Blockchain and ai-based ship carbon emission credible monitoring method and system

Abstract

The application discloses a kind of based on blockchain and artificial intelligence's ship carbon emission credible monitoring method and system, belong to ship carbon emission monitoring and green shipping technical field, including by collecting engine vibration signal and fuel consumption data, and obtain speed, voyage and load information in combination with AIS system;Vibration signal is analyzed in frequency domain and constructs gray spectrum diagram, and fuel type is identified using convolutional neural network;Based on the deviation of real-time speed and declared speed and in combination with historical data to set dynamic threshold, realize abnormal behavior identification, carbon intensity index is calculated by fusing fuel type, voyage and load data, and grade division is completed according to IMO standard;Key data hash value is stored in chain and certified, and green certification and carbon credit distribution are automatically completed by smart contract.The application can effectively improve the accuracy and credibility of ship carbon emission monitoring.

Description

Technical Field

[0001] This invention relates to the field of ship carbon emission monitoring and green shipping technology, and in particular to a reliable method and system for ship carbon emission monitoring based on blockchain and AI. Background Technology

[0002] With the continuous advancement of global shipping industry carbon reduction policies, ship carbon emission monitoring and accounting technologies are gradually evolving from traditional manual statistical methods towards intelligent, digital, and reliable approaches. Regulatory frameworks such as the Carbon Intensity Index (CII) and Energy Efficient Design Index (EEDI) proposed by the International Maritime Organization (IMO) require refined quantification and tiered assessment of ship carbon emissions during navigation. Against this backdrop, existing technologies are increasingly incorporating IoT sensors to collect real-time data on multi-dimensional operational parameters such as ship fuel consumption, speed, and load, and using centralized platforms for carbon emission calculation and data management. Simultaneously, some research is attempting to introduce blockchain technology to enhance the immutability of data storage and artificial intelligence algorithms to improve data analysis and prediction capabilities. However, from an overall application perspective, existing technologies still generally suffer from problems such as difficulty in ensuring the authenticity of data sources, reliance on manual reporting for fuel type identification, difficulty in automatically identifying abnormal behavior, and low efficiency in multi-party collaboration. Especially when key parameters for carbon emission accounting (such as fuel type and fuel consumption data) are subject to human intervention, it can easily lead to distorted carbon emission results, thereby affecting the effectiveness of carbon trading, public credibility certification, and regulatory decisions.

[0003] Furthermore, existing technological systems lack an effective closed-loop mechanism in the "data acquisition-analysis modeling-result authentication" chain. On the one hand, traditional sensor systems can only acquire raw physical quantity data, lacking the ability to deeply analyze the semantic level of the data, and cannot deduce fuel type or operating status from implicit features such as vibration and acoustics. On the other hand, centralized database architectures are susceptible to single-point tampering risks during data storage and management, making it difficult to meet the trusted auditing requirements in cross-entity, multi-role scenarios. In addition, most existing carbon emission monitoring systems rely on manual review and offline authentication processes, lacking automated smart contract mechanisms, resulting in long authentication cycles and low efficiency, making it difficult to support the real-time monitoring needs of high-frequency voyage scenarios. Therefore, how to achieve intelligent identification, anomaly detection, and trusted storage of ship carbon emission data while ensuring data authenticity, and to build an automated authentication and incentive mechanism, has become a key issue that urgently needs to be addressed in the current technological field.

[0004] CN120069336B discloses a blockchain-based trusted management method for supply chain carbon data. This method collects energy consumption data from production equipment through an edge computing gateway and combines this data with data from logistics, transportation, and recycling to construct a multi-source data collection system. Blockchain is used for trusted data storage, and smart contracts enable compliance auditing. This method achieves end-to-end management and trusted processing of carbon data at the supply chain level, improving the completeness and audit efficiency of carbon emission accounting to some extent. However, this method mainly relies on structured energy consumption data input and lacks the ability to deeply mine raw physical signals (such as vibration and acoustic characteristics). It also lacks an automatic identification mechanism for key parameters (such as fuel type), and the authenticity of the data source still depends on external input, making it difficult to prevent data fraud at the source.

[0005] CN120069294A discloses a blockchain-based ship carbon tax accounting and carbon trading system. This system records ship carbon emissions through onboard terminals and generates a carbon footprint on the blockchain, enabling carbon rights allocation, carbon tax calculation, and carbon trading. This method uses blockchain technology to ensure the security and integrity of carbon emission data and establishes a carbon trading incentive mechanism, demonstrating certain application value in the shipping carbon management field. However, the system still primarily relies on carbon emission data input from onboard terminals or manually, lacking automatic verification methods for data authenticity and failing to incorporate artificial intelligence models to identify and analyze key factors such as fuel type and abnormal behavior. Furthermore, the method does not address dynamic detection of abnormal navigation behavior (such as deviations in speed declarations) or construct a refined carbon intensity assessment model based on multi-source data fusion, resulting in deficiencies in anti-fraud capabilities and data accuracy. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the problems that existing ship carbon emission monitoring technologies generally suffer from, such as reliance on manual reporting of data sources, difficulty in verifying the authenticity of key parameters, inability to automatically identify fuel types, and the ease with which data can be tampered with and low authentication efficiency under multi-party collaboration, this invention is proposed.

