Green low-carbon port monitoring and evaluation method and system

Abstract

The invention relates to the technical field of smart ports and low-carbon environmental protection, and discloses a green low-carbon port monitoring and evaluation system, which comprises an Internet of Things data acquisition module for acquiring multi-dimensional environmental data, carbon emission data and energy consumption data in real time through multi-type sensor equipment deployed at a port; the data processing and analysis module is used for cleaning and preprocessing the collected data, and mining the relationship between energy use and carbon emission by using a big data analysis technology and an artificial intelligence algorithm; the carbon footprint calculation module is used for calculating the carbon emission of the port activity through an accurate model according to the processed data; by deploying multiple types of sensors, multi-dimensional data in port operation can be acquired in real time, a comprehensive carbon emission evaluation system is formed, the limitation that a traditional method can only monitor a single parameter is overcome through the multi-dimensional data comprehensive acquisition mode, and the problem of information fragmentation is effectively solved; and a solid foundation is provided for subsequent data analysis and decision making.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent ports and low-carbon environmental protection, and specifically provides a method and system for monitoring and evaluating green and low-carbon ports. Background Art

[0002] With the increasing severity of global climate change issues, the international community's call for reducing greenhouse gas emissions and promoting sustainable development is growing louder. As an important node in global logistics transportation, ports play a crucial role in the global supply chain. However, the high energy consumption and high emissions problems of ports have gradually become the focus of attention in various countries. The operation mode of traditional ports often relies on a large amount of fossil fuels, resulting in energy waste and environmental pollution, especially a significant increase in carbon emissions. Therefore, how to achieve green and low-carbon development in port operations has become an important issue in current port management and technological innovation.

[0003] Currently, the environmental protection and low-carbon management of ports mainly rely on single monitoring means, such as independently monitoring items such as air quality, electricity consumption, and fuel consumption. However, these fragmented data are difficult to comprehensively reflect the overall carbon emissions of ports and cannot provide effective decision-making bases for port managers for low-carbon development. The existing technical means generally have the following deficiencies:

[0004] 1. Existing systems often can only monitor single environmental parameters and fail to achieve comprehensive collection and analysis of multi-dimensional data during port operations. For example, air quality monitoring and energy consumption monitoring are often independent of each other, making it difficult to form a comprehensive carbon emissions assessment system.

[0005] 2. Although some ports have introduced basic environmental monitoring systems, these systems usually lack data integration capabilities and cannot uniformly process multi-source data. In addition, traditional monitoring systems lack intelligent analysis functions and are difficult to dynamically evaluate the carbon footprint and energy utilization efficiency of ports.

[0006] 3. Existing systems are mainly used for passive data collection and lack active prediction and early warning functions, and cannot provide effective decision-making support tools for port managers, especially in formulating green development plans and optimizing resource allocation, they appear powerless.

[0007] Therefore, those skilled in the art provide a monitoring and evaluation system for green and low-carbon ports to solve the above-mentioned problems. Summary of the Invention

[0008] Aiming at the deficiencies of the existing technology, the present invention provides a monitoring and evaluation system for green and low-carbon ports to solve the problems raised in the above background art.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A green and low-carbon port monitoring and evaluation system, comprising:

[0010] An Internet of Things data acquisition module, which collects multi-dimensional environmental data, carbon emission data, and energy consumption data in real time through various types of sensor devices deployed at the port;

[0011] A data processing and analysis module, which cleans and preprocesses the collected data, and uses big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions;

[0012] A carbon footprint calculation module, which calculates the carbon emissions of port activities through an accurate model based on the processed data;

[0013] A green port level evaluation module, which conducts a comprehensive evaluation based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction, evaluates the operating performance of the port in terms of green and low-carbon, and helps the port obtain the corresponding green port certification level;

[0014] An intelligent optimization and decision support module, which combines artificial intelligence algorithms to model and predict port operation data, and provides an intelligent decision support function;

[0015] A real-time warning module, which provides warning information by real-time monitoring of abnormal data during port operation, and ensures that the port can take timely measures to cope with environmental risks.

[0016] Preferably, the Internet of Things data acquisition module comprises:

[0017] A sensor unit, comprising: an energy consumption sensor for monitoring the energy consumption of various devices in the port; a carbon emission sensor for real-time monitoring of the emissions of carbon dioxide and other greenhouse gases during port operation; a waste emission sensor for monitoring the waste generated by the port and its emissions; an environmental quality sensor, further comprising: an air quality sensor, a water quality monitoring sensor, and a noise monitoring sensor for monitoring the ambient air, water quality, and environmental noise in the port area;

[0018] A data acquisition and transmission system unit, where the sensors collect data on energy consumption, carbon emissions, and waste emissions in real time, generate an original data stream, and use wireless technology to transmit the data collected by the sensors to the data platform in real time,

[0019] A multi-source data integration unit, which integrates data from different types of sensors, and its fusion steps include:

[0020] Step 1.1 Determine the weight of each sensor data in data fusion by calculating the information entropy of the sensor data, and its formula is:

[0021] For the data sequence x of each sensor i i , the data is discretized into M intervals, and then the probability of the data appearing in each interval is calculated:

[0022] where p ij represents the probability of sensor i in the j-th interval, and N ij represents the count of the data of sensor i in the j-th interval, and N i represents the total number of data of sensor i;

[0023] Step 1.2 Calculate the information entropy according to the probability distribution:

[0024] where H i represents the information entropy of sensor i;

[0025] Step 1.3 Use the information entropy to determine the weight. The lower the weight, the greater the data fluctuation and the higher the uncertainty, and its influence is relatively reduced:

[0026] where w i represents the weight of sensor i, and n represents the total number of sensors;

[0027] Step 1.4 Obtain the fusion result by weighted average:

[0028] where x f represents the fused data value.

[0029] Preferably, the data processing and analysis module includes:

[0030] A data cleaning and preprocessing unit that uses a Hampel filter to remove outliers, local weighted regression to fill in missing data, and exponential normalization to unify the data format, including:

[0031] The algorithm formula of the Hampel filter in Step 2.1 is:

[0032] Within the sliding window centered on the target data point x i , calculate the median of the data within the window:

[0033] median(x i-k ,…,x i+k ),

[0034] Calculate the median absolute deviation (MAD):

[0035] MAD = median(|x j - median(x i-k ,…,x i+k)|),

[0036] Among them, x j represents each data point within the window;

[0037] Determine whether the deviation of each data point from the median exceeds the threshold MAD:

[0038]

[0039] Among them, k represents the sliding window size, λ represents the threshold factor, and x i represents the original data,

[0040] represents the data after filtering;

[0041] In step 2.11, the linear interpolation method is used to fill in the missing values according to the local trend of the data. The interpolation formula is: If known at two points x i-1 and x i+1 , the estimated value of the missing point x i is:

[0042]

[0043] In step 2.12, the exponential normalization method is used to map data with different dimensions to the same standard interval, which can not only compress the influence of extreme values but also retain the subtle differences in the data. The normalization algorithm formula:

[0044]

[0045] Among them, x norm represents the normalized data value, x represents the original sensor data,

[0046] x min and x max represent the minimum and maximum values of the sensor data;

[0047] The data feature extraction and correlation analysis unit extracts key features through ICA and quantitatively evaluates the correlation degree between energy use and carbon emissions using grey correlation analysis. The key features extracted by ICA include:

[0048] In step 2.2, the multi-dimensional data is decomposed into several independent components. It is assumed that the observed data is composed of a mixture of multiple statistically independent source signals.

