1; if the evaluation index value is outside the 95% confidence interval, it is considered that the measurement data is low-quality data, and the evaluation result is recorded as 0; The quality of the measurement data is labeled, the peak trend is obvious, and the SO2 and CO2 average value sequence has high synchronicity, which is considered as high-quality data, and the quality label is 1; if the SO2 and CO2 average value sequence has large difference in change trend and changes sharply, it is considered as low-quality data, and the quality label is 0; The evaluation accuracy of each evaluation index is calculated by combining the evaluation results of the evaluation indexes and the quality labels, and the evaluation index set with higher accuracy is selected to jointly evaluate the quality of the measurement data; The distance between the evaluation index value of the measurement data and the overall mean value of the center position of the confidence interval is calculated, and the ratio of the distance to the one-sided length of the confidence interval is obtained, the closer the ratio is to 0, the better the quality of the measurement data, and the closer the ratio is to 1, the worse the quality of the measurement data, and the result with a ratio greater than 1 is re-assigned as 1; In the joint evaluation, if the calculation ratio of all evaluation indexes is 1, it is judged that the quality of the measurement data is poor, otherwise the mean value of all ratios less than 1 is calculated, the closer the mean value is to 0, the better the quality of the measurement data is considered by the evaluation index set. Based on the evaluation result of the measurement data, the calculation of the ship emission factor is optimized.

9. A ship emission factor calculation system characterized by, The method comprises a data measurement module, a calculation module and an optimization module. The data measurement module is used to measure the exhaust gas of the ship emission and obtain the ship exhaust emission data measurement value. The calculation module calculates the ship emission factor based on the ship exhaust emission data measurement value, comprising the following steps: Determine the integral interval length, convert the measurement value into the average value of the integral interval length, and obtain the average value sequence; Extract all SO2 peak points and CO2 peak points in the average value sequence; Use the dynamic time warping algorithm to correct the average value sequence and establish the matching relationship between each SO2 peak point and each CO2 peak point; Determine the optimal matching result from the matching relationship as the global optimal peak point; Calculate the ship emission factor based on the global optimal peak point; The optimization module is used to propose evaluation indexes, evaluate the data quality of the ship exhaust emission measurement data, and evaluate the credibility of the calculation result of the ship emission factor, so as to optimize the calculation of the ship emission factor. The determination of the integral interval comprises the following steps: S1, if a complete measurement process has n measurement values, use the sliding window algorithm to traverse all measurement values, and the window size is the candidate integral interval length, and the window moves one measurement value each time; S2, if a measurement value point is greater than the left and right adjacent measurement value points, the measurement value point is defined as a peak point; calculate the peak value density of each window, and the peak value density calculation formula of the window is: wherein denotes the peak density of the jth window, denotes the number of peak points of the jth window, denotes the total number of measured points of the jth window; S3, calculate the peak value density standard deviation of all windows in the measurement process: wherein is the average of the peak density of all windows during the segment of the measurement process; S4, if there are N complete measurement processes, execute steps S1 to S3 for each measurement process to obtain N peak value density standard deviations, and calculate the average value of the N peak value density standard deviations; S5, use multiple candidate integral interval lengths to execute steps S1 to S4 respectively, compare the peak value density standard deviations of each candidate integral interval length, and select the candidate integral interval length with the largest peak value density standard deviation as the optimal integral interval length.

Images