Cold-chain logistics low-carbon temperature control data sharing method and device based on federated learning

Abstract

The application provides a cold-chain logistics low-carbon temperature control data sharing method and device based on federated learning, relates to the technical field of federated learning, and solves the technical problems that the existing technology has high communication energy consumption in the cold-chain logistics low-carbon temperature control data sharing process and lacks low-carbon guidance. The method specifically comprises the following steps: obtaining cold-chain logistics temperature control data; determining a carbon influence factor based on the data characteristics of the cold-chain logistics temperature control data; performing screening processing on local model parameters based on the carbon influence factor to obtain target model parameters; and uploading the target model parameters to a server. The application is used for cold-chain logistics low-carbon temperature control data sharing.

Description

Technical Field

[0001] This application relates to the field of federated learning technology, and in particular to a method and apparatus for sharing low-carbon temperature control data in cold chain logistics based on federated learning. Background Technology

[0002] With the rapid development of the cold chain logistics industry, the demand for accurate collection and sharing of temperature control data during the logistics process is increasing. Cold chain logistics involves multiple links and participants, and the data from each party often forms data silos. Federated learning, as a privacy-preserving machine learning technology, can achieve multi-party joint modeling without sharing the original data. However, existing federated learning solutions usually pursue the maximization of model accuracy, often neglecting the energy consumption of the communication process during model parameter updates. Especially in the cold chain logistics scenario, temperature control data is characterized by large volume and high update frequency. Directly applying traditional federated learning methods will lead to huge communication overhead and energy consumption, which contradicts the current development goal of low-carbon logistics. In addition, existing data sharing solutions do not fully utilize the strong correlation between cold chain temperature control data (such as temperature fluctuation rate and cooling duration) and carbon emissions, resulting in a lack of targeting in the data sharing process and the transmission of a large amount of low-value redundant data. Therefore, existing technologies suffer from high communication energy consumption and a lack of low-carbon orientation in the low-carbon temperature control data sharing process of cold chain logistics. Summary of the Invention

[0003] This application provides a method and apparatus for sharing low-carbon temperature control data in cold chain logistics based on federated learning, which solves the technical problems of high communication energy consumption and lack of low-carbon orientation in the existing technology of sharing low-carbon temperature control data in cold chain logistics.

[0004] To achieve the above objectives, this application adopts the following technical solution: Firstly, a method for sharing low-carbon temperature control data in cold chain logistics based on federated learning is provided, including: acquiring cold chain logistics temperature control data; determining carbon impact factors based on the data characteristics of the cold chain logistics temperature control data; filtering and processing local model parameters based on the carbon impact factors to obtain target model parameters; and uploading the target model parameters to the server.

[0005] The above scheme introduces a carbon impact factor to evaluate the value of model parameters and selects high-value parameters for uploading based on this factor, which effectively reduces the amount of data transmission, thereby reducing communication energy consumption and realizing the decarbonization of the cold chain logistics data sharing process.

[0006] In conjunction with the first aspect above, in one possible implementation, the data features include at least one of a first feature and a second feature, wherein the first feature is used to characterize temperature deviation and the second feature is used to characterize energy consumption ratio; based on the data features of cold chain logistics temperature control data, determining the carbon impact factor includes: calculating the carbon impact factor based on the correlation between the first feature or the second feature and carbon emissions.

[0007] In conjunction with the first aspect above, in one possible implementation, the local model parameters are screened based on the carbon impact factor to obtain the target model parameters, including: in response to the carbon impact factor meeting the preset screening conditions, the local model parameters are determined as the target model parameters.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, a carbon impact factor is calculated based on the correlation between the first or second feature and carbon emissions, including: obtaining the temperature fluctuation rate and refrigeration duration of cold chain logistics temperature control data; and calculating the carbon impact factor based on the temperature fluctuation rate, refrigeration duration, and correlation.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the method further includes: dynamically determining a screening threshold based on at least one of historical energy consumption data or current network status; the screening threshold is used to determine whether the carbon impact factor meets the preset screening conditions.

[0010] In conjunction with the first aspect mentioned above, in one possible implementation, before uploading the target model parameters to the server, the method further includes: processing the identity identifier of the cold chain logistics temperature control data based on a hash algorithm to obtain a hash ID; and performing privacy alignment processing on the target model parameters based on the hash ID.

[0011] In conjunction with the first aspect above, in one possible implementation, the data characteristics include at least one of temperature fluctuation rate, cooling duration, temperature deviation, and energy consumption ratio.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, before filtering the local model parameters based on the carbon impact factor, the method further includes: performing local training on the initial model based on cold chain logistics temperature control data to obtain local model parameters.

[0013] In conjunction with the first aspect mentioned above, in one possible implementation, the method further includes: receiving global model parameters issued by the server, wherein the global model parameters are obtained by the server aggregating target model parameters uploaded by multiple clients; and updating the initial model based on the global model parameters.