[0049] First, subtract the mean of each variable: x centered = x - E[x],

[0050] Use eigenvalue decomposition to transform the data into a unit covariance matrix:

[0051]

[0052] Among them, E represents the eigenvector matrix, and D represents the eigenvalue matrix;

[0053] Next, Fast ICA optimizes the demixing matrix through fixed-point iteration, with the goal of maximizing the non-Gaussianity of the signal. The non-linear function g(u) is used:

[0054] w new = E[zg(w T z)] - E[g ′ (w T z)]w,

[0055] Among them, g(u) is the selected non-linear function, and tanh(u) is usually used;

[0056] Finally, the weight vector w obtained through iteration forms the demixing matrix W, and the independent components are finally extracted: s = Wz;

[0057] Step 2.21 adopts the method used in traditional statistical grey relational analysis to quantitatively measure the degree of association between each index:

[0058] For the reference sequence x0(k) and the comparison sequence x i (k), calculate the differences at each moment, and the grey relational coefficient is:

[0059]

[0060] For each comparison sequence, calculate its correlation degree:

[0061] Among them, x0(k) represents the data of the reference sequence at moment k, x i (k) represents the data of the i-th comparison sequence at moment k, n represents the total number of observations, and ρ represents the identification coefficient;

[0062] The data modeling and intelligent prediction unit uses robust regression based on the Huber loss function to construct a prediction model, and at the same time detects abnormal patterns with the improved Local Outlier Factor (LOF) and tunes the model through cross-validation, specifically including:

[0063] Step 2.3 adopts robust regression based on the Huber loss function to reduce the influence of outliers on model fitting, and its algorithm formula is:

[0064]

[0065] Among them, the Huber loss function is defined as:

[0066]

[0067] Among them, y i represents the actual output of the i-th sample, and x ij represents the j-th feature of the i-th sample. β0 and β j represent the model intercept and regression coefficient respectively. δ represents the critical value, which is set according to the data noise level;

[0068] Step 2.31 The Local Outlier Factor (LOF) algorithm is used to identify local abnormal patterns in energy usage and carbon emission data. Its algorithm formula is as follows:

[0069]

[0070] Among them, the local reachability density lrd(p) is defined as:

[0071] Among them, p represents the target data point, and N k (p) represents the k nearest neighbors of the data point p,

[0072] reach-dis(p,q) represents the reachability distance between the point p and its neighbor q, and k represents the number of nearest neighbors;

[0073] The algorithm formula for cross-validation in Step 2.32 is:

[0074]

[0075] Among them, K represents the number of folds of cross-validation, and V k represents the set of sample indices included in the K-th fold validation set. |V k | represents the number of samples contained in the K-th fold validation set (V k ).

[0076] y i represents the true label or true value of the sample i in the validation set, represents the output obtained by predicting the sample i in the validation set using the model trained with all the other folds except the K-th fold validation set. L(·) is a measure of the difference between the true value and the predicted value, and CV K represents the average loss value of the entire K-fold cross-validation process;

[0077] The result visualization and decision support unit visually displays the analysis results, generates a decision support report, and provides a basis for actual management and policy optimization, including:

[0078] Step 2.4 Data visualization: Plot trend charts, scatter plots, heat maps, etc. to display data distribution, variable correlation, and the comparison between the prediction results and the actual values. Use time series charts to analyze the dynamic change laws of energy usage and carbon emissions;

[0079] Step 2.41 Decision Support Report Generation: Comprehensively analyze the output results of the model, generate a comprehensive report including risk warnings, optimization suggestions, etc., transform data insights into management decision-making suggestions, and provide a quantitative basis for energy management and low-carbon strategies.

[0080] Preferably, the carbon footprint calculation module includes:

[0081] A data input and verification unit that extracts and verifies various data related to port activities based on the processed data to ensure data quality;

[0082] An activity volume and energy consumption data integration unit that summarizes, merges, and standardizes multi-source data, quantifies the energy consumption indicators of various activities, and the processing method is as follows:

[0083] Step 3.1 For each port activity category, perform weighted averaging on data from different sources, and the formula is as follows:

[0084]

[0085] Among them, A integrated represents the integrated activity volume or energy consumption indicator, w i represents the weight of the i-th data source, and A i represents the activity volume and energy consumption data of the i-th data source;

[0086] Step 3.11 Use fuzzy logic to perform standardization processing on the possible scale differences and uncertainties between different data sources. The standardization formula is:

[0087]

[0088] Among them, A normal ized represents the standardized activity volume or energy consumption value,

[0089] μ j represents the membership degree of the j-th data source, and A j represents the activity volume or energy consumption data of the j-th data source, and m represents the total number of data sources participating in the standardization;

[0090] Step 3.12 When integrating multiple data sources, combine the weighted average and fuzzy logic methods to obtain the final integrated data, and the calculation formula is:

[0091]

[0092] Among them, A final represents the final integrated data, and w j represents the weight of the j-th data source;

[0093] Step 3.13 In practical applications, it is necessary to adjust the membership function and weighting coefficient of the fuzzy logic to adapt to the characteristics of different data sources and activity types. The adjustment formula is as follows:

[0094]

[0095] α represents the translation parameter of the membership function, and β represents the shape parameter of the membership function.

[0096] Preferably, the carbon emission factor matching and adjustment unit uses a non-linear correction model and a fuzzy comprehensive evaluation method to perform non-linear correction and multi-factor dynamic adjustment on the basic emission factor respectively.

[0097] The algorithm formula of the non-linear correction model is:

[0098]

[0099] A i represents the quantification index of the i-th type of activity.

[0100] EF i represents the basic carbon emission factor of the i-th type of activity, β represents the non-linear correction coefficient, and A ref represents the reference activity amount.

[0101] Regarding the influence of multiple factors on the carbon emission factor in port operations, a fuzzy comprehensive evaluation is used to dynamically adjust the emission factor. The formula is defined as:

[0102]

[0103] Among them, EF j represents the carbon emission factor value under the condition of factor j. The adjusted factor EF fuzzy will be used for subsequent carbon emission calculations.

[0104] The carbon emission calculation unit combines the integrated data with the adjusted emission factor, and uses an exact formula to calculate the carbon emissions of each activity and accumulate them to obtain the total emissions. The calculation steps are as follows:

[0105] Step 3.2 Match the integrated activity data A i with the corresponding basic emission factor EF i and the factor EF fuzzy after fuzzy adjustment.

[0106] Step 3.21 Use the above non-linear correction formula to calculate the carbon emissions of each activity, and then accumulate them to obtain the total emissions.

[0107] Step 3.22 Normalize and convert the unit of the calculation result to ensure that the final output meets international or industry standards.

[0108] The model verification and error analysis unit verifies the model accuracy and optimizes and adjusts the parameters through historical data comparison, sensitivity analysis, and uncertainty assessment. The steps include:

[0109] Step 3.3 performs random sampling on the model input parameters and uses Monte Carlo simulation to calculate the uncertainty of the model output. Assume the model is: y = f(θ1, θ2, …, θ p ),

[0110] where θ1 represents the uncertain parameter,

[0111] After N samplings, the simulated output y k (k = 1, 2, …, N) is obtained, along with the output mean and standard deviation;

[0112]

[0113] where y k represents the model output value obtained from the k-th simulation, represents the mean of the simulated output,

[0114] σ MC represents the uncertainty estimate of the model output

[0115] Step 3.31 compares the model prediction value with the historical monitoring data and uses the root mean square error (RMSE) as the error measurement index. The calculation formula is:

[0116]

[0117] where n represents the number of historical data samples, y sim,i represents the prediction value of the model for the i-th sample, and y hist,i represents the historical observation value corresponding to the i-th sample;

[0118] Step 3.32 combines the model prediction error with the Monte Carlo uncertainty estimate to construct a comprehensive error index E total , and the formula is as follows: E total = A·RMSE + (1 - A)·σ MC ,

[0119] where A represents the weighting coefficient, and E total represents the overall error index comprehensively reflecting the model prediction error and the parameter uncertainty;

[0120] Step 3.4 is to calculate the first-order sensitivity index of each parameter using Sobol sensitivity analysis to clarify the contribution of each uncertain parameter to the model output. The formula is:

[0121]

[0122] Among them, S i represents the first-order sensitivity index of parameter θ i , Var(y) represents the total variance of the model output,

[0123] represents the model output obtained by taking the expectation of the uncertainty of other parameters under the condition of fixing parameter θ i , represents the variance of the mean value of the model output when θ i changes.

