Cold-chain logistics transportation data block chain selective evidence storage method based on dynamic weight

Abstract

The invention relates to a cold-chain logistics transportation data block chain selective evidence storage method based on dynamic weight. The method comprises the following steps: acquiring multi-dimensional state feature data such as a temperature and humidity threshold deviation ratio, a geographic position displacement vector, an equipment power consumption fluctuation coefficient and a communication link quality index in a cold chain transportation process in real time, performing time sequence feature extraction and dynamic association analysis on the multi-dimensional state feature data, extracting key features, and performing enhancement and correction to generate an optimized evidence storage weight value; and comparing the evidence storage weight value based on a preset threshold value, allocating the block chain storage resources and generating an encrypted evidence storage path, and dynamically adjusting the block chain fragment resource allocation proportion and monitoring the idle condition in combination with the spatial and temporal distribution characteristics and the service fluctuation trend of the encrypted evidence storage path to generate a resource scheduling scheme. According to the method, selective uplink of the cold-chain transportation data is realized, the utilization efficiency of block chain fragment resources is optimized and improved, the method is suitable for complex and changeable cold-chain transportation scenes, and an efficient and economic solution is provided for cold-chain logistics data management.

Description

Technical Field

[0001] This invention belongs to the field of cold chain logistics data processing, and in particular relates to a selective blockchain evidence storage method for cold chain logistics transportation data based on dynamic weights. Background Technology

[0002] In cold chain logistics operations, the accurate recording and secure storage of transportation data are crucial for ensuring the quality and safety of goods. Traditional cold chain logistics data largely relies on centralized database storage, such as some companies uploading data to the cloud storage of third-party cold chain logistics monitoring platforms (e.g., YiLiu, JieYi platform). However, this model has serious drawbacks; data is easily tampered with. For example, some logistics companies can have third-party platforms modify temperature records when transportation temperatures are not up to standard, causing the data to lose credibility and making it difficult to provide effective data support to cargo owners and regulatory authorities. To solve the data credibility problem, blockchain technology has been introduced into the field of cold chain logistics data management. For instance, patent CN201910888451.3 proposes uploading logistics information to blockchain storage, leveraging the immutability of blockchain data to determine the compliance of goods transportation. However, this method does not consider the high cost of blockchain storage. In practical applications, if all the massive amounts of data generated by cold chain logistics were uploaded to the blockchain, storage and maintenance costs would increase dramatically. Meanwhile, some companies have attempted to put only critical data on the blockchain, but the lack of a dynamic evaluation mechanism for determining critical data makes it difficult to adapt to the complex and ever-changing cold chain transportation scenarios, as different types of goods and transportation stages have varying requirements for data importance. Furthermore, in terms of resource scheduling, existing solutions have failed to fully integrate the spatiotemporal distribution of encrypted evidence storage paths and business fluctuation trends for dynamic optimization, resulting in unreasonable allocation of blockchain sharding resources and the occurrence of resource idleness or overload. Summary of the Invention

[0003] Therefore, it is necessary to provide a selective blockchain evidence storage method for cold chain logistics transportation data based on dynamic weights, which can utilize blockchain to ensure the immutability and traceability of key data and improve the utilization efficiency of blockchain shard resources, in order to address the above-mentioned technical problems.

[0004] Firstly, this application provides a method for selective blockchain-based notarization of cold chain logistics transportation data based on dynamic weights, including:

[0005] Acquire multi-dimensional status feature data of cold chain transportation collected in real time by edge computing nodes; the multi-dimensional status feature data includes temperature and humidity threshold deviation rate, geographical location displacement vector, equipment power consumption fluctuation coefficient and communication link quality index.

[0006] We perform time-series feature extraction and dynamic correlation analysis on multidimensional state feature data, extract key features for enhancement and correction, and generate optimized evidence preservation weight values.

[0007] The evidence storage weight values ​​are compared based on a preset threshold, and blockchain storage resources are allocated and encrypted evidence storage paths are generated based on the comparison results.

[0008] By combining the spatiotemporal distribution characteristics of encrypted evidence storage paths with business fluctuation trends, the resource allocation ratio of blockchain sharding is dynamically adjusted and idle status is monitored to generate a resource scheduling scheme.

[0009] In one embodiment, temporal feature extraction and dynamic correlation analysis are performed on multidimensional state feature data, key features are extracted, enhanced, and corrected, and optimized evidence storage weight values ​​are generated, including:

[0010] Based on multidimensional state feature data, the combined weights of each dimension feature are calculated using an improved analytic hierarchy process to generate an initial weight vector.

[0011] The initial weight vector is input into the pre-trained Long Short-Term Memory (LSTM) network to extract temporal feature dependencies and generate a temporally optimized adjusted weight matrix. The parameters of the LSM network are obtained through pre-training with historical cold chain transportation data.

[0012] The attention mechanism enhances the graph neural network fusion adjustment of the temporal correlation between the weight matrix and features to generate the evidence storage weight tensor; the evidence storage weight tensor quantifies the evidence storage priority of features of each dimension in different spatiotemporal scenarios.

[0013] If the component values ​​of the evidence storage weight tensor are lower than the adaptive threshold, wavelet transform and feature reconstruction are performed on the corresponding dimension features, and an optimized evidence storage weight value is generated by combining the historical anomaly contribution correction coefficient.

[0014] In one embodiment, the temporal correlation between features is calculated using the following formula, including:

[0015]

[0016] Where, r ij (t) represents the temporal correlation between feature i and feature j at time t, ω i (tk) represents the weight value of feature i at time tk, ω i (t) represents the weight value of feature i at time t, obtained after time-series optimization using a Long Short-Term Memory (LSTM) network, h t-1 W represents the hidden state of the LSTM at time t-1. l U l V l b l Let ω represent the learnable parameters of the l-th layer of the LSTM, (·) represent taking the i-th component of vector i, and ω 初始，i Represents the initial weight vector. This represents the average weight of feature i within the time window T, where T represents the size of the sliding time window dynamically set based on the feature's fluctuation period.