[0008] Therefore, the problem to be solved by this invention is how to achieve highly reliable collection of ship carbon emission data, intelligent identification of fuel type, automatic detection of abnormal behavior, and tamper-proof evidence storage and efficient authentication throughout the entire process in complex shipping environments.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a reliable monitoring method for ship carbon emissions based on blockchain and AI, comprising, S1: Collect vibration signals and fuel consumption data during engine operation, and synchronously obtain the ship's real-time speed, distance traveled, and load data via the AIS system; S2: Perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector. Convert the frequency domain feature vector into a grayscale spectrum and input it into a convolutional neural network to obtain the fuel type identification result. S3: Calculate the speed deviation between the real-time speed and the speed declared by the crew, set a dynamic threshold based on historical navigation records, and determine the abnormal behavior and generate an abnormal marker when the speed deviation is greater than or equal to the dynamic threshold. S4: Based on the fuel type identification results and range and load data, calculate the carbon intensity index value, and classify the carbon intensity index value according to the benchmark value specified by the IMO; S5: Upload the hash values ​​of vibration signals, fuel type identification results, anomaly markers, and grade classification results to the consortium blockchain for evidence storage, complete green certification through smart contracts, and issue corresponding carbon credits.

[0010] Secondly, embodiments of the present invention provide a trusted monitoring system for ship carbon emissions based on blockchain and AI, comprising: The data acquisition module is used to collect vibration signals and fuel consumption data during engine operation, and synchronously acquire the ship's real-time speed, distance traveled, and load data via the AIS system. The fuel identification module is used to perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector. After converting the frequency domain feature vector into a grayscale spectrum, it is input into a convolutional neural network to obtain the fuel type identification result. The anomaly monitoring module is used to calculate the speed deviation between the real-time speed and the speed declared by the crew. It sets a dynamic threshold based on historical navigation records. When the speed deviation is greater than or equal to the dynamic threshold, it is judged as abnormal behavior and an anomaly mark is generated. The carbon emission calculation module calculates the carbon intensity index value based on the fuel type identification result, flight distance and load data, and classifies the carbon intensity index value into levels according to the benchmark value stipulated by the IMO. The blockchain evidence storage module is used to upload the hash values ​​of vibration signals, fuel type identification results, anomaly markers, and grade classification results to the consortium blockchain for evidence storage, and complete green certification and issue corresponding carbon credits through smart contracts.

[0011] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the steps of the aforementioned blockchain and AI-based trusted monitoring system for ship carbon emissions.

[0012] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any of the steps of the aforementioned blockchain and AI-based trusted monitoring system for ship carbon emissions.

[0013] Compared with existing technologies, the advantages of this invention are as follows: By synchronously acquiring engine vibration signals and speed, range, and load data from the AIS system, spatiotemporal alignment and unified benchmarks of multi-source heterogeneous data are achieved, providing a reliable and complete basic data source for subsequent analysis. This avoids errors caused by data silos or time deviations, improving the overall data quality and consistency of carbon emission monitoring. The vibration signals are subjected to Fast Fourier Transform to obtain frequency domain feature vectors, which are then converted into grayscale spectrograms and input into a convolutional neural network for automatic feature extraction and classification. This enables non-invasive real-time identification of fuel types, accurately distinguishing between heavy and light fuels without the need for expensive fuel mass flow meters. This provides a precise basis for selecting carbon emission factors, significantly reducing hardware deployment costs and improving identification efficiency. The deviation between real-time speed and crew-reported speed is calculated, and a dynamic threshold is set based on historical voyage records. This adapts to the normal fluctuation range under different ship and navigation conditions, effectively avoiding false alarms or omissions caused by fixed thresholds. The system enables sensitive detection and labeling of abnormal behaviors such as falsified declared speeds, enhancing the credibility and anti-fraud capabilities of the carbon emission monitoring process. By integrating fuel type identification results, voyage distance, and load data to calculate carbon intensity index values, and classifying them into five levels (A to E) according to IMO-stipulated benchmarks, complex carbon emission data is transformed into standardized performance levels. This provides an intuitive and comparable quantitative assessment of the carbon emission efficiency of a single voyage, facilitating ship owners and management agencies to quickly grasp emission levels and formulate targeted improvement measures. By calculating hash values ​​for vibration signals, fuel type identification results, abnormal labeling, and level classification results, and uploading them to the consortium blockchain for evidence storage, and using smart contracts to automatically complete green certification and carbon credit issuance, the system ensures the immutability and traceability of the entire process from data collection, processing, to evaluation. At the same time, the use of automated smart contracts eliminates human intervention and intermediate links, greatly improving certification transparency and issuance efficiency, and effectively incentivizing ship owners to proactively reduce carbon emissions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1A flowchart illustrating a reliable monitoring method for ship carbon emissions based on blockchain and AI; Figure 2 Network structure diagram of CNN fuel identification model for a trusted monitoring method of ship carbon emissions based on blockchain and AI; Figure 3 A timeline diagram for smart contract authentication of a blockchain and AI-based trusted monitoring method for ship carbon emissions; Figure 4 A screenshot of the shipowner's homepage interface for a trusted monitoring method for ship carbon emissions based on blockchain and AI; Figure 5 Core page diagram for certification bodies of trusted monitoring methods for ship carbon emissions based on blockchain and AI; Figure 6 Core page diagram for port management based on a trusted monitoring method for ship carbon emissions using blockchain and AI; Figure 7 For regulatory agencies seeking a trusted monitoring method for ship carbon emissions based on blockchain and AI, please refer to the core page diagram. Figure 8 This is a structural diagram of a trusted monitoring system for ship carbon emissions based on blockchain and AI. Detailed Implementation

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0016] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0018] As mentioned in the background section, existing ship carbon emission monitoring technologies suffer from significant problems, including insufficient assurance of data authenticity, reliance on manual declaration of fuel type, lack of effective means of identifying abnormal behavior, and susceptibility to data tampering and low authentication efficiency in multi-party collaboration. To address these issues, this invention provides a reliable ship carbon emission monitoring method based on blockchain and AI.