[0124] Preferably, the green port level evaluation module includes:

[0125] An index construction and weight assignment unit, which constructs a quantitative model for each evaluation index of the green port and assigns the weights of each index. The method is as follows:

[0126] Step 4.1 realizes the construction of the quantitative model by using data normalization and fuzzification processing;

[0127] Step 4.2 uses mutual information weight calculation, introduces a reference variable Y, where Y is the expert evaluation, historical comprehensive score, and green certification result, and calculates the mutual information between the j-th index and Y:

[0128]

[0129] Normalize the mutual information to obtain the mutual information weight component:

[0130] Among them, MI j represents the mutual information between the j-th index and the reference variable Y, and p(z ij , y) represents the joint probability distribution of the index normalized value and Y, and p(z ij ) and p(y) represent the marginal probability distributions respectively, represents the weight score obtained based on the normalization of the mutual information;

[0131] Step 4.21 obtains the final weight by using a weighted combination method. Let the adjustment parameter be λ, then the weight of the j-th index finally is:

[0132]

[0133] Among them, λ represents the adjustment parameter;

[0134] The comprehensive scoring and ranking unit uses the index weights and the standardized evaluation scores to comprehensively score the overall green and low-carbon operation level of the port, and gives a grade division through the multi-index fuzzy comprehensive evaluation algorithm. The formula for the comprehensive scoring algorithm is:

[0135] The comprehensive score S of each port i is obtained by weighted summation of the scores of each index. The formula is:

[0136] where S i represents the comprehensive score of the i-th port,

[0137] μ ij represents the standardized score of the i-th port under the j-th index;

[0138] The multi-index fuzzy comprehensive evaluation algorithm is:

[0139] where B represents the fuzzy adjustment coefficient;

[0140] The model verification and feedback adjustment unit verifies, analyzes errors and makes feedback adjustments to the entire green port evaluation model to ensure that the evaluation results not only conform to historical data comparison but also can adapt to future dynamic changes.

[0141] Preferably, the intelligent optimization and decision support module includes:

[0142] The data analysis and problem modeling unit conducts a preliminary analysis of the operation data of the green port, identifies the important factors affecting green operation, and constructs a mathematical model according to the actual problems. The formula of its model algorithm is:

[0143] minf(x) = α1·C(x) + α2·E(x) + α3·F(x),

[0144] where C(x) is the carbon emission function, E(x) represents the energy consumption function, F(x) represents the relevant costs of facility construction and improvement, and α1, α2, α3 represent the weight parameters;

[0145] The optimization algorithm and model solution unit solves the multi-objective optimization model through the particle swarm optimization algorithm to obtain the best decision-making plan. The update formula is:

[0146]

[0147] x i (k + 1) = x i (k) + v i (k + 1),

[0148] where v i(k) represents the velocity of the i-th particle, and x i (k) represents the position of the i-th particle.

[0149] represents the historical best position of the i-th particle, represents the global best position.

[0150] c1 and c2 represent learning factors, r1 and r2 represent random numbers, and ω represents the inertia weight.

[0151] Preferably, the decision support and optimization result analysis unit visualizes and deeply analyzes the optimization results to help the decision maker understand and select the optimal solution. Among them,

[0152] Step 5.1 Visualization: Use graphical tools to display the optimization results;

[0153] Step 5.11 Conduct a sensitivity analysis on the key parameters of the optimization model to explore the impact of changes in these parameters on the optimization results. Use the Sobol method for global sensitivity analysis:

[0154] Among them, V i represents the contribution of the i-th parameter to the output, and V total represents the total contribution of all parameters to the output;

[0155] The adaptive adjustment and optimization update unit dynamically optimizes the decision support module using a reinforcement learning algorithm, enabling the model to adaptively adjust in different scenarios. Its algorithm formula is as follows:

[0156]

[0157] Q(s,a) represents the value of taking action a in state s, R(s,a) represents the immediate reward, γ represents the discount factor, and α represents the learning rate.

[0158] Preferably, the real-time warning module includes:

[0159] The anomaly detection and pattern recognition unit uses the Z-score detection method to detect anomalies in real-time data and identify values that are significantly different from the historical data pattern. The algorithm formula of the Z-score detection method is:

[0160]

[0161] Among them, X represents the current monitored value, μ represents the mean of the historical data, and σ represents the standard deviation of the historical data;

[0162] The risk assessment and early warning push unit combines the detected abnormal data with the environmental risk model to evaluate whether the abnormal event may cause a major environmental risk. The judgment algorithm formula is:

[0163] R = λ1·ΔC + λ2·ΔE + λ3·ΔF,

[0164] where R represents the risk score, ΔC represents the change in carbon emissions, and ΔE represents the change in energy consumption.

[0165] ΔF represents the change in facility failures or damages, and λ1, λ2, and λ3 represent weight parameters;

[0166] The dynamic adjustment and feedback mechanism unit. After the early warning is triggered, the system will provide real-time feedback on the specific abnormal data and risk assessment results to provide decision-making support for management personnel. And according to the early warning feedback, the port management system can make automatic adjustments. During the adjustment process, through methods such as reinforcement learning, continuously optimize the early warning and adjustment mechanisms.

[0167] The report generation and historical data analysis unit generates a detailed report on the real-time early warning data, risk assessment results, and adjustment measures, and provides it to port managers and relevant departments.

[0168] Preferably, a method for monitoring and evaluating a green and low-carbon port includes the following steps:

[0169] Step S1, through multi-type sensor devices deployed at the port, collect multi-dimensional environmental data, carbon emission data, and energy consumption data in real time;

[0170] Step S2, clean and preprocess the collected data, and use big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions;

[0171] Step S3, according to the processed data, calculate the carbon emissions of port activities through an accurate model;

[0172] Step S4, based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction, conduct a comprehensive evaluation to evaluate the operating performance of the port in terms of green and low-carbon, and help the port obtain the corresponding green port certification level;

[0173] Step S5, the system combines artificial intelligence algorithms to model and predict the port operation data, and provides an intelligent decision-making support function;

[0174] Step S6, by real-time monitoring of abnormal data during the port operation process, provide early warning information to ensure that the port can take timely measures to deal with environmental risks.

[0175] The present invention provides a green and low-carbon port monitoring and evaluation system. It has the following beneficial effects:

[0176] 1. By deploying multiple types of sensors, the present invention can collect multi-dimensional data during port operation in real time, forming a comprehensive carbon emission assessment system. This method of comprehensive collection of multi-dimensional data overcomes the limitation of traditional methods that can only monitor a single parameter, effectively making up for the problem of information fragmentation, and providing a solid foundation for subsequent data analysis and decision-making.

[0177] 2. The present invention uses the information entropy method to assign weights to sensor data, and combines multiple technologies such as weighted average, Hampel filtering, locally weighted regression, ICA, and grey relational analysis to clean, preprocess, and extract features from the data. This not only ensures the reliability of data quality, but also can deeply explore the relationship between energy use and carbon emissions, so as to provide a scientific and dynamic low-carbon operation assessment for port managers.

[0178] 3. By integrating various activity volume and energy consumption data, and combining fuzzy logic and non-linear correction models, the system of the present invention can accurately calculate the carbon emissions of various activities and perform unified standardization processing on the data. Compared with traditional methods, this accurate carbon footprint calculation method can more comprehensively and accurately reflect the overall carbon emission level of the port, providing reliable data support for green certification and carbon management.