[0017] In one embodiment, the evidence preservation weight value is obtained through the following steps:

[0018] Obtain the component value data of the evidence storage weight tensor; the component value data is used for anomaly detection in the feature dimension.

[0019] Adaptive threshold filtering is applied to the component value data. If the value is below the threshold, the corresponding dimension feature is marked as an anomaly, thus obtaining the anomaly feature.

[0020] Wavelet transform is applied to the anomalous features to obtain the feature data after time-frequency domain decomposition.

[0021] Based on the feature data, feature reconstruction is performed to generate a denoised reconstructed feature matrix.

[0022] The reconstructed feature matrix is ​​combined with the historical anomaly contribution correction coefficient to obtain the optimized evidence preservation weight value.

[0023] The optimized weight values ​​are subjected to distribution consistency verification. If there is a deviation, a support vector machine is used to build a data storage model for classification optimization to obtain the weight distribution data.

[0024] The parameters of the evidence preservation model are updated based on the weight distribution data to determine the final evidence preservation weights.

[0025] In one embodiment, the historical anomaly contribution correction coefficient is calculated using the following formula:

[0026]

[0027] β i =ρ·f i +(1-ρ)e j

[0028] Where, β i f represents the historical anomaly contribution correction coefficient for feature i. i c represents the proportion of abnormal frequencies of feature i. i Let i represent the total number of times feature i appears anomaly in historical data, n represent the total number of features, and e represent the total number of features. j I represents the degree of anomalous influence of feature i. i,t This represents an indicator function, which takes a value of 1 when feature i experiences an anomaly at time t, and a value of 0 otherwise, |Δy t | represents the system output deviation at time t, T represents the total statistical time length, and ρ represents the balance coefficient.

[0029] In one embodiment, the evidence storage weight value is compared based on a preset threshold, and blockchain storage resources are allocated and an encrypted evidence storage path is generated according to the comparison result, including:

[0030] The evidence storage weight value is compared with the preset threshold in a hierarchical manner to obtain the comparison result containing the weight interval identifier.

[0031] Dynamic allocation of blockchain storage resources is performed based on the comparison results. If data falls into the high-weight range, it is marked as high-priority data.

[0032] High-priority data is matched with a multi-node redundant storage resource pool that meets its attribute requirements.

[0033] An encrypted evidence storage path is generated based on the attribute requirements of the storage resource pool and high-priority data; the encrypted evidence storage path includes the node access sequence and data shard index of the corresponding storage resource pool.

[0034] In one embodiment, by combining the spatiotemporal distribution characteristics of the encrypted evidence storage path with business fluctuation trends, the resource allocation ratio of blockchain shards is dynamically adjusted and idle status is monitored to generate a resource scheduling scheme, including:

[0035] Obtain the spatiotemporal distribution characteristics and business fluctuation trends of the encrypted evidence storage path; the spatiotemporal distribution characteristics include the geographical distribution of path-related nodes and the access frequency during the time period.

[0036] By conducting a correlation analysis between the spatiotemporal distribution characteristics and business fluctuation trends, the matching relationship between resource demand and business fluctuations can be obtained.

[0037] The allocation ratio of blockchain shard resources is dynamically adjusted based on the matching relationship to obtain the resource allocation adjustment result.

[0038] The idle status of resources in each segment is monitored in real time to obtain idle monitoring data.

[0039] A resource scheduling plan is generated by combining the resource allocation adjustment results with idle monitoring data; the resource scheduling plan includes dynamic expansion and contraction strategies for fragmented resources.

[0040] In one embodiment, the method further includes:

[0041] The execution effect of the optimal resource scheduling scheme is quantitatively evaluated to obtain quantitative evaluation indicators and results. The quantitative evaluation indicators include blockchain storage resource utilization, data storage response latency, abnormal data tracing accuracy, and shard resource load balancing.

[0042] By using edge computing nodes to monitor the dynamic changes in the evidence storage weight values ​​in real time, and combining the evaluation results to adjust the parameters of the evidence storage model, an optimized evidence storage model is obtained.

[0043] The optimized evidence storage model will be synchronized to the edge nodes of the entire cold chain transportation chain. Based on the evidence storage weight value of each node and the actual storage path, storage cost data including the proportion of blockchain and database storage will be generated.

[0044] Secondly, this application also provides 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 the method described above.

[0045] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method.

[0046] The aforementioned method for selective blockchain-based evidence storage of cold chain logistics transportation data, along with the computer equipment and storage media, utilizes edge computing nodes to collect multi-dimensional state characteristic data in real time during cold chain transportation, including temperature and humidity threshold deviation rates, geographical location displacement vectors, equipment power consumption fluctuation coefficients, and communication link quality indices. Temporal feature extraction and dynamic correlation analysis are performed on these data to extract key features and enhance and correct them to generate optimized evidence storage weight values. Based on preset thresholds, the evidence storage weight values ​​are compared, and blockchain storage resources are allocated accordingly to generate encrypted evidence storage paths. Furthermore, by combining the spatiotemporal distribution characteristics of the encrypted evidence storage paths with business fluctuation trends, the allocation ratio of blockchain shard resources is dynamically adjusted, and idle status is monitored to generate resource scheduling schemes. This method achieves selective on-chain storage of cold chain transportation data through dynamic weight evaluation. It leverages blockchain to ensure the immutability and traceability of critical data while controlling blockchain storage costs through differentiated storage of non-critical data. Simultaneously, resource scheduling optimization improves the utilization efficiency of blockchain shard resources, making it suitable for complex and ever-changing cold chain transportation scenarios and providing an efficient and economical solution for cold chain logistics data management. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart of a blockchain selective evidence storage method for cold chain logistics transportation data based on dynamic weights, provided in an embodiment of the present invention;

[0049] Figure 2 This is a flowchart provided by an embodiment of the present invention for extracting time-series features and performing dynamic correlation analysis on multi-dimensional state feature data, extracting key features for enhancement and correction, and generating optimized evidence storage weight values. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] In one embodiment, such as Figure 1 As shown, this application provides a method for selective blockchain storage of cold chain logistics transportation data based on dynamic weights, which may include the following steps:

[0052] Step S101: Obtain multi-dimensional status feature data of cold chain transportation collected in real time by the edge computing node; the multi-dimensional status feature data includes temperature and humidity threshold deviation rate, geographical location displacement vector, equipment power consumption fluctuation coefficient and communication link quality index.