[0019] Reference Figures 1-7 , Figure 1This is a flowchart illustrating a reliable monitoring method for ship carbon emissions based on blockchain and AI, according to an embodiment of the present invention. Figure 1 As shown, a reliable monitoring method for ship carbon emissions based on blockchain and AI includes: S1: Collect vibration signals and fuel consumption data during engine operation, and synchronously obtain the ship's real-time speed, range, and load data via the AIS system; Specifically, multi-source data is collected and a raw dataset is constructed: vibration sensors are installed on the surface of the ship's main engine cylinder head to obtain vibration signals, fuel consumption data is obtained through mass flow meters installed in the fuel line to form a raw fuel consumption sequence, and the ship's real-time speed, distance traveled, and load data are obtained synchronously with the acquisition time of the vibration sensors through the ship's deployed AIS system. Furthermore, the vibration signal, fuel consumption sequence, speed, range, and load together constitute the original dataset; if there are missing data segments in the fuel consumption sequence due to sensor failure or signal interruption, the missing time period and location are recorded and filled in using the LSTM model in subsequent step S2.5; if there are no missing segments, proceed directly to the next step. S2: Perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector. Convert the frequency domain feature vector into a grayscale spectrum and input it into a convolutional neural network to obtain the fuel type identification result. Furthermore, a Fast Fourier Transform is performed on the vibration signals in the original dataset to transform the vibration signals from the time domain to the frequency domain, obtaining the frequency domain feature vector. The specific formula is as follows: ; in, This is a frequency domain eigenvector, representing the intensity distribution of the vibration signal at various frequencies, where ω is the angular frequency. The signal value collected by the vibration sensor at time t. This is a complex exponential function used to decompose a signal into sine waves of different frequencies. The imaginary unit; Specifically, the frequency domain feature vector is moduloed and scaled to a preset pixel size, and then converted into a grayscale spectrum, where the pixel values ​​in the grayscale spectrum reflect the amplitude intensity of the vibration signal at the corresponding frequency. Furthermore, the grayscale spectrogram is input into a pre-trained convolutional neural network, such as... Figure 2 As shown, the convolutional neural network sequentially passes through an input layer, an alternating combination of convolutional and max-pooling layers, a fully connected layer, and a forward propagation operation of the output layer, outputting the probability of heavy fuel oil and the probability of light fuel oil. The fuel type identification result is based on a preset probability threshold. When the probability of heavy fuel oil is greater than or equal to the preset probability threshold, the fuel type is heavy fuel oil (HFO); otherwise, the fuel type is light fuel oil (MDO). It should be noted that the preset pixel size is 128×128 pixels, where the grayscale spectrum is stored in the form of a single-channel grayscale image; the preset probability threshold is based on the output probability properties of the Softmax function and Bayesian decision theory, so that the sum of the heavy oil probability and the light oil probability is 1. Using 0.5 as the dividing line can achieve the optimal decision of minimizing the classification error rate. When the heavy oil probability is greater than or equal to 0.5, the fuel type identification result is heavy oil (HFO); otherwise, it is light oil (MDO). Furthermore, the training of the convolutional neural network employs a cross-entropy loss function. The training samples include vibration spectrum maps of both heavy oil and light oil to ensure that the convolutional neural network's ability to classify fuel type from grayscale spectrum maps meets the preset accuracy requirements. The specific formula for the cross-entropy loss function is as follows: ; in, This represents the total number of training samples, for example, 8000 spectrograms. The sample number. The true labels for the samples are 1 for heavy oil and 0 for light oil. The value is between 0 and 1, representing the probability that this sample belongs to heavy oil. This is the cross-entropy loss value, and the training process aims to minimize this loss value. Specifically, for the missing fuel consumption data locations marked in step S1.3, the historical fuel consumption time series within a preset time period before the missing fuel consumption data locations and the real-time speed and load at the corresponding times are used as input sequences and input to the Long Short-Term Memory Network Model. The Long Short-Term Memory Network Model is subjected to gating operations through forget gate, input gate, candidate memory and output gate, and then outputs the completed fuel consumption value through a fully connected layer. The completed fuel consumption value is filled into the corresponding missing position of the original dataset to obtain the complete fuel consumption sequence. Preferably, the preset duration is 60 minutes; the gating operation of the Long Short-Term Memory network model includes four stages: forget gate, input gate, candidate memory, and output gate. The calculation formulas for each gating signal are as follows: in, This refers to the time step number, such as minute 1, minute 2, etc. This is the input vector at the current moment, containing information such as speed, load, and time features (e.g., hours). This is the short-term memory (hidden state) from the previous moment, with a dimension of 64. This represents long-term memory (cellular state) from the previous moment, with a dimension of 64. , , These are the gate control signals for the forget gate, input gate, and output gate, respectively. Their values ​​range from 0 to 1, and they are used to control the flow and retention of information. Candidate memories represent new information that you wish to remember at the current moment. For the updated long-term memory, It serves as the short-term memory for the current moment, and is also the output of the LSTM. , , , This is the weight matrix (trainable parameters) for each gating unit. , , , These are the bias terms (trainable parameters) for each gating unit. This is the Sigmoid function, with an output range of 0 to 1. It is a hyperbolic tangent function, with an output range of -1 to 1. This is the element-wise multiplication operator; Furthermore, the complete fuel consumption value is output by the fully connected layer using the following formula: ; in, To complete the fuel consumption value at time step t, This is the weight matrix of the fully connected layer, used to represent the hidden states output by the LSTM. A linear mapping to the dimension space of fuel consumption values. This is a bias term for fully connected layers, used to increase the model's fitting ability; S3: Calculate the speed deviation between the real-time speed and the speed declared by the crew, set a dynamic threshold based on historical navigation records, and determine the abnormal behavior and generate an abnormal marker when the speed deviation is greater than or equal to the dynamic threshold. Furthermore, the real-time speed is extracted from the original dataset, and the declared speeds submitted by crew members at the same time are read to calculate the speed deviation between the real-time speed and the declared speed. The speed deviation reflects the real-time speed. (Sourced from the ship's GNSS receiver, representing objective physical measurements, used as a benchmark) and declared speed The absolute difference between (manually filled out, as the objects to be inspected) is calculated using the following formula: ; Preferably, speed deviation samples of all normal voyages that were not judged as abnormal within a preset statistical period are extracted from historical voyage records, the mean and standard deviation of the speed deviation samples are calculated, and a dynamic threshold is constructed by a linear combination of the mean and standard deviation. It should be noted that the preset statistical period is the past 30 days, and the dynamic threshold is... The specific formula is as follows: ; in, This represents the mean of the speed deviation samples. The standard deviation of the speed deviation sample. It is a constant with a value of 2.5. The constant can be adjusted according to the historical deviation distribution characteristics of different ship types or different routes so that the dynamic threshold can be adapted to the anomaly detection sensitivity requirements under different application scenarios. Specifically, the speed deviation is compared with a dynamic threshold. When the speed deviation is greater than the dynamic threshold, it is determined that there is abnormal behavior at the current moment, the abnormal flag Aanomaly=1 is set, and the timestamp of the abnormal moment, real-time speed, reported speed and speed deviation are recorded to generate abnormal behavior monitoring results. When the speed deviation is less than or equal to the dynamic threshold, it is determined that the current moment is normal navigation, and the abnormal flag Aanomaly=0 is set. S4: Based on the fuel type identification results and range and load data, calculate the carbon intensity index value, and classify the carbon intensity index value according to the benchmark value specified by the IMO; Furthermore, the emission factor corresponding to the fuel type used in this voyage is determined based on the fuel type identification result: if the fuel type identification result is heavy fuel oil (HFO), then the emission factor corresponding to heavy fuel oil is used; if the fuel type identification result is light fuel oil (MDO), then the emission factor corresponding to light fuel oil is used. It should be noted that the emission factor for heavy oil HFO is 3.114 tons