[0179] 4. The present invention introduces intelligent algorithms such as robust regression based on the Huber loss function, local outlier factor (LOF), particle swarm optimization, and reinforcement learning, which can not only accurately predict historical data, but also monitor anomalies in real time and give early warnings. These intelligent optimization and decision support functions enable the port to shift from passive monitoring to active prediction, providing a scientific basis and operable suggestions for green development planning and optimal allocation of resources.

[0180] 5. By using the Z-score detection method to monitor anomalies in real-time data and combining with a risk assessment model, the system can quickly identify potential environmental risks and feedback information in a timely manner through a dynamic adjustment mechanism. This function greatly improves the response speed of the port to environmental anomalies and risks, ensuring that effective measures can be taken immediately to prevent accidents, guaranteeing the safety of port operation and maintaining the sustainability of low-carbon management. Description of the Drawings

[0181] Figure 1 is the system flow chart of the present invention;

[0182] Figure 2 is the method flow chart of the present invention. Detailed Embodiments

[0183] To enable those skilled in the art to understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0184] The following will describe the present invention in detail with reference to the accompanying drawings:

[0185] Embodiment:

[0186] Please refer to the attached Figure 1 , the embodiment of the present invention provides a green and low-carbon port monitoring and evaluation system, including:

[0187] The Internet of Things data acquisition module, through various types of sensor devices deployed in the port, collects multi-dimensional environmental data, carbon emission data, and energy consumption data in real time;

[0188] The data processing and analysis module cleans and preprocesses the collected data, and uses big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions;

[0189] The carbon footprint calculation module calculates the carbon emissions of port activities through an accurate model based on the processed data;

[0190] The green port level evaluation module conducts a comprehensive evaluation based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction, evaluates the operation performance of the port in terms of green and low-carbon, and helps the port obtain the corresponding green port certification level;

[0191] The intelligent optimization and decision support module combines artificial intelligence algorithms to model and predict the port operation data and provides an intelligent decision support function;

[0192] The real-time warning module provides warning information by real-time monitoring of abnormal data during the port operation process to ensure that the port can take timely measures to cope with environmental risks.

[0193] The Internet of Things data acquisition module includes:

[0194] The sensor unit includes: an energy consumption sensor for monitoring the energy consumption of various devices in the port; a carbon emission sensor for real-time monitoring of the emissions of carbon dioxide and other greenhouse gases during the port operation process; a waste emission sensor for monitoring the waste generated by the port and its emissions; an environmental quality sensor, which further includes: an air quality sensor and a water quality monitoring sensor for monitoring the environmental quality around the port, especially the pollutant concentrations in the air and water area;

[0195] The data acquisition and transmission system unit, where sensors collect data on energy consumption, carbon emissions, and waste emissions in real time, generate an original data stream, and use wireless technology to transmit the data collected by the sensors to the data platform in real time.

[0196] The multi-source data integration unit integrates data from different types of sensors. The fusion steps include:

[0197] Step 1.1 Determine its weight in data fusion by calculating the information entropy of each sensor's data. The formula is:

[0198] For the data sequence x of each sensor i i , discretize the data into M intervals, and then calculate the probability of the data appearing in each interval:

[0199] Among them, p ij represents the probability of sensor i in the jth interval, N ij represents the count of sensor i's data in the jth interval, and N i represents the total number of sensor i's data;

[0200] Step 1.2 Calculate the information entropy according to the probability distribution:

[0201] Among them, H i represents the information entropy of sensor i;

[0202] Step 1.3 Determine the weight using the information entropy. The lower the weight, the greater the data fluctuation and uncertainty, and its influence is relatively reduced:

[0203] Among them, w i represents the weight of sensor i, and n represents the total number of sensors;

[0204] Step 1.4 Obtain the fusion result by using the weighted average method:

[0205] Among them, x f represents the fused data value.

[0206] Specifically, by deploying multiple types of sensors, it is possible to collect multi-dimensional data during port operation in real time, forming a comprehensive carbon emission assessment system. This method of comprehensive multi-dimensional data collection overcomes the limitation of traditional methods that can only monitor a single parameter, effectively makes up for the problem of information fragmentation, and provides a solid foundation for subsequent data analysis and decision-making.

[0207] The data processing and analysis module includes:

[0208] Data cleaning and preprocessing unit, which uses the Hampel filter to remove outliers, locally weighted regression to fill in missing data, and exponential normalization to unify the data format, including:

[0209] Step 2.1 The algorithm formula of the Hampel filter is:

[0210] Within the sliding window centered on the target data point x i , calculate the median of the data within the window:

[0211] median(x i-k ,…,x i+k ),

[0212] Calculate the median absolute deviation (MAD):

[0213] MAD = median(|x j - median(x i-k ,…,x i+k )|),

[0214] where x j represents each data point within the window;

[0215] Judge whether the deviation of each data point from the median exceeds the threshold MAD:

[0216]

[0217] where k represents the sliding window size, λ represents the threshold factor, x i represents the original data,

[0218] represents the data after filtering;

[0219] Step 2.11 Use the linear interpolation method to fill in the missing values according to the local trend of the data. The interpolation formula is: If x i-1 and x i+1 are known at two points, the estimated value of the missing point x i is:

[0220]

[0221] Step 2.12 Use the exponential normalization method to map data with different dimensions to the same standard interval, which can not only compress the influence of extreme values but also retain the subtle differences in the data. The normalization algorithm formula:

[0222]

[0223] where x norm represents the normalized data value, and x represents the original sensor data.

[0224] x min represents the minimum and maximum values of the sensor data; max represents the minimum and maximum values of the sensor data;

[0225] The data feature extraction and correlation analysis unit extracts key features through ICA and quantitatively evaluates the correlation degree between energy use and carbon emissions using grey relational analysis. The key features extracted by ICA include:

[0226] Step 2.2 decomposes the multi-dimensional data into several independent components. It is assumed that the observed data is a mixture of multiple statistically independent source signals.

[0227] First, subtract the mean of each variable: x centered = x - E[x],

[0228] Use eigenvalue decomposition to transform the data into a unit covariance matrix:

[0229]

[0230] where E represents the eigenvector matrix and D represents the eigenvalue matrix;

[0231] Next, Fast ICA optimizes the demixing matrix through fixed-point iteration with the goal of maximizing the non-Gaussianity of the signal. Use the nonlinear function g(u):

[0232] w new = E[zg(w T z)] - E[g ′ (w T z)]w,

[0233] where g(u) is the selected nonlinear function, usually tanh(u);

[0234] Finally, through the weight vector w obtained by iteration, form the demixing matrix W, and finally extract the independent component: s = Wz;

[0235] Step 2.21 adopts the method used in traditional statistical grey relational analysis to quantitatively measure the correlation degree between each index:

[0236] For the reference sequence x0(k) and the comparison sequence x i (k), calculate the difference at each moment. The grey correlation coefficient is:

[0237]

[0238] For each comparison sequence, calculate its correlation degree:

[0239] Among them, x0(k) represents the data of the reference sequence at time k, and x i (k) represents the data of the i-th comparison sequence at time k, n represents the total number of observations, and ρ represents the identification coefficient;

[0240] The data modeling and intelligent prediction unit uses robust regression based on the Huber loss function to construct a prediction model, and at the same time detects abnormal patterns by means of an improved Local Outlier Factor (LOF), and tunes the model through cross-validation, specifically including:

[0241] Step 2.3 adopts robust regression based on the Huber loss function to reduce the influence of outliers on model fitting. Its algorithm formula is:

[0242]

[0243] Among them, the Huber loss function is defined as:

[0244]

[0245] Among them, y i represents the actual output of the i-th sample, and x ij represents the j-th feature of the i-th sample, and β0, β j represent the model intercept and regression coefficient respectively, and δ represents the critical value, which is set according to the data noise level;