[0053] Specifically, edge computing nodes are deployed on cold chain transportation vehicles (such as refrigerated trucks) or monitoring points along the route. They collect four types of core data in real time via sensors: temperature and humidity threshold deviation rate (reflecting the degree of deviation between actual temperature and humidity and preset thresholds), geographic location displacement vector (including real-time latitude and longitude and direction of movement speed), device power consumption fluctuation coefficient (reflecting the energy consumption stability of refrigeration equipment, etc.), and communication link quality index (quantifying the signal strength and packet loss rate of data transmission). This data is then used to generate structured data packets at fixed time intervals (e.g., every 30 seconds).

[0054] Step S102: Perform time-series feature extraction and dynamic correlation analysis on the multidimensional state feature data, extract key features for enhancement and correction, and generate optimized evidence storage weight values.

[0055] Using preprocessed multidimensional state feature data as input, time-series feature extraction algorithms (such as sliding window statistics) are employed to capture the changing trends of each dimension of data over time (such as the continuous expansion or contraction of temperature deviation). Simultaneously, dynamic correlation analysis models (such as cross-feature matrices) are used to uncover the intrinsic relationships between different dimensions of data (such as the correlation between sudden increases in equipment power consumption and rising temperature deviation rates). For the key features identified (such as significant fluctuations affecting cargo quality), feature enhancement techniques (such as weighted amplification of abnormal fluctuations) and correction mechanisms (such as eliminating false alarms from sensors) are employed. Finally, a weighted calculation model generates the evidence weight value for each data item, quantifying its importance in cold chain traceability.

[0056] Step S103: Compare the evidence storage weight values ​​based on a preset threshold, allocate blockchain storage resources according to the comparison results, and generate an encrypted evidence storage path.

[0057] The generated evidence storage weight values ​​are compared with the system's preset multi-level thresholds (such as high, medium, and low levels) to determine the weight range to which each data item belongs (e.g., values ​​above 80 are in the high-weight range). Based on the comparison results, differentiated resource allocation is performed: high-weight data is directly allocated blockchain storage resources to ensure its immutability; medium- and low-weight data are allocated to traditional databases. For on-chain data, the corresponding storage resource configuration is further matched according to its weight level (e.g., high-weight data is allocated multi-node redundant storage), and an encrypted evidence storage path containing node access permissions and data shard location indexes is generated to ensure the security and traceability of data access.

[0058] Step S104: Combining the spatiotemporal distribution characteristics of the encrypted evidence storage path with business fluctuation trends, dynamically adjust the blockchain sharding resource allocation ratio and monitor idle status to generate a resource scheduling scheme.

[0059] Indicatively, based on the encrypted evidence storage path, the spatiotemporal distribution characteristics of data storage (such as the storage load of nodes in a certain region and the access frequency during a specific period) are parsed. Simultaneously, combined with the fluctuating trends of cold chain operations (such as peak holiday transportation periods and changes in cargo volume on specific routes), a resource scheduling model dynamically adjusts the resource allocation ratio of each blockchain shard (e.g., increasing the storage capacity of high-access shards). The resource idle rate of each shard (e.g., nodes with storage space utilization below 30%) is monitored in real time, and idle resources are reallocated to shards with higher loads. Finally, a scheduling scheme including shard expansion / contraction thresholds and resource migration paths is generated to achieve efficient utilization of blockchain storage resources.

[0060] The aforementioned method for selective blockchain-based evidence storage of cold chain logistics transportation data, based on dynamic weights, collects multi-dimensional state characteristic data in real time from edge computing nodes. This data includes temperature and humidity threshold deviation rates, geographical location displacement vectors, equipment power consumption fluctuation coefficients, and communication link quality indices during the cold chain transportation process. Temporal feature extraction and dynamic correlation analysis are performed on these data to extract key features and enhance and correct them, generating optimized evidence storage weight values. Based on preset thresholds, the evidence storage weight values ​​are compared, and blockchain storage resources are allocated accordingly, generating encrypted evidence storage paths. Furthermore, the spatiotemporal distribution characteristics of the encrypted evidence storage paths and business fluctuation trends are combined to dynamically adjust the allocation ratio of blockchain shard resources and monitor idle status to generate resource scheduling schemes. This method achieves selective on-chain storage of cold chain transportation data through dynamic weight evaluation. It utilizes blockchain to ensure the immutability and traceability of critical data while controlling blockchain storage costs through differentiated storage of non-critical data. Simultaneously, resource scheduling optimization improves the utilization efficiency of blockchain shard resources. This method is suitable for complex and ever-changing cold chain transportation scenarios, providing an efficient and economical solution for cold chain logistics data management.

[0061] In one embodiment, such as Figure 2As shown, performing time-series feature extraction and dynamic correlation analysis on multidimensional state feature data, extracting key features for enhancement and correction, and generating optimized evidence preservation weight values ​​may include the following steps:

[0062] Step S201: Based on the multidimensional state feature data, the combined weights of each dimension feature are calculated using the improved hierarchical analysis method to generate an initial weight vector.