[0014] Secondly, a data sharing device for low-carbon temperature control in cold chain logistics based on federated learning is provided, comprising: a communication unit and a processing unit; the communication unit is used to acquire cold chain logistics temperature control data; the processing unit is used to determine carbon impact factors based on the data characteristics of the cold chain logistics temperature control data; to filter and process local model parameters based on the carbon impact factors to obtain target model parameters; and to control the communication unit to upload the target model parameters to the server.

[0015] This application acquires cold chain logistics temperature control data and determines carbon impact factors based on data characteristics. These factors are then used to filter local model parameters, uploading only high-value target model parameters. This innovative approach combines the low-carbon goals of cold chain logistics with the parameter update mechanism of federated learning. It utilizes the correlation between temperature control data characteristics (such as temperature deviation and energy consumption ratio) and carbon emissions to guide parameter selection, eliminating the transmission of low-value or redundant parameters. This not only significantly reduces the amount of communication data during data sharing, thereby directly reducing communication energy consumption and carbon emissions, but also improves model convergence efficiency by selecting high-value parameters, achieving a dual improvement in low-carbon environmental protection and data sharing efficiency.

[0016] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a low-carbon temperature control data sharing method for cold chain logistics based on federated learning, provided for an embodiment of this application; Figure 2 A system architecture diagram of a low-carbon temperature control data sharing system for cold chain logistics based on federated learning is provided for embodiments of this application. Figure 3 A schematic diagram of a low-carbon temperature control data sharing device for cold chain logistics based on federated learning, provided in an embodiment of this application; Figure 4This is a schematic diagram of the hardware structure of a low-carbon temperature control data sharing device for cold chain logistics based on federated learning, provided in an embodiment of this application. Detailed Implementation

[0018] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0019] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0020] Example 1: like Figure 1 As shown, this embodiment provides a low-carbon temperature control data sharing method for cold chain logistics based on federated learning. This method aims to address the problems of high communication energy consumption and lack of low-carbon guidance in existing federated learning processes. By introducing a carbon impact factor to screen model parameters, it achieves a low-carbon data sharing process. The method specifically includes the following steps: S101. Obtain temperature control data for cold chain logistics.

[0021] Among them, cold chain logistics temperature control data refers to the set of data related to temperature control generated in the cold chain logistics transportation and warehousing process.

[0022] In this embodiment, the data can be collected in real time by various sensors (such as temperature sensors, humidity sensors, and energy consumption monitors) deployed at nodes such as cold chain transport vehicles and cold storage facilities, or by reading historical records from the database of the logistics management system. Specifically, the acquired data may include raw data such as ambient temperature, set temperature, refrigeration equipment operating power, and refrigeration duration.

[0023] It should be noted that the acquired data usually has time-series characteristics, and the amount of data increases linearly with the transportation time, which provides a foundation for subsequent data feature extraction.

[0024] As an example, a cold chain transport vehicle collects temperature data and refrigeration unit energy consumption data inside the compartment every 5 minutes through an on-board terminal, forming a continuous temperature control data stream.

[0025] Based on the above steps, by obtaining comprehensive temperature control data, the necessary data support was provided for the subsequent calculation of the carbon impact factor, ensuring the accuracy of the analysis results.

[0026] S102. Determine the carbon impact factor based on the data characteristics of cold chain logistics temperature control data.

[0027] Among them, the carbon impact factor is a quantitative indicator used to characterize the contribution of the current batch of temperature control data to carbon emissions. Data features are key information extracted from the raw temperature control data that reflects the quality of temperature control and energy consumption levels.

[0028] In this embodiment, the process of determining the carbon impact factor is essentially the process of establishing a mapping relationship between temperature control data characteristics and carbon emissions. The raw data acquired in S101 is preprocessed and features are extracted, such as calculating temperature fluctuation rate and energy consumption ratio. Then, based on a preset evaluation model, these feature values ​​are converted into a specific carbon impact factor value. The higher this value, the greater the carbon emission tendency caused by temperature control behavior during that time period, or the richer the high-energy-consumption information contained in that data segment.

[0029] It should be noted that the purpose of introducing the carbon impact factor is to transform the abstract "low carbon" goal into a calculable and comparable numerical indicator, thereby guiding the subsequent data screening process and changing the traditional federated learning's single evaluation dimension that only focuses on model accuracy.

[0030] As an example, the temperature deviation and energy consumption ratio features of the current data segment are extracted, and a carbon impact factor between 0 and 1 is calculated by weighted summation.

[0031] Based on the above steps, by quantifying the contribution of data to carbon emissions, an objective basis for judgment is provided for the subsequent screening of high-value parameters, thus achieving a deep integration of low-carbon goals and data processing.