[0246] Step 2.31 The Local Outlier Factor (LOF) algorithm is used to identify local abnormal patterns that appear in energy usage and carbon emission data. Its algorithm formula is as follows:

[0247]

[0248] Among them, the local reachability density lrd(p) is defined as:

[0249] Among them, p represents the target data point, and N k (p) represents the k nearest neighbors of the data point p,

[0250] reach-dis(p,q) represents the reachable distance between the point p and its neighbor q, and k represents the number of nearest neighbors;

[0251] The algorithm formula for step 2.32 cross-validation is:

[0252]

[0253] Among them, K represents the number of folds of cross-validation, and V k represents the set of sample indices included in the K-th fold validation set, and |V k | represents the K-th fold validation set (V kThe number of samples contained in

[0254] y i represents the true label or true value of sample i in the validation set, represents the output obtained by predicting sample i in the validation set using the model trained with all the remaining folds except the K-th fold validation set. L(·) is a measure of the difference between the true value and the predicted value, CV K represents the average loss value of the entire K-fold cross-validation process;

[0255] Result visualization and decision support unit, which visually displays the analysis results, generates a decision support report, and provides a basis for actual management and strategy optimization, including:

[0256] Step 2.4 Data visualization: Plot trend charts, scatter plots, heat maps, etc. to display data distribution, variable correlations, and the comparison between predicted results and actual values, and use time series charts to analyze the dynamic change laws of energy use and carbon emissions;

[0257] Step 2.41 Generation of decision support report: Comprehensively analyze the model output results, generate a comprehensive report including risk warnings, optimization suggestions, etc., transform data insights into management decision-making suggestions, and provide a quantitative basis for energy management and low-carbon strategies.

[0258] Specifically, the present invention uses the information entropy method to assign weights to sensor data, and combines multiple technologies such as weighted average, Hampel filtering, locally weighted regression, ICA, and grey relational analysis to clean, preprocess, and extract features from the data. This not only ensures the reliability of data quality, but also can deeply explore the relationship between energy use and carbon emissions, so as to provide scientific and dynamic low-carbon operation evaluations for port managers.

[0259] The carbon footprint calculation module includes:

[0260] Data input and verification unit, which extracts and verifies various data related to port activities according to the processed data to ensure data quality;

[0261] Activity volume and energy consumption data integration unit, which summarizes, merges, and standardizes multi-source data, quantifies the energy consumption indicators of various activities, and the processing method is as follows:

[0262] Step 3.1 For each port activity category, perform weighted average on data from different sources, and the formula is as follows:

[0263]

[0264] Among them, A integrated represents the integrated activity volume or energy consumption indicator, w iRepresents the weight of the i-th data source, A i Represents the activity volume and energy consumption data of the i-th data source;

[0265] In step 3.11, fuzzy logic is used to standardize the data to handle the possible scale differences and uncertainties between different data sources. The standardization formula is:

[0266]

[0267] Among them, A normal ized Represents the standardized activity volume or energy consumption value,

[0268] μ j Represents the membership degree of the j-th data source, A j Represents the activity volume or energy consumption data of the j-th data source, and m represents the total number of data sources participating in the standardization;

[0269] In step 3.12, when integrating multiple data sources, the weighted average and fuzzy logic methods are combined to obtain the final integrated data. The calculation formula is:

[0270]

[0271] Among them, A final Represents the final integrated data, w j Represents the weight of the j-th data source;

[0272] In practical applications, it is necessary to adjust the membership function and weighted coefficient of fuzzy logic to adapt to the characteristics of different data sources and activity types. The adjustment formula is:

[0273]

[0274] α represents the translation parameter of the membership function, and β represents the shape parameter of the membership function.

[0275] The carbon emission factor matching and adjustment unit uses a non-linear correction model and a fuzzy comprehensive evaluation method to perform non-linear correction and multi-factor dynamic adjustment on the basic emission factor respectively,

[0276] The algorithm formula of the non-linear correction model is:

[0277]

[0278] A i Represents the quantization index of the i-th type of activity,

[0279] EF i Represents the basic carbon emission factor of the i-th type of activity, β represents the non-linear correction coefficient, A ref Represents the reference activity volume;

[0280] Regarding the influence of multiple factors on the carbon emission factor in port operations, fuzzy comprehensive evaluation is used to dynamically adjust the emission factor, and the formula is defined as:

[0281]

[0282] Among them, EF j represents the carbon emission factor value under factor j, and the adjusted factor EF fuzzy will be used for subsequent carbon emission calculations;

[0283] The carbon emission calculation unit combines the integrated data with the adjusted emission factor, and uses an exact formula to calculate the carbon emissions of each activity and accumulate them to obtain the total emissions. The calculation steps are as follows:

[0284] Step 3.2 Matches the integrated activity data A i with the corresponding basic emission factor EF i and the fuzzy-adjusted factor EF fuzzy ;

[0285] Step 3.21 Calculates the carbon emissions of each activity using the above non-linear correction formula, and then accumulates them to obtain the total emissions;

[0286] Step 3.22 Normalizes and converts the units of the calculation results to ensure that the final output meets international or industry standards;

[0287] The model verification and error analysis unit verifies the accuracy of the model and optimizes and adjusts the parameters through historical data comparison, sensitivity analysis, and uncertainty assessment. The steps include:

[0288] Step 3.3 Calculates the uncertainty of the model output using Monte Carlo simulation by randomly sampling the input parameters of the model. Assume the model is: y = f(θ1, θ2, …, θ p ),

[0289] Among them, θ1 represents the uncertain parameter,

[0290] After N samplings, the simulated output y k (k = 1, 2, …, N) is obtained, and the output mean and standard deviation are calculated;

[0291]

[0292] Among them, y k represents the model output value obtained from the k-th simulation, represents the mean of the simulated output,

[0293] σ MCThe uncertainty estimation value representing the model output

[0294] In step 3.31, the model prediction value is compared with the historical monitoring data, and the root mean square error (RMSE) is used as the error measurement index. The calculation formula is as follows:

[0295]

[0296] where n represents the number of historical data samples, and y sim,i represents the predicted value of the model for the i-th sample, and y hist,i represents the historical observation value corresponding to the i-th sample;

[0297] In step 3.32, the model prediction error is combined with the Monte Carlo uncertainty estimation to construct a comprehensive error index E total , and the formula is as follows: E total = A·RMSE+(1 - A)·σ MC ,

[0298] where A represents the weighting coefficient, and E total COMPREHENSIVELY reflects the overall error index brought by the model prediction error and parameter uncertainty;

[0299] In step 3.4, to clarify the contribution of each uncertain parameter to the model output, the first-order sensitivity index of each parameter is calculated using Sobol sensitivity analysis. The formula is as follows:

[0300]

[0301] where S i represents the first-order sensitivity index of the parameter θ i , Var(y) represents the total variance of the model output,

[0302] represents the model output obtained by taking the expectation of the uncertainty of other parameters under the condition of fixing the parameter θ i , represents the variance of the mean value of the model output when θ i changes.

[0303] Specifically, by integrating various activity volume and energy consumption data, and combining fuzzy logic and non-linear correction models, the system of the present invention can accurately calculate the carbon emissions of each activity and perform unified standardization processing on the data. Compared with traditional methods, this accurate carbon footprint calculation method can more comprehensively and accurately reflect the overall carbon emission level of the port, providing reliable data support for green certification and carbon management.