[0063] Preferably, the weight calculation method optimized based on the traditional analytic hierarchy process (AHP) constructs a judgment matrix of multi-dimensional features and introduces dynamic correction factors (such as the influence coefficient of features on cargo quality) to reduce subjective assignment bias and more accurately calculate the combined weights of features of various dimensions such as temperature and humidity threshold deviation rate and geographical location displacement vector, providing a quantitative basis for the initial weight vector.

[0064] Step S202: Input the initial weight vector into the pre-trained long short-term memory network, extract the temporal feature dependencies, and generate the temporally optimized adjusted weight matrix; the parameters of the long short-term memory network are obtained by pre-training with historical cold chain transportation data.

[0065] Furthermore, the initial weight vector is optimized by LSTM to generate an adjusted weight matrix. Compared with the initial weight vector, it adds time dimension information, which can reflect the weight difference of the same feature at different times (such as the weight of temperature deviation rate at the end of transportation is higher than that at the beginning), which is more in line with the dynamic characteristics of cold chain transportation.

[0066] Step S203: The temporal correlation between the weight matrix and features is adjusted by fusion of graph neural networks enhanced by attention mechanism to generate evidence storage weight tensor; evidence storage weight tensor quantifies the evidence storage priority of features of each dimension in different spatiotemporal scenarios.

[0067] The model introduces an attention mechanism on the basis of graph neural networks. The graph neural network is used to model the correlation between features in various dimensions (such as the correlation between power consumption fluctuation and temperature deviation), while the attention mechanism assigns higher weights to feature pairs with high correlation (such as prioritizing the correlation between temperature and power consumption of cooling equipment). Finally, the weight matrix and the temporal correlation between features are integrated and adjusted to generate a storage weight tensor.

[0068] The evidence preservation weight tensor contains tensor data with three dimensions of information: space (multi-dimensional features), time (different moments), and priority (importance of evidence preservation). It can quantify the evidence preservation priority of each dimension of features in different spatiotemporal scenarios (such as the evidence preservation priority of temperature data in high-temperature areas during transportation).

[0069] Step S204: If the component values ​​of the evidence storage weight tensor are lower than the adaptive threshold, then perform wavelet transform and feature reconstruction on the corresponding dimension features, and generate optimized evidence storage weight values ​​by combining the historical anomaly contribution correction coefficient.

[0070] The adaptive threshold is a weighted judgment value dynamically adjusted according to the cold chain transportation scenario. It is jointly determined by the distribution of historical abnormal data (such as the frequency of temperature exceeding the standard) and real-time business requirements (such as the freshness level of the goods). For example, the threshold is higher when transporting vaccines than when transporting ordinary goods to ensure that critical data is not misjudged as low priority.

[0071] The historical anomaly contribution correction coefficient is a dynamic adjustment factor calculated based on the historical anomaly records of features. It comprehensively considers the frequency of feature anomalies (such as the proportion of historical anomalies in temperature deviation) and the degree of impact of anomalies on evidence weight (such as the impact of temperature exceeding the standard on the quality of goods). It is used to correct the evidence weight value to ensure that the weight assessment of anomaly features is more in line with the actual impact.

[0072] Specifically, firstly, the importance of each dimension of features is quantified using an improved analytic hierarchy process (AHP), and combined weights are calculated to generate an initial weight vector. This initial weight vector is then input into a long short-term memory (LSTM) network pre-trained with historical cold chain transportation data. Leveraging the network's ability to capture temporal feature dependencies, a temporally optimized adjusted weight matrix is ​​output. Subsequently, an attention-enhanced graph neural network is used to fuse the adjusted weight matrix with the temporal correlation between features (e.g., the correlation between temperature and humidity deviations at different times and equipment power consumption), generating a storage weight tensor that quantifies the storage priority of each dimension of features in different spatiotemporal scenarios. If a component value in the storage weight tensor is lower than an adaptive threshold, wavelet transform is performed on the corresponding dimension feature to decompose the time-frequency domain information. After feature reconstruction to remove noise, and combined with a historical anomaly contribution correction coefficient (calculated based on the frequency and impact of feature anomalies), an optimized storage weight value is finally generated.

[0073] This embodiment achieves a reasonable allocation of initial weights through the analytic hierarchy process (AHP), uses a long short-term memory network to capture temporal dependencies to optimize the temporal dynamics of the weights, utilizes a graph neural network enhanced by an attention mechanism to strengthen the influence of inter-feature correlations on the weights, and processes low-weight anomalous features through wavelet transform and anomaly correction mechanisms. The final generated evidence preservation weight values ​​can accurately reflect the evidence preservation value of each data in different scenarios, providing a scientific basis for subsequent selective evidence preservation. This ensures the priority of key data and improves the adaptability of weight evaluation to complex cold chain scenarios.

[0074] In one embodiment, the temporal correlation between features is calculated using the following formula, including:

[0075]

[0076] Where, r ij (t) represents the temporal correlation between feature i and feature j at time t, ω i (tk) represents the weight value of feature i at time tk, ωi (t) represents the weight value of feature i at time t, obtained after time-series optimization using a Long Short-Term Memory (LSTM) network, h t-1 W represents the hidden state of the LSTM at time t-1. l U l V l b l Let ω represent the learnable parameters of the l-th layer of the LSTM, (·) represent taking the i-th component of vector i, and ω 初始，i Represents the initial weight vector. This represents the average weight of feature i within the time window T, where T represents the size of the sliding time window dynamically set based on the feature's fluctuation period.

[0077] Preferably, Among them, a ij H represents the importance score of feature i relative to feature j in the improved analytic hierarchy process (AHP) judgment matrix. i Let θ represent the information entropy of feature i, and θ∈[0,1] represent the fusion coefficient, balancing subjective judgment and objective data.