[0032] S103. Based on the carbon impact factor, the local model parameters are screened to obtain the target model parameters.

[0033] In this context, local model parameters refer to the model weights or gradient data obtained by the client based on local data under the federated learning framework. Target model parameters, on the other hand, are a selected subset of high-value parameters intended to be uploaded to the server.

[0034] In this embodiment, the screening process is a key step in reducing communication energy consumption. The carbon impact factor calculated in S102 is used as the screening criterion to determine whether the current local model parameters belong to high-carbon emission-related data or high-value data. Specifically, a screening threshold can be set. When the carbon impact factor meets a preset condition (such as exceeding a certain threshold), it is determined that the parameter has a significant contribution to model updates and a high correlation with carbon emission characteristics, and it is retained and marked as a target model parameter; otherwise, it can be temporarily discarded or its upload postponed. In this way, instead of blindly uploading all parameters, parameters strongly correlated with low-carbon characteristics are selectively transmitted, thereby reducing the amount of data transmitted at the source.

[0035] It should be noted that this screening mechanism differs from traditional data compression technology. It is based on the carbon emission value of the data content, which retains key information that helps improve the accuracy of the model while eliminating redundant low-value data, thus achieving a balance between low carbon emissions and high efficiency.

[0036] As an example, after completing local training, the client calculates the carbon impact factor for each parameter, retains only the top 20% of parameters by carbon impact factor as the target model parameters, and does not upload the remaining parameters.

[0037] Based on the above steps, the filtering mechanism effectively reduces the amount of data that needs to be uploaded, directly reducing energy consumption during communication, which is the core means to achieve low-carbon data sharing in cold chain logistics.

[0038] S104. Upload the target model parameters to the server.

[0039] In this context, the server refers to the central or coordinating node in the federated learning architecture that is responsible for aggregating the parameters of each client model.

[0040] In this embodiment, the client sends the target model parameters, filtered in step S103, to the server via the network communication module. Since the uploaded data has already been filtered, the network transmission load is significantly reduced compared to uploading all local model parameters. After receiving the target model parameters uploaded by each client, the server performs an aggregation operation to update the global model.

[0041] It should be noted that the uploaded data is the filtered "target model parameters" rather than the original temperature control data. This complies with the privacy protection principle of federated learning that "data does not leave the domain". At the same time, combined with the low-carbon screening strategy of this invention, it achieves the dual effect of privacy protection and low-carbon transmission.

[0042] As an example, the client sends the filtered gradient parameters, which are then packaged and encrypted, to a cloud server for aggregation via a 5G network or satellite communication link.

[0043] Based on the above steps, by uploading only the filtered parameters, the network bandwidth usage and transmission energy consumption are significantly reduced, thereby achieving the low-carbon emission target in the cold chain logistics data sharing process.

[0044] Based on the above technical solution, this embodiment acquires cold chain logistics temperature control data and determines carbon impact factors based on data characteristics. These factors are then used to filter local model parameters, uploading only high-value target model parameters. This solution creatively combines the low-carbon goals of cold chain logistics with the parameter update mechanism of federated learning. It utilizes the correlation between temperature control data characteristics and carbon emissions to guide parameter selection, eliminating the transmission of low-value or redundant parameters. This not only significantly reduces the amount of communication data during data sharing, thereby directly reducing communication energy consumption and carbon emissions, but also improves model convergence efficiency by selecting high-value parameters, achieving a dual improvement in low-carbon environmental protection and data sharing efficiency.

[0045] Example 2: Based on Example 1 above, this example elaborates on the specific implementation of determining the carbon impact factor in S102. Specifically, the data features include at least one of a first feature and a second feature, where the first feature characterizes temperature deviation and the second feature characterizes energy consumption ratio. Furthermore, to more comprehensively assess the impact of carbon emissions, the data features may also include temperature fluctuation rate and cooling duration.

[0046] Temperature deviation refers to the difference between the actual temperature and the preset standard temperature during cold chain logistics. It is usually used to measure the control accuracy of the temperature control system. A larger deviation often means a higher risk of potential energy waste. Energy efficiency ratio is the ratio of the cooling capacity output by the refrigeration equipment to the input power, directly reflecting the equipment's energy efficiency level. Temperature volatility characterizes the degree of temperature change; a higher volatility indicates more frequent start-ups and shutdowns of the refrigeration equipment, usually resulting in greater energy loss. Cooling duration refers to the cumulative time the refrigeration equipment operates continuously; a longer duration naturally leads to higher accumulated energy consumption and carbon emissions.

[0047] In this embodiment, the carbon impact factor is calculated based on the correlation between the first feature or the second feature and carbon emissions. Specifically, this includes: obtaining the temperature fluctuation rate and refrigeration duration of cold chain logistics temperature control data; and calculating the carbon impact factor based on the temperature fluctuation rate, refrigeration duration, and correlation.