[0304] The green port grade evaluation module includes:

[0305] The index construction and weight assignment unit constructs a quantization model for various evaluation indexes of a green port and assigns the weights of each index. The method is as follows:

[0306] Step 4.1 realizes the construction of the quantization model by using data normalization and fuzzification processing;

[0307] Step 4.2 adopts mutual information weight calculation, introduces a reference variable Y, where Y is the expert evaluation, historical comprehensive score, and green certification result, and calculates the mutual information between the j-th index and Y:

[0308]

[0309] Normalize the mutual information to obtain the mutual information weight component:

[0310] Among them, MI j represents the mutual information between the j-th index and the reference variable Y, and p(z ij ,y) represents the joint probability distribution of the index normalization value and Y, and p(z ij ) and p(y) represent the marginal probability distributions respectively. represents the weight score obtained based on the mutual information normalization;

[0311] Step 4.21 obtains the final weight in a weighted combination manner. Let the adjustment parameter be λ, then the weight of the j-th index finally is:

[0312]

[0313] Among them, λ represents the adjustment parameter;

[0314] The comprehensive scoring and ranking unit uses the index weights and the standardized evaluation scores to comprehensively score the overall green and low-carbon operation level of the port, and gives a grade division through the multi-index fuzzy comprehensive evaluation algorithm. The comprehensive scoring algorithm formula is:

[0315] The comprehensive score S of each port i is obtained by weighted summation of the scores of each index. The formula is:

[0316] Among them, S i represents the comprehensive score of the i-th port,

[0317] μ ij represents the standardized score of the i-th port under the j-th index;

[0318] The multi-index fuzzy comprehensive evaluation algorithm is:

[0319] Among them, B represents the fuzzy adjustment coefficient;

[0320] The model verification and feedback adjustment unit verifies, analyzes errors, and makes feedback adjustments to the entire green port evaluation model to ensure that the evaluation results not only conform to historical data comparison but also can adapt to future dynamic changes.

[0321] The intelligent optimization and decision support module includes:

[0322] The data analysis and problem modeling unit conducts preliminary analysis on the operation data of the green port, identifies important factors affecting green operation, and constructs a mathematical model based on actual problems. The model algorithm formula is:

[0323] minf(x) = α1·C(x) + α2·E(x) + α3·F(x),

[0324] where C(x) is the carbon emission function, E(x) represents the energy consumption function, F(x) represents the relevant costs of facility construction and improvement, and α1, α2, α3 represent weight parameters;

[0325] The optimization algorithm and model solution unit solves the multi-objective optimization model through the particle swarm optimization algorithm to obtain the best decision-making plan. The update formula is:

[0326]

[0327] x i (k + 1) = x i (k) + v i (k + 1),

[0328] where v i (k) represents the velocity of the i-th particle, and x i (k) represents the position of the i-th particle.

[0329] represents the historical optimal position of the i-th particle, represents the global optimal position.

[0330] c1 and c2 represent learning factors, r1 and r2 represent random numbers, and ω represents the inertia weight.

[0331] The decision support and optimization result analysis unit visualizes and deeply analyzes the optimization results to help decision-makers understand and select the optimal plan. Among them,

[0332] Step 5.1 Visualization: Use graphical tools to display the optimization results;

[0333] Step 5.11 Conduct sensitivity analysis on the key parameters of the optimization model to explore the impact of changes in these parameters on the optimization results. The Sobol method is used for global sensitivity analysis:

[0334] Among them, V i represents the contribution of the i-th parameter to the output, and V total represents the total contribution of all parameters to the output;

[0335] An adaptive adjustment and optimization update unit uses a reinforcement learning algorithm to dynamically optimize the decision support module, enabling the model to adaptively adjust in different scenarios. Its algorithm formula is as follows:

[0336]

[0337] Q(s,a) represents the value of taking action a in state s, R(s,a) represents the immediate reward, γ represents the discount factor, and α represents the learning rate.

[0338] Specifically, the present invention introduces intelligent algorithms such as robust regression based on the Huber loss function, local outlier factor (LOF), particle swarm optimization, and reinforcement learning, which can not only accurately predict historical data but also monitor anomalies in real time and give early warnings. These intelligent optimization and decision support functions enable the port to shift from passive monitoring to active prediction, providing a scientific basis and operational suggestions for green development planning and resource optimization allocation

[0339] The real-time warning module includes:

[0340] An anomaly detection and pattern recognition unit uses the Z-score detection method to detect anomalies in real-time data and identify values that are significantly different from the historical data pattern. The algorithm formula of the Z-score detection method is:

[0341]

[0342] Among them, X represents the current monitored value, μ represents the mean of historical data, and σ represents the standard deviation of historical data;

[0343] A risk assessment and warning push unit combines the detected abnormal data with the environmental risk model to evaluate whether the abnormal event is likely to cause a large environmental risk. The judgment algorithm formula is:

[0344] R = λ1·ΔC + λ2·ΔE + λ3·ΔF,

[0345] Among them, R represents the risk score, ΔC represents the change in carbon emissions, ΔE represents the change in energy consumption,

[0346] ΔF represents the change in facility failure or damage, and λ1, λ2, λ3 represent weight parameters;

[0347] The dynamic adjustment and feedback mechanism unit. After the early warning is triggered, the system will provide real-time feedback on specific abnormal data and risk assessment results to provide decision-making support for management personnel. And according to the early warning feedback, the port management system can make automatic adjustments. During the adjustment process, through methods such as reinforcement learning, the early warning and adjustment mechanisms are continuously optimized.

[0348] The report generation and historical data analysis unit generates a detailed report on real-time early warning data, risk assessment results and adjustment measures and provides it to port managers and relevant departments.

[0349] Specifically, the present invention uses the Z-score detection method to monitor abnormal real-time data and combines a risk assessment model. The system can quickly identify potential environmental risks and provide timely feedback information through a dynamic adjustment mechanism. This function greatly improves the port's response speed to environmental anomalies and risks, ensures that effective measures can be taken in the first time to prevent accidents, and guarantees the safety of port operations and maintains the sustainability of low-carbon management.

[0350] A method for monitoring and evaluating a green and low-carbon port includes the following steps:

[0351] Step S1, through multi-type sensor devices deployed at the port, real-time collect multi-dimensional environmental data, carbon emission data and energy consumption data;

[0352] Step S2, clean and preprocess the collected data, and use big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions;

[0353] Step S3, according to the processed data, calculate the carbon emissions of port activities through an accurate model;

[0354] Step S4, based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction, conduct a comprehensive evaluation to evaluate the port's operation performance in terms of green and low-carbon, and help the port obtain the corresponding green port certification level;

[0355] Step S5, the system combines artificial intelligence algorithms to model and predict the port operation data and provide an intelligent decision-making support function;

[0356] Step S6, through real-time monitoring of abnormal data during port operation, provide early warning information to ensure that the port can take timely measures to respond to environmental risks.

[0357] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A green and low-carbon port monitoring and evaluation system, characterized in that: include: The IoT data collection module collects multi-dimensional environmental data, carbon emission data, and energy consumption data in real time through various types of sensor devices deployed at the port; The data processing and analysis module cleans and preprocesses the collected data, and uses big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions; Carbon footprint calculation module, which calculates the carbon emissions of port activities through accurate models based on the processed data; The green port rating evaluation module conducts a comprehensive evaluation based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction to assess the port's green and low-carbon operational performance and help ports obtain the corresponding green port certification level; Intelligent optimization and decision support module: the system combines artificial intelligence algorithms to model and predict port operation data and provide intelligent decision support functions; The real-time early warning module provides early warning information by monitoring abnormal data during port operations in real time, ensuring that the port can take timely measures to deal with environmental risks.

2. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The Internet of Things data acquisition module includes: Sensor units include: energy consumption sensors, which are used to monitor the energy consumption of various equipment in the port; carbon emission sensors, which are used to monitor the emission of carbon dioxide and other greenhouse gases generated during the port operation in real time; waste emission sensors, which are used to monitor the waste generated by the port and its emission; environmental quality sensors, including: air quality sensors, water quality monitoring sensors and noise monitoring sensors, which are used to monitor the ambient air, water quality and environmental noise in the port area; Data collection and transmission system unit, sensors collect data on energy consumption, carbon emissions, and waste emissions in real time, generate raw data streams, and use wireless technology to transmit the data collected by sensors to the data platform in real time. The multi-source data integration unit integrates data from different types of sensors. The fusion steps include: Step 1.1: Determine the weight of each sensor data in data fusion by calculating its information entropy. The formula is: For each sensor i’s data sequence x i , discretize the data into M intervals, and then calculate the probability of data occurrence in each interval: Among them, p ij represents the probability that sensor i is in the jth interval, N ij represents the count of sensor i’s data in the jth interval, N i Represents the total number of sensor i data; Step 1.2 Calculate the information entropy based on the probability distribution: Among them, H i represents the information entropy of sensor i; Step 1.3 uses information entropy to determine the weight. The lower the weight, the greater the data fluctuation and the higher the uncertainty, and the lower its influence: Among them, w i represents the weight of sensor i, and n represents the total number of sensors; Step 1.4 uses weighted average to obtain the fusion result: Among them, x f Represents the fused data value.

3. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The data processing and analysis module includes: The data cleaning and preprocessing unit uses the Hampel filter to remove outliers, local weighted regression to fill in missing data, and uses exponential normalization to unify the data format, including: Step 2.1 The Hampel filter algorithm formula is: At the target data point x i In a sliding window centered on , calculate the median of the data in the window: median(x i-k ,…,x i+k ), Calculate the median absolute deviation (MAD): MAD=median(|x j -median(x i-k ,…,x i+k )|), Among them, x j represents each data point within the window; Determine whether the deviation of each data point from the median exceeds the threshold MAD: Among them, k represents the sliding window size, λ represents the threshold factor, and x i Represents the original data, Represents the data after filtering; Step 2.11 uses linear interpolation to fill in missing values ​​based on the local trend of the data. The interpolation formula is: i-1 and x i+1 Two points are known, and the missing point is x i The estimated value of is: Step 2.12 uses the exponential normalization method to map data of different dimensions to the same standard range, which can not only compress the influence of extreme values ​​but also retain the subtle differences in data. The normalization algorithm formula is: Among them, x norm represents the normalized data value, x represents the original sensor data, x min With x max Indicates the minimum and maximum values ​​of sensor data; The data feature extraction and correlation analysis unit extracts key features through ICA and uses grey correlation analysis to quantitatively evaluate the correlation between energy use and carbon emissions. The key features extracted by ICA include: Step 2.2 decomposes the multidimensional data into several independent components, assuming that the observed data is a mixture of multiple statistically independent source signals. First, the mean of each variable is subtracted: centered =xE[x], Use eigenvalue decomposition to transform the data into a unit covariance matrix: Where E represents the eigenvector matrix, and D represents the eigenvalue matrix; Next, FastICA optimizes the unmixing matrix through fixed-point iterations, with the goal of maximizing the non-Gaussianity of the signal. The nonlinear function g(u) is used: w new =E[zg(w T z)]-E[g ′ (w T z)]w, Among them, g(u) is the selected nonlinear function, usually tanh(u); Finally, the weight vector w obtained by iteration forms the unmixing matrix W, and finally extracts the independent components: s = Wz; Step 2.21 uses the grey correlation analysis method used in traditional statistics to quantitatively measure the degree of correlation between the indicators: For the reference sequence x0(k) and the comparison sequence x i (k), calculate the difference at each moment, and the grey correlation coefficient is: For each comparison sequence, calculate its association: Among them, x0(k) represents the data of the reference sequence at time k, x i (k) represents the data of the ith comparison sequence at time k, n represents the total number of observations, and ρ represents the identification coefficient; The data modeling and intelligent prediction unit uses robust regression based on the Huber loss function to build a prediction model, detects abnormal patterns with the help of an improved local outlier factor (LOF), and performs model tuning through cross-validation, including: Step 2.3 uses robust regression based on Huber loss function to reduce the impact of outliers on model fitting. The algorithm formula is: The Huber loss function is defined as: Among them, y i represents the actual output of the i-th sample, x ij represents the jth feature of the i-th sample, β0, β j Represent the model intercept and regression coefficient, respectively, and δ represents the critical value, which is set according to the data noise level; Step 2.31 The local outlier factor (LOF) algorithm is used to identify local abnormal patterns in energy usage and carbon emission data. The algorithm formula is as follows: The local reachability density lrd(p) is defined as: Among them, p represents the target data point, N k (p) represents the k nearest neighbors of data point p, reach-dis(p,q) represents the reachable distance between point p and its neighbor q, and k represents the number of neighbors; Step 2.32 The algorithm formula for cross validation is: Among them, K represents the number of cross-validation folds, V k represents the sample index set contained in the K-fold validation set, |V k | represents the K-fold validation set (V k ), y i represents the true label or true value of sample i in the validation set, It represents the output obtained by predicting sample i in the validation set using the model trained with all the remaining folds except the K-th fold validation set. L(·) is a measure of the difference between the true value and the predicted value. CV K Represents the average loss value of the entire K-fold cross validation process; The result visualization and decision support unit displays the analysis results intuitively and generates decision support reports to provide a basis for actual management and strategy optimization, including: Step 2.4 Data visualization: Draw trend graphs, scatter plots, heat maps, etc. to show data distribution, variable correlation, and comparison between predicted results and actual values. Use time series graphs to analyze the dynamic changes in energy use and carbon emissions. Step 2.41 Decision support report generation: Comprehensively analyze the output results of the model to generate a comprehensive report containing risk warnings, optimization suggestions, etc., convert data insights into management decision recommendations, and provide a quantitative basis for energy management and low-carbon strategies.

4. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The carbon footprint calculation module includes: The data input and verification unit extracts and verifies various data related to port activities based on the processed data to ensure data quality; The activity volume and energy consumption data integration unit aggregates, merges and standardizes multi-source data to quantify the energy consumption indicators of various activities. The processing method is as follows: Step 3.1 For each port activity category, a weighted average of the data from different sources is calculated using the following formula: Among them, A integrated represents the integrated activity or energy consumption index, w i represents the weight of the i-th data source, A i represents the activity and energy consumption data of the i-th data source; Step 3.11: Use fuzzy logic to standardize the data to deal with the scale differences and uncertainties that may exist between different data sources. The standardization formula is: Among them, A normalized Represents the standardized activity or energy consumption value, μ j represents the membership degree of the jth data source, A j represents the activity or energy consumption data of the jth data source, and m represents the total number of data sources involved in standardization; Step 3.12: When integrating multiple data sources, the weighted average and fuzzy logic methods are combined to obtain the final integrated data. The calculation formula is: Among them, A final represents the final integrated data, w j represents the weight of the j-th data source; Step 3.13 In practical applications, the membership function and weighting coefficient of fuzzy logic need to be adjusted to adapt to the characteristics of different data sources and activity types. The adjustment formula is: α represents the translation parameter of the membership function, and β represents the shape parameter of the membership function.