[0078] This embodiment, by introducing optimized feature weights, hidden states, and a dynamic sliding time window using a Long Short-Term Memory (LSTM) network, can accurately capture the strength of correlations between features at different times. Combining the initial weight vector with the average weight of features within the time window, it preserves the basic importance information of features while adapting to historical data patterns through learnable parameters. This allows the correlation calculation to simultaneously consider temporal dynamics and the intrinsic correlation of features, providing a reliable quantitative basis for the subsequent generation of evidence-based weight tensors and improving the accuracy of multi-dimensional feature fusion.

[0079] In one embodiment, the evidence preservation weight value can be obtained through the following steps:

[0080] Step S301: Obtain the component value data of the evidence storage weight tensor; the component value data is used for anomaly detection in the feature dimension.

[0081] Preferably, the specific values ​​of the evidence storage weight tensor in each dimension, each component value corresponds to the weight parameter of a certain feature in a specific spatiotemporal scenario (such as the weight value of the temperature and humidity threshold deviation rate at time t), which is the basic data for judging whether the feature is abnormal and directly reflects the evidence storage priority of the feature in that dimension.

[0082] Step S302: Perform adaptive threshold filtering on the component value data. If the value is below the threshold, mark the corresponding dimension feature as an anomaly to obtain the anomaly feature.

[0083] Adaptive threshold filtering is a filtering mechanism that dynamically adjusts the threshold based on the historical data distribution of features (such as the frequency of anomalies) and real-time scenario requirements (such as the freshness grade of goods). By comparing component value data with the threshold, it automatically filters out abnormal features below the threshold, avoiding the inadequacy of fixed thresholds for complex scenarios.

[0084] Step S303: Perform wavelet transform processing on the abnormal features to obtain the feature data after time-frequency domain decomposition.

[0085] Step S304: Based on the feature data, perform feature reconstruction to generate a denoised reconstructed feature matrix.

[0086] Step S305: The reconstructed feature matrix and the historical anomaly contribution correction coefficient are fused together to obtain the optimized evidence storage weight value.

[0087] Step S306: Perform a distribution consistency check on the optimized weight values. If there is a deviation, use a support vector machine to build a storage model for classification optimization to obtain weight distribution data.

[0088] Furthermore, the distribution consistency verification is a process of statistically testing the optimized weight values. By comparing the actual weight distribution with the theoretical distribution (such as normal distribution or preset threshold interval distribution), it is determined whether there is a deviation in the weight values ​​(such as excessive concentration or excessive dispersion), thus ensuring the rationality of the weight distribution.

[0089] When there is a bias in the distribution of weight values, a classification model is constructed using the support vector machine algorithm. By finding the optimal classification hyperplane, the weight values ​​are divided into reasonable intervals (such as high, medium, and low weight intervals), thereby optimizing the classification of the weight distribution and obtaining weight distribution data that meets the needs of the scenario.

[0090] Weight distribution data reflects the distribution of each feature's weight values ​​in different intervals (such as the proportion of high weight values, the standard deviation of weight values, etc.), and is used to quantify the rationality of the weight distribution.

[0091] Step S307: Update the parameters of the evidence preservation model based on the weight distribution data to determine the final evidence preservation weight.

[0092] Starting with the component value data of the evidence storage weight tensor, anomaly detection in the feature dimensions is first performed using the component value data. The component value data is then filtered using an adaptive threshold, marking features in the corresponding dimensions below the threshold as anomalies, thus obtaining anomalous features. Wavelet transform processing is performed on the anomalous features to decompose them into time-frequency domain feature data. Based on this data, feature reconstruction is performed to generate a denoised reconstructed feature matrix. The reconstructed feature matrix is ​​then fused with historical anomaly contribution correction coefficients to obtain optimized evidence storage weight values. The optimized weight values ​​are then subjected to distribution consistency verification. If deviations exist, a support vector machine is used to construct an evidence storage model for classification optimization, obtaining weight distribution data. Finally, the evidence storage model parameters are updated based on the weight distribution data to determine the final evidence storage weights.

[0093] This embodiment accurately identifies abnormal features through adaptive thresholding, effectively reduces noise interference by combining wavelet transform and feature reconstruction, enhances the rationality of weight values ​​by incorporating historical anomaly contribution correction coefficients, and ensures the reliability of weight distribution through distribution consistency verification and support vector machine optimization. The final determined evidence storage weights can more accurately reflect the actual importance of each dimension of features in cold chain traceability, providing a scientific and reliable basis for subsequent selective evidence storage, and effectively improving the targeting and accuracy of data evidence storage.

[0094] In one embodiment, the historical anomaly contribution correction coefficient can be calculated using the following formula:

[0095]

[0096] β i =ρ·f i +(1-ρ)e j

[0097] Where, β i f represents the historical anomaly contribution correction coefficient for feature i. i c represents the proportion of abnormal frequencies of feature i. i Let i represent the total number of times feature i appears anomaly in historical data, n represent the total number of features, and e represent the total number of features. j I represents the degree of anomalous influence of feature i. i,t This represents an indicator function, which takes a value of 1 when feature i experiences an anomaly at time t, and a value of 0 otherwise, |Δy t | represents the system output deviation at time t, T represents the total statistical time length, and ρ represents the balance coefficient.

[0098] This embodiment quantifies the probability of historical anomalies by the proportion of anomaly frequency, reflects the actual impact of anomalies on system output by combining anomaly impact degree, and adjusts the weights of both with a balancing coefficient to form a dynamic correction factor. It preserves the historical statistical patterns of feature anomalies while also taking into account the real-time impact of anomalies on the system, enabling the correction factor to accurately adapt to the actual anomaly characteristics of features in cold chain transportation. This provides a scientific basis for optimizing evidence weight values ​​and effectively improves the accuracy and rationality of anomaly feature weight assessment.

[0099] In one embodiment, comparing the evidence storage weight values ​​based on a preset threshold, allocating blockchain storage resources according to the comparison results, and generating an encrypted evidence storage path may include the following steps:

[0100] Step S401: The evidence storage weight value is compared with the preset threshold in a hierarchical manner to obtain the comparison result containing the weight interval identifier.