of CO2 / ton of oil, and the emission factor for light oil MDO is 3.206 tons of CO2 / ton of oil. Furthermore, based on the complete fuel consumption series, the fuel consumption for this voyage is cumulatively calculated according to fuel type, and combined with the determined emission factors, the CO2 emissions for this voyage are calculated. The specific formula is as follows: ; in, Let J represent the consumption of the j-th type of fuel, in tons. As emission factors, This represents the CO2 emissions for this voyage, expressed in grams. Specifically, the voyage distance and payload data are extracted from the original dataset, and combined with the CO2 emissions of this voyage, the carbon intensity index value of this voyage is calculated according to the carbon intensity index calculation formula specified by the IMO. The specific formula is as follows: ; in, This is the carbon intensity index value for this voyage, expressed in grams of CO2 per ton-nautical mile. Load capacity, in tons. This refers to the distance traveled, expressed in nautical miles. Furthermore, the benchmark values ​​stipulated by the IMO should be consulted based on the ship type and tonnage. The ratio of the carbon intensity index value to the benchmark value is calculated, and the carbon intensity index value is divided into five levels from A to E according to the preset level boundaries, resulting in the level classification results: ; Preferably, the preset grade boundaries and corresponding grade classification conditions are as follows: if the carbon intensity index value is less than 0.86 times the benchmark value, it is classified as Grade A; if the carbon intensity index value is greater than or equal to 0.86 times the benchmark value and less than 0.94 times the benchmark value, it is classified as Grade B; if the carbon intensity index value is greater than or equal to 0.94 times the benchmark value and less than 1.06 times the benchmark value, it is classified as Grade C; if the carbon intensity index value is greater than or equal to 1.06 times the benchmark value and less than 1.18 times the benchmark value, it is classified as Grade D; if the carbon intensity index value is greater than or equal to 1.18 times the benchmark value, it is classified as Grade E. The grade classification results will be used as one of the on-chain data for blockchain notarization in step S5, and will be read by the smart contract to trigger the green certificate issuance judgment and carbon credit calculation. It should be noted that in the rating results, Grade A indicates excellent carbon emission intensity for this voyage, meaning the carbon emission intensity for this voyage is more than 14% lower than the benchmark level stipulated by the IMO; Grade B indicates good carbon emission intensity for this voyage, meaning the carbon emission intensity for this voyage is 6% to 14% lower than the benchmark value; Grade C indicates acceptable carbon emission intensity for this voyage, meaning the carbon emission intensity for this voyage fluctuates within 6% above or below the benchmark value; Grade D indicates poor carbon emission intensity for this voyage, meaning the carbon emission intensity for this voyage is 6% to 18% higher than the benchmark value; Grade E indicates very poor carbon emission intensity for this voyage, meaning the carbon emission intensity for this voyage is more than 18% higher than the benchmark value. According to IMO regulations, vessels that are continuously rated as Grade D or E must develop a rectification plan, otherwise they will face penalties such as restrictions on navigation. Therefore, the rating results have a direct binding effect on the compliant operation of shipping companies. Furthermore, a feature vector composed of planned speed, load, predicted wave height, predicted wind speed, and predicted temperature is input to a pre-trained Long Short-Term Memory (LSTM) network prediction model. The LTM network prediction model outputs a probability distribution of CII levels for five future flight segments, from A to E. The level with the highest probability in the CII level probability distribution is used as the predicted CII level for the future flight segment. Specifically, this includes: The specific formula for the eigenvector is as follows: ; Where t is the time step number, i.e., the sampling point of the t-th flight segment. For feature vectors, Planned speed, in knots. Load capacity, in tons. For wave height forecasts, the unit is meters. Forecast wind speed, unit is meters per second. Temperature forecasts are in degrees Celsius. Output CII Rank Probability Distribution of Long Short-Term Memory Network Prediction Model The sum of all components in the CII rank probability distribution is 1. The training loss function of the Long Short-Term Memory network prediction model uses the mean squared error, calculated as follows: ; in, The training loss function value for the prediction model. This represents the total number of training samples, expressed in segments. For the actual observed CII level unique thermal encoding, The probability distribution of CII level output by the Long Short-Term Memory Network prediction model; It should be noted that the Long Short-Term Memory Network prediction model and the Long Short-Term Memory Network model used for fuel consumption data completion in step S2.4 are two independent models. The former takes feature vectors as input and CII level probability distribution as output, while the latter takes historical fuel consumption time series as input and complete fuel consumption value as output. The input features, output targets and training data of the two are different. In an optional embodiment, the shipowner can adjust the planned speed or optimize the load allocation scheme in advance based on the probability of predicting the CII level as A or B in the CII level probability distribution, so that the carbon intensity index value of the actual voyage falls into a better level range, thereby triggering the issuance of green certificates and obtaining higher carbon credits in step S5. S5: Upload the hash values ​​of vibration signals, fuel type identification results, anomaly markers, and grade classification results to the consortium blockchain for evidence storage, complete green certification through smart contracts, and issue corresponding carbon credits; Specifically, the fuel type identification results, abnormal behavior monitoring results, grade classification results, and vibration signal feature fingerprints (hash values ​​of grayscale spectrograms) are concatenated with their corresponding timestamps and node identifiers, and the hash values ​​of each data item are calculated. The specific formulas are as follows: ; in, For Unix timestamps This serves as the node identifier for the current edge AI gateway. This is a string representing the pixel matrix of a grayscale spectrum. The hash value of each of the above data items; It should be noted that the hash value length is fixed at 256 bits, and any change to the data content will cause the hash value to change. The pixel matrix string of the 128×128 grayscale spectrum image; Furthermore, the edge AI gateway calls the evidence storage smart contract deployed on the consortium blockchain through the Web3 interface, uploads the hash value H to the consortium blockchain, and retains the original data of vibration signal, fuel type identification result, abnormal behavior monitoring result and grade classification result in local or private IPFS, with only the hash value stored on the chain. Furthermore, the consortium blockchain is built on FISCO BCOS, with nodes corresponding to four types of participants: ship owners, ports, certification bodies, and regulatory agencies. The hash value H is written into the block and then jointly maintained by the nodes of the four types of participants. It should be noted that the design of storing only the hash value H on the chain, rather than the original data, ensures that the data is immutable and traceable while also taking into account data privacy protection and evidence storage efficiency. Specifically, such as Figure 3 As shown, after the voyage ends, the shipowner node submits the voyage end event to the consortium blockchain. The evidence storage smart contract automatically reads the rating results stored on the chain. If the rating result is A or B, the evidence storage smart contract automatically forges a green certificate in NFT format and stores the certificate hash value of the green certificate on the chain. If the rating result is C, D or E, the evidence storage smart contract does not issue a green certificate. Preferably, the metadata of the green certificate includes the voyage ID, the vessel's IMO number, the carbon intensity index value, and the issuance time; port nodes can query the on-chain green certificate information before the vessel arrives at the port and give priority berthing rights to vessels holding valid green certificates; auditing nodes can verify the consistency between the original data of any voyage and the on-chain evidence at any time through hash value. Furthermore, based on the grade classification results, the evidence storage smart contract calculates and issues carbon credits to the shipowner according to the preset calculation rules, and the carbon credit issuance record is simultaneously written to the consortium blockchain. Furthermore, the specific formula for the carbon integral is as follows: ; in, For carbon integral, The base score is 100 points by default. This is the reward coefficient, defaulting to 20. This is the carbon intensity index value for this voyage. The reference value specified by the IMO; It should be noted that the ratio of the carbon intensity index value to the benchmark value reflects the carbon emission level of this voyage relative to the benchmark. The lower the carbon intensity index value, the higher the carbon credit. Carbon credits and green certificates together form an incentive loop for the carbon emission performance of ships. The carbon credit issuance records stored on the consortium blockchain can be queried by port nodes to grant differentiated berthing rights to ships with higher carbon credits.