5. A green and low-carbon port monitoring and evaluation system according to claim 4, characterized in that: The carbon emission factor matching and adjustment unit adopts a nonlinear correction model and a fuzzy comprehensive evaluation method to perform nonlinear correction and multi-factor dynamic adjustment on the basic emission factor respectively. The algorithm formula of the nonlinear correction model is: A i represents the quantitative indicator of the i-th type of activity, EF i represents the basic carbon emission factor of the i-th activity, β represents the nonlinear correction coefficient, and A ref represents the reference activity amount; In view of the impact of multiple factors on carbon emission factors in port operations, fuzzy comprehensive evaluation is used to dynamically adjust the emission factors. The formula is defined as: Among them, EF j Indicates the carbon emission factor value under factor j, the adjusted factor EF fuzzy Will be used for subsequent carbon emissions calculations; The carbon emission calculation unit combines the integrated data with the adjusted emission factor, uses an accurate formula to calculate the carbon emissions of each activity and accumulates them to obtain the total emissions. The calculation steps are as follows: Step 3.2: The integrated activity data A i and the corresponding basic emission factor EF i and the fuzzy adjusted factor EF fuzzy Make a match; Step 3.21 Use the above nonlinear correction formula to calculate the carbon emissions of each activity, and then add them up to get the total emissions; Step 3.22 Normalize and convert the calculation results to ensure that the final output complies with international or industry standards; The model verification and error analysis unit verifies the accuracy of the model and optimizes the parameters through historical data comparison, sensitivity analysis and uncertainty assessment. The steps include: Step 3.3: Use Monte Carlo simulation to calculate the uncertainty of the model output by randomly sampling the model input parameters. Suppose the model is: y = f(θ1, θ2, …, θ p ), Among them, θ1 represents the uncertain parameter, After N samplings, the analog output y is obtained. k (k=1,2,…,N), output mean and standard deviation; Among them, y k represents the model output value obtained from the kth simulation, represents the mean value of the simulation output, σ MC Represents the uncertainty estimate of the model output Step 3.31 compares the model prediction value with the historical monitoring data, and uses the root mean square error (RMSE) as the error measurement indicator. The calculation formula is: Where n represents the number of historical data samples, y sim,i Represents the model's predicted value for the i-th sample, y hist,i Represents the historical observation value corresponding to the i-th sample; Step 3.32 Combine the model prediction error with the Monte Carlo uncertainty estimate to construct a comprehensive error index E total , the formula is as follows: E total =A·RMSE+(1-A)·σ MC , Among them, A represents the weighting coefficient, E total BIAOSHI comprehensively reflects the overall error index caused by model prediction error and parameter uncertainty; Step 3.4 To clarify the contribution of each uncertain parameter to the model output, Sobol sensitivity analysis is used to calculate the first-order sensitivity index of each parameter. The formula is: Among them, S i Denotes the parameter θ i The first-order sensitivity index of , Var(y) represents the total variance of the model output, Indicates that at a fixed parameter θ i Under the condition of , the model output is obtained by taking the expectation of the uncertainty of other parameters. Indicates that the change θ i The variance of the model output mean when .

6. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The green port grade evaluation module includes: The indicator construction and weight allocation unit will construct a quantitative model for each evaluation indicator of the green port and allocate the weight of each indicator as follows: Step 4.1: Use data normalization and fuzzification processing to build a quantitative model; Step 4.2 uses mutual information weight calculation, introduces reference variable Y, Y is the expert evaluation, historical comprehensive score and green certification result, and calculates the mutual information between the jth indicator and Y: Normalize the mutual information to get the mutual information weight component: Among them, MI j represents the mutual information between the jth indicator and the reference variable Y, p(z ij ,y) represents the joint probability distribution of the normalized value of the indicator and Y, p(z ij ) and p(y) represent the marginal probability distribution, Represents the weight score obtained after normalization based on mutual information; Step 4.21 uses weighted combination to get the final weight. Assume the adjustment parameter λ, then the final weight of the jth indicator is: Among them, λ represents the adjustment parameter; The comprehensive scoring and ranking unit uses the indicator weights and standardized evaluation scores to comprehensively score the overall green and low-carbon operation level of the port, and gives a grade classification through a multi-indicator fuzzy comprehensive evaluation algorithm. The comprehensive scoring algorithm formula is: The overall score of each port is S i It is obtained by weighted summation of scores of various indicators, and the formula is: Among them, S i represents the comprehensive score of the i-th port, μ ij represents the standardized score of the i-th port under the j-th indicator; The multi-index fuzzy comprehensive evaluation algorithm is: Wherein, B represents the fuzzy adjustment coefficient; The model verification and feedback adjustment unit verifies, analyzes errors and makes feedback adjustments to the entire green port evaluation model to ensure that the evaluation results are consistent with historical data comparisons and can adapt to future dynamic changes.

7. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The intelligent optimization and decision support module includes: The data analysis and problem modeling unit conducts preliminary analysis on the operation data of the green port, identifies the important factors affecting green operation, and builds a mathematical model based on actual problems. The model algorithm formula is as follows: minf(x)=α1·C(x)+α2·E(x)+α3·F(x), Among them, C(x) represents the carbon emission function, E(x) represents the energy consumption function, F(x) represents the relevant costs of facility construction and improvement, and α1, α2, and α3 represent weight parameters; The optimization algorithm and model solving unit solves the multi-objective optimization model through the particle swarm optimization algorithm to obtain the best decision-making solution. The update formula is: x i (k+1)=x i (k)+v i (k+1), Among them, v i (k) represents the velocity of the ith particle, x i (k) represents the position of the i-th particle. represents the historical optimal position of the i-th particle, represents the global optimal position. c1, c2 represent learning factors, r1, r2 represent random numbers, and ω represents the inertia weight.

8. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The decision support and optimization result analysis unit visualizes and deeply analyzes the optimization results to help decision makers understand and select the optimal solution, wherein: Step 5.1 Visualization: Use graphical tools to display the optimization results; Step 5.11: Perform sensitivity analysis on the key parameters of the optimization model to explore the impact of these parameter changes on the optimization results. Use the Sobol method to perform global sensitivity analysis: Among them, V i represents the contribution of the ith parameter to the output, V total Represents the total contribution of all parameters to the output; The adaptive adjustment and optimization update unit dynamically optimizes the decision support module using a reinforcement learning algorithm, so that the model can be adaptively adjusted in different scenarios. The algorithm formula is as follows: Q(s,a) represents the value of taking action a in state s, R(s,a) represents the immediate reward, γ represents the discount factor, and α represents the learning rate.

9. A green and low-carbon port monitoring and evaluation system according to claim 1, characterized in that: The real-time warning module includes: The anomaly detection and pattern recognition unit uses the Z-score detection method to detect anomalies in real-time data and identify values ​​that are significantly different from historical data patterns. The algorithm formula of the Z-score detection method is: Among them, X represents the current monitoring value, μ represents the mean of historical data, and σ represents the standard deviation of historical data; The risk assessment and early warning push unit combines the detected abnormal data with the environmental risk model to assess whether the abnormal event may cause a greater environmental risk. The judgment algorithm formula is: R=λ1·ΔC+λ2·ΔE+λ3·ΔF, Among them, R represents the risk score, ΔC represents the change in carbon emissions, and ΔE represents the change in energy consumption. ΔF represents the change of facility failure or damage, λ1, λ2, λ3 represent weight parameters; Dynamic adjustment and feedback mechanism unit. After the early warning is triggered, the system will provide real-time feedback on specific abnormal data and risk assessment results to provide decision support for managers. Based on the early warning feedback, the port management system can be automatically adjusted. During the adjustment process, the early warning and adjustment mechanism is continuously optimized through reinforcement learning and other methods. The report generation and historical data analysis unit generates detailed reports with real-time warning data, risk assessment results and adjustment measures, and provides them to port managers and relevant departments.

10. A green and low-carbon port monitoring and evaluation method, using a green and low-carbon port monitoring and evaluation system according to any one of claims 1 to 9, characterized in that: The steps include: Step S1, real-time collection of multi-dimensional environmental data, carbon emission data and energy consumption data through multiple types of sensor equipment deployed at the port; Step S2, cleaning and preprocessing the collected data, using big data analysis technology and artificial intelligence algorithms to explore the relationship between energy use and carbon emissions; Step S3, calculating the carbon emissions of port activities through an accurate model based on the processed data; Step S4, conduct a comprehensive evaluation based on multi-dimensional indicators such as carbon emissions, energy efficiency, and green facility construction to assess the port's green and low-carbon operational performance and help the port obtain the corresponding green port certification level; Step S5, the system combines artificial intelligence algorithms to model and predict port operation data and provide intelligent decision support functions; Step S6, by real-time monitoring of abnormal data during port operations, early warning information is provided to ensure that the port can take timely measures to deal with environmental risks.

Images