[0101] Preferably, the preset threshold is a pre-set multi-level weight judgment standard (such as high, medium and low three-level thresholds), which is determined according to the cold chain transportation scenario (such as cargo type, transportation stage) and business needs (such as regulatory requirements, storage costs), and is used to divide the data storage priority range (such as the high weight range above 80 points).

[0102] Step S402: Based on the comparison results, perform dynamic allocation of blockchain storage resources. If the data falls into the high-weight range, it is marked as high-priority data.

[0103] Based on the hierarchical comparison results, the allocation strategy of blockchain storage resources is flexibly adjusted: more storage resources are allocated to high-weight data (such as multi-node redundant storage), and basic storage resources are allocated to medium and low-weight data or transferred to the database, so as to realize the on-demand allocation of resources and reduce the overall storage cost.

[0104] Step S403: Match high-priority data with a multi-node redundant storage resource pool that meets its attribute requirements.

[0105] A multi-node redundant storage resource pool is a storage cluster composed of multiple blockchain nodes. It provides redundant storage for high-priority data (the same data is backed up on multiple nodes) to avoid data loss due to single point of failure. Its configuration needs to match the attribute requirements of the data (such as more nodes to back up data with high security level).

[0106] Step S404: Generate an encrypted evidence storage path based on the attribute requirements of the storage resource pool and high-priority data; the encrypted evidence storage path includes the node access sequence and data shard index of the corresponding storage resource pool.

[0107] The encrypted evidence storage path is path information that encrypts the storage location and access method of high-priority data. It contains two core parts: node access sequence (the order of nodes that need to be verified when reading data to ensure access permissions) and data shard index (the specific storage location of each shard in the nodes after data splitting). It not only ensures data security, but also supports fast location and traceability.

[0108] First, the data is compared with preset multi-level thresholds to determine the weight range (e.g., high, medium, and low levels) to which each data item belongs, generating a comparison result containing weight range identifiers. Based on this result, dynamic allocation of blockchain storage resources is performed. Data falling into the high-weight range is marked as high-priority data. For high-priority data, a multi-node redundant storage resource pool that meets its attribute requirements (e.g., security level, access frequency) is matched to ensure the reliability of data storage. Finally, based on the node distribution of the storage resource pool and the attribute requirements of the high-priority data, an encrypted evidence path containing the node access sequence (the node order in which data is read) and the data shard index (the storage location of each shard in the node) is generated to ensure the security and traceability of data access.

[0109] This embodiment achieves differentiated classification of data storage through hierarchical threshold comparison. High-priority data is matched with a multi-node redundant storage resource pool to ensure the reliability and immutability of critical data. The encrypted evidence storage path achieves secure data storage and efficient traceability through node access sequences and shard indexes. The overall process reduces storage costs by selectively allocating blockchain resources and ensures the security and accessibility of high-priority data through structured resource matching and path design, providing an efficient and reliable solution for hierarchical evidence storage of cold chain logistics data.

[0110] In one embodiment, by combining the spatiotemporal distribution characteristics of the encrypted evidence storage path with business fluctuation trends, the resource allocation ratio of blockchain shards is dynamically adjusted and idle status is monitored to generate a resource scheduling scheme, which may include the following steps:

[0111] Step S501: Obtain the spatiotemporal distribution characteristics and business fluctuation trends of the encrypted evidence storage path; the spatiotemporal distribution characteristics include the geographical distribution and time period access frequency of the path-related nodes.

[0112] Preferably, the spatiotemporal distribution characteristics of the encrypted evidence storage path include geographical distribution: the geographical location information of the path-related nodes (such as the city and region where the nodes are located), reflecting the physical location distribution of data storage and access; and time period access frequency: the number of times the path is accessed or the data flow within a specific time period (such as the morning of a weekday or the night of a holiday), reflecting the temporal regularity of data access.

[0113] Business fluctuation trends are the changes in cold chain logistics business volume over time, such as cyclical fluctuations in transportation order volume (e.g., weekend peaks), seasonal fluctuations (e.g., a surge in refrigerated demand in summer), or sudden changes caused by specific events (e.g., a sudden surge in vaccine transportation demand during the pandemic), used to predict dynamic changes in resource demand.

[0114] Step S502: Perform correlation analysis on the spatiotemporal distribution characteristics and business fluctuation trends to obtain the matching relationship between resource demand and business fluctuations.

[0115] The quantitative relationships derived from correlation analysis clarify the demand standards for various resources (such as storage capacity and computing power) under different business scenarios (such as peak periods and off-peak periods) (e.g., "for every 10% increase in order volume, the corresponding regional shard resources need to be expanded by 5%), providing a basis for dynamic scheduling.

[0116] Step S503: Dynamically adjust the allocation ratio of blockchain shard resources based on the matching relationship to obtain the resource allocation adjustment result.

[0117] Furthermore, the percentage allocation of total blockchain resources to each shard (e.g., shard A accounts for 30% and shard B accounts for 50%) is dynamically adjusted based on the matching relationship to form a new resource allocation scheme (e.g., during peak business periods, the proportion of shard B is increased to 70%).

[0118] Step S504: Monitor the resource idle status of each segment in real time to obtain idle monitoring data.

[0119] Idle monitoring data is real-time data collected on the usage status of each resource segment, including storage space occupancy, node response time, network bandwidth utilization, etc., providing real-time status basis for resource scheduling.

[0120] Step S505: Combine the resource allocation adjustment results with idle monitoring data to generate a resource scheduling scheme; the resource scheduling scheme includes dynamic expansion and contraction strategies for fragmented resources.

[0121] Dynamic expansion and contraction of sharded resources is a mechanism that automatically adjusts the scale of sharded resources according to real-time demand: Expansion: Enhances sharding processing capabilities by increasing the number of nodes and improving storage capacity; Contraction: Releases resources and reduces operation and maintenance costs by migrating data and shutting down redundant nodes.