[0020] Specifically, based on the hash value H, green certificates, and carbon credit issuance records stored on the consortium blockchain, a visual query interface is provided to four types of participants—shipowners, ports, certification bodies, and regulatory agencies—to display their respective role permissions. Each participant can obtain on-chain data matching their role permissions through this visual query interface, including: Shipowners can use a visual query interface to check their ship's real-time carbon intensity index, historical carbon emission trends, list of voyages awaiting certification, number of green certificates, and carbon credit balance, such as... Figure 4 The image shown is a screenshot of the shipowner's homepage; certification bodies can query the voyage review list through a visual query interface, such as... Figure 5 The image shown is a core page diagram of the certification body; ports can query the green vessel berthing priority list through a visual query interface. The green vessel berthing priority list is sorted from high to low carbon credits, and vessels holding valid green certificates and with higher carbon credits receive priority berthing rights, such as... Figure 6 The image shown is a core page diagram of port management. Regulatory agencies can query carbon emission data by time range and vessel identifier through a visual query interface, and output statistical reports in the form of daily emission bar charts and CII level distribution charts, such as... Figure 7 The image shown is a screenshot of the core page for regulatory agencies' queries. When a certification body discovers an anomaly flag of Aanomaly=1 for a certain voyage through a visual query interface, it can retrieve a comparison trajectory map of the real-time speed and the declared speed for that voyage. The comparison trajectory map visually presents the abnormal behavior monitoring results by showing the degree of deviation between the AIS measured trajectory and the declared trajectory, which can be manually reviewed by the certification body. It should be noted that each record in the voyage audit list includes the vessel's IMO number, voyage time, fuel type identification result, anomaly flag (Aanomaly) for abnormal behavior monitoring results, carbon intensity index value, and audit operation entry point. In an optional embodiment, regulatory agencies can export carbon emission statistics reports in PDF format through a visual query interface. All data in the carbon emission statistics reports can be traced and verified through the hash value H stored on the blockchain, thereby providing ESG investors, green financing institutions, and third-party auditing institutions with carbon emission data backed by the blockchain.

[0021] Based on the teachings of the above embodiments, other aspects disclosed in the embodiments of the present invention also propose a trusted monitoring system for ship carbon emissions based on blockchain and AI, such as... Figure 8 As shown, it includes: The data acquisition module is used to collect vibration signals and fuel consumption data during engine operation, and synchronously acquire the ship's real-time speed, distance traveled, and load data via the AIS system. The fuel identification module is used to perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector. After converting the frequency domain feature vector into a grayscale spectrum, it is input into a convolutional neural network to obtain the fuel type identification result. The anomaly monitoring module is used to calculate the speed deviation between the real-time speed and the speed declared by the crew. It sets a dynamic threshold based on historical navigation records. When the speed deviation is greater than or equal to the dynamic threshold, it is judged as abnormal behavior and an anomaly mark is generated. The carbon emission calculation module calculates the carbon intensity index value based on the fuel type identification result, flight distance and load data, and classifies the carbon intensity index value into levels according to the benchmark value stipulated by the IMO. The blockchain evidence storage module is used to upload the hash values ​​of vibration signals, fuel type identification results, anomaly markers, and grade classification results to the consortium blockchain for evidence storage, and complete green certification and issue corresponding carbon credits through smart contracts.

[0022] In summary, this invention achieves spatiotemporal alignment and a unified benchmark for multi-source heterogeneous data by simultaneously acquiring engine vibration signals and speed, range, and load data from the AIS system. This provides a reliable and complete data source for subsequent analysis, avoiding errors caused by data silos or time deviations, and improving the overall data quality and consistency of carbon emission monitoring. By performing a fast Fourier transform on the vibration signals to obtain frequency domain feature vectors and converting them into grayscale spectrograms, which are then input into a convolutional neural network for automatic feature extraction and classification, non-intrusive real-time identification of fuel type is achieved. This accurately distinguishes between heavy and light fuel without the need for expensive fuel mass flow meters, providing a precise basis for carbon emission factor selection, significantly reducing hardware deployment costs and improving identification efficiency. Furthermore, by calculating the deviation between real-time speed and crew-reported speed and setting dynamic thresholds based on historical voyage records, this invention can adapt to the normal fluctuation range under different ship and navigation conditions, effectively avoiding false alarms or missed alarms caused by fixed thresholds. It enables sensitive detection and marking of abnormal behaviors such as tampering with declared speed, enhancing the credibility and anti-fraud capabilities of the carbon emission monitoring process. By integrating fuel type identification results, voyage distance, and load data to calculate carbon intensity index values, and classifying them into five levels (A to E) according to the benchmark values ​​stipulated by the IMO, it transforms complex carbon emission data into standardized performance levels, providing intuitive and comparable quantitative assessment results for the carbon emission efficiency of a single voyage. This facilitates ship owners and management agencies to quickly grasp emission levels and formulate targeted improvement measures. By calculating hash values ​​for vibration signals, fuel type identification results, abnormal markings, and level classification results and uploading them to the consortium blockchain for evidence storage, and using smart contracts to automatically complete green certification and carbon credit issuance, it ensures the immutability and traceability of the entire process from data collection, processing to evaluation. At the same time, the use of automated smart contracts eliminates human intervention and intermediate links, greatly improving certification transparency and issuance efficiency, and effectively incentivizing ship owners to proactively reduce carbon emissions.

[0023] This embodiment also provides a computer device suitable for a blockchain and AI-based trusted monitoring system for ship carbon emissions, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the blockchain and AI-based trusted monitoring system for ship carbon emissions as proposed in the above embodiment.

[0024] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0025] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the trusted monitoring system for ship carbon emissions based on blockchain and AI as proposed in the above embodiments.

[0026] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments. Example 2

[0027] Referring to Table 1, which is the second embodiment of the present invention, this embodiment provides a reliable monitoring method for ship carbon emissions based on blockchain and AI. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0028] Specifically, an ADXL345 vibration sensor is installed on the surface of the ship's main engine cylinder head, with a sampling frequency set to 1kHz, to capture subtle vibration signals generated by different fuel combustion during engine operation in real time. Simultaneously, an NVIDIA Jetson Xavier NX is deployed as an edge AI gateway, pre-installed with the Ubuntu 18.04 operating system and the TensorFlow 2.4 deep learning framework, and loaded with pre-trained CNN fuel identification models, LSTM data completion models, and carbon emission prediction models. Based on the FISCO BCOS 2.8.0 enterprise-grade blockchain platform, a consortium blockchain is built on four cloud servers (Alibaba Cloud ECS, configured with 2 cores and 4GB of RAM). The four nodes correspond to the shipowner company, the target port, a third-party certification body, and a maritime regulatory agency, respectively. A smart contract for automatic authentication and points distribution (written in Solidity 0.6.10, with a deployment gas consumption of approximately 1.2M) is deployed on the chain. All sensors, the AIS system, the edge gateway, and the blockchain nodes are connected via a secure network channel.