[0122] Specifically, the process involves acquiring the spatiotemporal distribution characteristics of the encrypted evidence storage path (including the geographical distribution and access frequency of the path's associated nodes) and business fluctuation trends (such as peak transportation periods and changes in cargo types). Through correlation analysis, the resource demand patterns under different business scenarios are clarified (e.g., the increased demand for temperature data storage resources during peak cold chain transportation seasons), resulting in a matching relationship between resource demand and business fluctuations. Based on this matching relationship, the resource allocation ratio of each blockchain shard is dynamically adjusted (e.g., increasing shard resources corresponding to high-access-frequency areas), forming a resource allocation adjustment result. Simultaneously, the resource idle status of each shard is monitored in real time (e.g., storage space utilization, node load), generating idle monitoring data. Finally, combining the resource allocation adjustment result and the idle monitoring data, a resource scheduling scheme is generated, incorporating strategies for dynamic expansion (e.g., increasing shard storage capacity before business peaks) and contraction (e.g., reducing redundant resources during business troughs).

[0123] This embodiment achieves on-demand allocation of blockchain sharding resources by associating spatiotemporal distribution characteristics with business fluctuation trends, avoiding resource waste and overload. Real-time monitoring and dynamic adjustment mechanisms ensure the flexibility and efficiency of resource utilization, and the rational allocation of idle resources further improves overall resource utilization. The overall process not only meets the dynamic demand for storage resources caused by fluctuations in cold chain logistics operations but also reduces the operational costs of blockchain storage through precise scheduling, providing reliable resource guarantees for the efficient storage of cold chain data.

[0124] In one embodiment, the method may further include the following steps:

[0125] Step S601: Quantitatively evaluate the execution effect of the optimal resource scheduling scheme to obtain quantitative evaluation indicators and evaluation results; the quantitative evaluation indicators include blockchain storage resource utilization rate, data storage response latency, abnormal data tracing accuracy rate, and shard resource load balancing degree.

[0126] Preferably, the evaluation results are conclusive data obtained after statistical analysis of quantitative evaluation indicators (such as "storage resource utilization rate of 85% and average response latency of 0.5 seconds"), which are used to judge the merits of the resource scheduling scheme.

[0127] Step S602: Use edge computing nodes to monitor the dynamic change trend of the evidence storage weight value in real time, and adjust the parameters of the evidence storage model in combination with the evaluation results to obtain an optimized evidence storage model.

[0128] The data storage weight value changes in real time with cold chain transportation scenarios (such as temperature fluctuations and equipment status) (e.g., the weight value increases sharply when the temperature is abnormal during transportation). This data is continuously collected and monitored by edge computing nodes to reflect the dynamic adjustment needs of data storage priority.

[0129] Step S603: The optimized evidence storage model is synchronized to the edge nodes of the entire cold chain transportation chain. Based on the evidence storage weight value of each node and the actual storage path, storage cost data including the proportion of blockchain and database storage is generated.

[0130] To illustrate, edge nodes in the entire cold chain transportation chain are edge computing devices distributed in various links of cold chain transportation (such as refrigerated trucks, transit cold storage, and distribution points). They are responsible for local data collection and weight calculation. After synchronously optimizing the evidence storage model, the consistency of the data evidence storage strategy across the entire chain can be achieved.

[0131] The actual storage path is the actual storage location information of the data (such as "high-weight data is stored in blockchain shard A, and low-weight data is stored in database server B"), which corresponds to the evidence storage weight value and is used to count the resource usage of different storage media.

[0132] Storage cost data is based on the evidence storage weight value of each node and the actual storage path, and is used to calculate the consumption ratio of blockchain and database storage resources (e.g., "blockchain storage accounts for 30%, database accounts for 70%) and the corresponding costs (e.g., storage capacity, maintenance fees).

[0133] Specifically, starting with a quantitative evaluation of the execution effect of the optimal resource scheduling scheme, the evaluation is carried out through indicators such as blockchain storage resource utilization, data storage response latency, abnormal data tracing accuracy, and sharding resource load balancing. The evaluation results are obtained by using edge computing nodes to monitor the dynamic change trend of the storage weight value in real time and adjusting the parameters of the storage model in combination with the evaluation results to form an optimized storage model. The optimized storage model is synchronized to the edge nodes of the entire cold chain transportation chain. Each node, based on its own storage weight value and actual storage path, counts and summarizes the storage ratio of the blockchain and the database to generate storage cost data.

[0134] This embodiment quantifies the effectiveness of resource scheduling, providing data support for model optimization; real-time monitoring of edge nodes ensures the model can quickly respond to business changes, improving the accuracy of data storage; end-to-end model synchronization and storage cost statistics ensure the consistency of data storage strategies while reducing overall costs through differentiated storage. It achieves dynamic optimization and refined cost management of cold chain data storage, effectively reducing storage resource consumption while ensuring data quality.

[0135] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0136] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the aforementioned method for selective blockchain-based evidence storage of cold chain logistics transportation data based on dynamic weights.

[0137] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0138] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0139] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for selective blockchain-based evidence storage of cold chain logistics transportation data based on dynamic weights, characterized in that: The method includes: Acquire multi-dimensional status feature data of cold chain transportation collected in real time by edge computing nodes; the multi-dimensional status feature data includes temperature and humidity threshold deviation rate, geographical location displacement vector, equipment power consumption fluctuation coefficient and communication link quality index; The multidimensional state feature data is subjected to time-series feature extraction and dynamic correlation analysis. Key features are extracted, enhanced, and corrected to generate optimized evidence storage weight values. The evidence storage weight value is compared based on a preset threshold, and blockchain storage resources are allocated and encrypted evidence storage paths are generated according to the comparison results. By combining the spatiotemporal distribution characteristics of the encrypted evidence storage path with business fluctuation trends, the resource allocation ratio of blockchain shards is dynamically adjusted and idle status is monitored to generate a resource scheduling scheme.