[0029] Furthermore, the vessel performs a typical voyage from Tianjin Port to Singapore Port. In step S1, the vibration sensor continuously collects data, generating a data packet every 10 seconds and uploading it to the edge gateway. Simultaneously, the real-time speed, distance traveled, and load data of the vessel are acquired through the AIS system. When the data stream enters step S2, the edge gateway first performs a Fast Fourier Transform on the vibration signal, converting the time-domain signal into a frequency-domain feature vector and generating a 128×128 pixel grayscale spectrum. This spectrum is input into a pre-trained CNN model. The model extracts deep features from the spectrum through alternating convolutional and pooling layers, and finally, the Softmax output layer outputs the fuel type probability. The experiment sets the probability of heavy oil as... The probability threshold is 0.5. When the model outputs a heavy fuel oil probability higher than this threshold, it is determined to be heavy fuel oil (HFO); otherwise, it is determined to be light fuel oil (MDO). In this voyage, the model identified the fuel type as light fuel oil (MDO) with a confidence level of 98.7%. If fuel consumption data is missing due to satellite signal fluctuations (such as during a 30-minute period), the LSTM data completion process in step S2 is executed. This LSTM model takes the historical fuel consumption, speed, and payload data of the 60 minutes before the missing time as the input sequence, updates the long-term memory and short-term memory through the gating mechanism of forget gate, input gate, and output gate, and finally outputs the completed fuel consumption value by the fully connected layer. The measured completion error is less than 3%, which effectively ensures the integrity of the fuel consumption sequence.

[0030] Further, in step S3, the system monitors abnormal behavior; it extracts the real-time speed (12 knots) recorded by the AIS and the crew's declared speed (9 knots) at the same time, calculating a speed deviation of 3 knots; based on the speed deviations of all normal voyages in the past 30 days, the system dynamically calculates a mean of 0.5 knots and a standard deviation of 0.6 knots, obtaining a dynamic threshold of 2.0 knots. Since the measured deviation of 3 knots is significantly greater than the dynamic threshold of 2.0 knots, the system automatically determines that there is a suspicion of fuel consumption fraud during this period and generates an anomaly marker (A_anomaly=1); subsequently, in step S4, based on the fuel type (MDO) and its corresponding emission factor (3.206 tons of CO2 / ton of fuel) identified in step S2, combined with the complete fuel consumption sequence, voyage distance (2,500 nautical miles), and load data, the system calculates the total CO2 emissions and CII value for this voyage; the calculated CII value is 18.68 g. CO2 / (t·nm); According to the table, the IMO benchmark value corresponding to this ship type is 20.0, and the ratio of the two is 0.934; Based on the preset grade boundaries (Grade A: <0.86, Grade B: 0.86-0.94, Grade C: 0.94-1.06), the system classifies this voyage as Grade B; In step S5, the edge gateway concatenates the vibration signal, fuel identification result, anomaly marker and grade classification result with the timestamp and node ID respectively, calculates their respective SHA-256 hash values, and calls the evidence storage smart contract through the Web3 interface to store these hash values ​​on the blockchain. After the voyage ends, the shipowner node submits a voyage completion event. The smart contract automatically reads the on-chain CII level (Level B), determines that it meets the A / B level standard, and then forges an NFT-formatted green certificate (certificate hash value example: 0x3f8e9a2b...). Based on the preset points rules (base score 100, reward coefficient 20), 101 carbon credits are calculated and issued. Before the ship arrives at the port, the port node queries the on-chain information and grants the ship priority berthing rights. The auditing agency can also verify the consistency between the on-chain evidence hash and the local original data at any time.

[0031] Specifically, as shown in Table 1, in terms of fuel type identification accuracy, the present invention improves by approximately 28.7 percentage points compared to manual reporting, and achieves a qualitative leap from "no automatic identification" to "high-precision identification." In terms of data tamper-proof capability, traditional centralized databases (methods A and B) are at risk of being tampered with by internal personnel or attacked by external hackers and cannot leave a trace. However, the blockchain hash evidence storage mechanism adopted by the present invention ensures that any data is immutable and fully traceable once it is on the chain, fundamentally establishing a trust foundation with the participation of multiple parties.

[0032] Table 1. Experimental Comparison of the Invention and Existing Technical Solutions Furthermore, and particularly crucially, the abnormal behavior detection rate is achieved. This invention, through dynamic threshold speed deviation detection, achieves a 100% detection rate (recall rate). This means that all implanted fuel consumption fraud behaviors are captured, while the detection rates of methods A and B are only 40% and 60%, respectively, with a large number of missed detections. This significant advantage directly breaks through common cheating methods such as "ghost navigation" and "speed fraud," providing regulatory agencies with a powerful technical tool. In terms of authentication efficiency, this invention, through smart contracts, shortens the authentication cycle from the traditional 7-15 days to minutes, saving up to 80% in labor costs and realizing an automated closed loop of "monitoring-auditing-authentication-incentive."

[0033] In summary, this invention is not a simple combination of existing technologies, but rather a systematic solution to the three major technical challenges of data authenticity, anti-fraud capabilities, and multi-party collaboration efficiency through the deep integration of AI and blockchain technologies. Its technical effects far exceed the sum of individual technical features, demonstrating significant creative progress and industrial applicability.