2. The method according to claim 1, characterized in that, The step of extracting time-series features and performing dynamic correlation analysis on the multidimensional state feature data, extracting key features for enhancement and correction, and generating optimized evidence preservation weight values ​​includes: Based on the multidimensional state feature data, the combined weights of each dimension feature are calculated using an improved analytic hierarchy process to generate an initial weight vector. The initial weight vector is input into a pre-trained long short-term memory network to extract temporal feature dependencies and generate a temporally optimized adjusted weight matrix; the parameters of the long short-term memory network are obtained through pre-training with historical cold chain transportation data. A graph neural network enhanced by an attention mechanism is used to fuse the temporal correlation between the adjusted weight matrix and the features to generate a storage weight tensor; the storage weight tensor quantifies the storage priority of features of each dimension under different spatiotemporal scenarios. If the component values ​​of the evidence storage weight tensor are lower than the adaptive threshold, wavelet transform and feature reconstruction are performed on the corresponding dimension features, and an optimized evidence storage weight value is generated by combining the historical anomaly contribution correction coefficient.

3. The method according to claim 2, characterized in that, The temporal correlation between the features is calculated using the following formula, including: Where, r ij (t) represents the temporal correlation between feature i and feature j at time t, ω i (tk) represents the weight value of feature i at time tk, ω i (t) represents the weight value of feature i at time t, obtained after time-series optimization using a Long Short-Term Memory (LSTM) network, h t-1 W represents the hidden state of the LSTM at time t-1. l U l V l b l Let ω represent the learnable parameters of the l-th layer of the LSTM, (·) represent taking the i-th component of vector i, and ω 初始，i Represents the initial weight vector. This represents the average weight of feature i within the time window T, where T represents the size of the sliding time window dynamically set based on the feature's fluctuation period.

4. The method according to claim 2, characterized in that, The evidence preservation weight value is obtained through the following steps: Obtain the component value data of the evidence storage weight tensor; the component value data is used for anomaly detection in the feature dimension; Adaptive threshold filtering is performed on the component value data. If the value is below the threshold, the corresponding dimension feature is marked as an anomaly to obtain the anomaly feature. The abnormal features are processed by wavelet transform to obtain the feature data after time-frequency domain decomposition; Based on the feature data, feature reconstruction is performed to generate a denoised reconstructed feature matrix; The reconstructed feature matrix is ​​fused with the historical anomaly contribution correction coefficient to obtain the optimized evidence storage weight value. The optimized weight values ​​are subjected to a distribution consistency check. If there is a deviation, a support vector machine is used to construct an evidence storage model for classification optimization to obtain the weight distribution data. The parameters of the evidence preservation model are updated based on the weight distribution data to determine the final evidence preservation weight.

5. The method according to claim 4, characterized in that, The historical anomaly contribution correction coefficient is calculated using the following formula: b i =ρ·f i +(1-p)e j Where, β i f represents the historical anomaly contribution correction coefficient for feature i. i c represents the percentage of abnormal frequencies of feature i. i Let i represent the total number of times feature i appears anomaly in historical data, n represent the total number of features, and e represent the total number of features. j I represents the degree of anomalous influence of feature i. i,t This represents an indicator function, which takes a value of 1 when feature i experiences an anomaly at time t, and a value of 0 otherwise, |Δy t | represents the system output deviation at time t, T represents the total statistical time length, and ρ represents the balance coefficient.

6. The method according to claim 1, characterized in that, The step of comparing the evidence storage weight value based on a preset threshold, allocating blockchain storage resources according to the comparison result, and generating an encrypted evidence storage path includes: The evidence storage weight value is compared with a preset threshold in a hierarchical manner to obtain a comparison result containing the weight interval identifier; Based on the comparison results, the blockchain storage resources are dynamically allocated. If the data falls into the high-weight range, it is marked as high-priority data. The high-priority data is matched with a multi-node redundant storage resource pool that meets its attribute requirements; An encrypted evidence storage path is generated based on the attribute requirements of the storage resource pool and the high-priority data; the encrypted evidence storage path includes the node access sequence and data shard index corresponding to the storage resource pool.

7. The method according to claim 1, characterized in that, The method combines the spatiotemporal distribution characteristics of the encrypted evidence storage path with business fluctuation trends to dynamically adjust the blockchain sharding resource allocation ratio and monitor idle status, generating a resource scheduling scheme, including: The spatiotemporal distribution characteristics and business fluctuation trends of the encrypted evidence storage path are obtained; the spatiotemporal distribution characteristics include the geographical distribution and time period access frequency of the path-related nodes. A correlation analysis was conducted between the spatiotemporal distribution characteristics and business fluctuation trends to obtain the matching relationship between resource demand and business fluctuations; Based on the matching relationship, the allocation ratio of blockchain shard resources is dynamically adjusted to obtain the resource allocation adjustment result; Real-time monitoring of resource idleness in each segment yields idleness monitoring data; A resource scheduling scheme is generated by combining the resource allocation adjustment results with idle monitoring data; the resource scheduling scheme includes dynamic expansion and contraction strategies for fragmented resources.

8. The method according to claim 1, characterized in that, The method further includes: The execution effect of the optimal resource scheduling scheme is quantitatively evaluated to obtain quantitative evaluation indicators and evaluation results; the quantitative evaluation indicators include blockchain storage resource utilization rate, data storage response latency, abnormal data tracing accuracy rate, and shard resource load balancing degree. The dynamic trend of the evidence storage weight value is monitored in real time using edge computing nodes, and the parameters of the evidence storage model are adjusted in combination with the evaluation results to obtain an optimized evidence storage model. The optimized evidence storage model will be synchronized to the edge nodes of the entire cold chain transportation chain. Based on the evidence storage weight value of each node and the actual storage path, storage cost data including the proportion of blockchain and database storage will be generated.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images