[0034] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A reliable monitoring method for ship carbon emissions based on blockchain and AI, characterized in that, include: S1: Collect vibration signals and fuel consumption data during engine operation, and synchronously obtain the ship's real-time speed, distance traveled, and load data via the AIS system; S2: Perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector, convert the frequency domain feature vector into a grayscale spectrum and input it into a convolutional neural network to obtain the fuel type identification result; S3: Calculate the speed deviation between the real-time speed and the speed declared by the crew, set a dynamic threshold based on historical navigation records, and determine the abnormal behavior and generate an abnormal marker when the speed deviation is greater than or equal to the dynamic threshold. S4: Based on the fuel type identification result and the range and load data, calculate the carbon intensity index value, and classify the carbon intensity index value into levels according to the benchmark value specified by the IMO; S5: Upload the vibration signal, the fuel type identification result, the anomaly marker, and the hash value of the grade classification result to the consortium blockchain for evidence storage, complete the green certification through the smart contract, and issue the corresponding carbon credits.

2. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 1, characterized in that, Step S5 specifically includes: The fuel type identification results, abnormal behavior monitoring results, grade classification results, and vibration signal feature fingerprints are concatenated with the corresponding timestamps and node identifiers, and the hash values ​​of each data item are calculated. Based on the grade classification results, the evidence storage smart contract calculates and issues carbon credits to the shipowner according to preset calculation rules, and the issuance record of the carbon credits is synchronously written to the consortium blockchain.

3. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 2, characterized in that, The hash value of each data item is calculated using the following formula: ； in, For Unix timestamps This serves as the node identifier for the current edge AI gateway. This is a string representing the pixel matrix of a grayscale spectrum. The hash value of each of the above data items; The specific formula for the carbon integral is as follows: ； in, For carbon integral, Basic integral, As the reward coefficient, This is the carbon intensity index value for this voyage. These are the benchmark values ​​specified by the IMO.

4. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 2, characterized in that, The ranking results include: The emission factor corresponding to the type of fuel used in this voyage is determined based on the fuel type identification results. Based on the complete fuel consumption series, the fuel consumption of this voyage is cumulatively calculated according to fuel type, and the CO2 emissions of this voyage are calculated in combination with the determined emission factors. Extract the voyage distance and payload data from the original dataset, combine them with the CO2 emissions of this voyage, and calculate the carbon intensity index value of this voyage according to the carbon intensity index calculation formula specified by the IMO. The reference values ​​stipulated by the IMO should be obtained based on the ship type and tonnage. The ratio of the carbon intensity index value to the benchmark value is calculated, and the carbon intensity index value is divided into five levels from A to E according to the preset level boundaries to obtain the level division result.

5. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 4, characterized in that, The historical fuel consumption time series within a preset time period before the missing fuel consumption data position, along with the real-time speed and load at the corresponding time, are used as the input sequence. This sequence is then input into a long short-term memory network model. After gating operations through forget gate, input gate, candidate memory, and output gate, the long short-term memory network model outputs a completed fuel consumption value through a fully connected layer. This completed fuel consumption value is then filled into the corresponding missing position in the original dataset to obtain a complete fuel consumption sequence.

6. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 5, characterized in that, The specific formula for completing the fuel consumption value is as follows: ； in, To complete the fuel consumption value at time step t, This is the weight matrix of the fully connected layer, used to represent the hidden states output by the LSTM. A linear mapping to the dimension space of fuel consumption values. This is a bias term for fully connected layers, used to increase the model's fitting ability.

7. The trusted monitoring system for ship carbon emissions based on blockchain and AI as described in claim 4, characterized in that, The specific formula for the CO2 emissions of this voyage is as follows: ； in, Let be the consumption of the j-th type of fuel. As emission factors, This represents the CO2 emissions for this voyage.

8. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 2, characterized in that, The fuel type identification result is based on a preset probability threshold. When the probability of heavy fuel oil is greater than or equal to the preset probability threshold, the fuel type is heavy fuel oil HFO; otherwise, the fuel type is light fuel oil MDO.

9. The reliable monitoring method for ship carbon emissions based on blockchain and AI as described in claim 8, characterized in that, The heavy oil probability includes: Perform a fast Fourier transform on the vibration signal to convert it from the time domain to the frequency domain, and obtain the frequency domain feature vector; The frequency domain feature vector is moduloed and scaled to a preset pixel size, then converted into a grayscale spectrum. The grayscale spectrum is input into a pre-trained convolutional neural network, which sequentially passes through an input layer, an alternating combination of convolutional and max-pooling layers, a fully connected layer, and a forward propagation operation of the output layer to output the probabilities of heavy oil and light oil.

10. A reliable monitoring system for ship carbon emissions based on blockchain and AI, based on the reliable monitoring method for ship carbon emissions based on blockchain and AI as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect vibration signals and fuel consumption data during engine operation, and synchronously acquire the ship's real-time speed, distance traveled, and load data via the AIS system. The fuel identification module is used to perform a fast Fourier transform on the vibration signal to obtain a frequency domain feature vector, convert the frequency domain feature vector into a grayscale spectrum, and then input it into a convolutional neural network to obtain the fuel type identification result. The anomaly monitoring module is used to calculate the speed deviation between the real-time speed and the speed declared by the crew, set a dynamic threshold based on historical navigation records, and determine the abnormal behavior and generate an anomaly mark when the speed deviation is greater than or equal to the dynamic threshold. The carbon emission calculation module calculates the carbon intensity index value based on the fuel type identification result and the range and load data, and classifies the carbon intensity index value into levels according to the benchmark value specified by the IMO. The blockchain evidence storage module is used to upload the hash values ​​of the vibration signal, the fuel type identification result, the anomaly marker, and the grade classification result to the consortium blockchain for evidence storage, and complete the green certification and issue the corresponding carbon credits through smart contracts.

Images