[0048] Specifically, when calculating the carbon impact factor, the mapping relationship between the above characteristics and carbon emissions will be taken into account.

[0049] As an example, the carbon impact factor satisfies the following formula:

[0050] in, Carbon impact factor, For temperature deviation, The preset temperature reference value, It is the reciprocal of the energy efficiency ratio (used to unify the units; a higher energy efficiency ratio usually means more energy saving, so the reciprocal or negative correlation is taken). This is a reference value for energy consumption. For temperature fluctuation rate, For cooling time, As a time reference base value, , , , These are the weight coefficients corresponding to each feature term, and + =1.

[0051] As another example, data collected from a cold chain transportation node shows that the current temperature fluctuation rate is high and the refrigeration time exceeds the normal range, resulting in a calculated carbon impact factor. The value is significantly higher than the historical average. This indicates that the data for this period contains high-energy-consuming characteristics, and the corresponding model parameters have higher weight value or screening priority in federated learning updates.

[0052] Based on the above steps, this embodiment introduces multi-dimensional data features such as temperature deviation, energy consumption ratio, temperature fluctuation rate, and cooling duration, and provides specific quantitative calculation formulas, thereby transforming the abstract concept of carbon emissions into a calculable numerical indicator. This not only solves the technical problem of "how to calculate" carbon impact factors, but also improves the accuracy and robustness of factor assessment through the fusion of multi-dimensional features, providing a solid data foundation for subsequent parameter selection.

[0053] Example 3: Based on Example 1 above, this example elaborates on the specific logic of the local model parameter screening process based on the carbon influence factor in S103. Specifically, the screening process is achieved by comparing the carbon influence factor with preset screening conditions.

[0054] The preset screening criteria refer to the standards used to determine whether model parameters are worth uploading, and are usually expressed in the form of screening thresholds. The screening threshold is a critical value that measures the level of the carbon impact factor. When the carbon impact factor reaches or exceeds this critical value, it means that the corresponding model parameters have high characterization value for carbon emission features and are worth uploading to participate in global model updates.

[0055] In this embodiment, during the screening process, the carbon impact factor corresponding to the model parameters of the current batch is calculated in real time and compared with the screening threshold. If the carbon impact factor meets the preset screening conditions (e.g., the carbon impact factor is greater than or equal to the screening threshold), the local model parameter is determined as the target model parameter and added to the upload queue; otherwise, if the carbon impact factor does not meet the preset screening conditions, the parameter is considered low-value or redundant data and is discarded or temporarily stored.

[0056] Furthermore, to improve the adaptability of the solution in different environments, this embodiment also provides a mechanism for dynamically determining the screening threshold. The method further includes: dynamically determining the screening threshold based on at least one of historical energy consumption data or the current network state.

[0057] Among them, the screening threshold is used to determine whether the carbon impact factor meets the preset screening conditions; historical energy consumption data refers to the energy consumption statistics recorded by the client in the past communication process, which reflects the long-term energy consumption level of the equipment; and the current network status refers to the real-time communication indicators such as network bandwidth, latency, and packet loss rate at the current moment.

[0058] Specifically, the screening threshold is not a fixed constant, but a variable that adaptively adjusts with changes in the environment. The system has a built-in adaptive adjustment strategy that establishes a mapping relationship between historical energy consumption data, current network status, and the screening threshold. When the current network status is detected to be congested (e.g., network latency exceeds a preset latency threshold, or available bandwidth is lower than a preset bandwidth threshold), the screening threshold is automatically increased. Increasing the threshold means stricter screening criteria; only a small number of core parameters with extremely high carbon impact factors can pass the screening, thus significantly reducing the amount of data that needs to be uploaded. This not only alleviates network congestion, but more importantly, reducing data transmission volume means reducing the transmission power and duration of communication modules, directly achieving low carbon emissions. Conversely, when the network status is good, the screening threshold can be appropriately lowered to allow more parameters to participate in the update, thereby accelerating model convergence. Similarly, if historical energy consumption data shows that recent equipment energy consumption is too high, the threshold will be increased accordingly to control energy consumption.

[0059] It should be noted that this dynamic adjustment mechanism tightly couples the communication process of federated learning with the actual operating environment of cold chain logistics (network fluctuations, energy consumption limitations), achieving synergistic optimization of "low-carbon" goals and "communication efficiency." Compared to traditional fixed-threshold screening methods, the dynamic threshold mechanism is more intelligent and robust.

[0060] As an example, a cold chain transport vehicle travels to a remote area where the network signal degrades from 4G to 2G, resulting in a significant decrease in network bandwidth. Detecting this change in network status triggers dynamic adjustment logic, raising the filtering threshold from the default 0.5 to 0.8. At this point, only model parameters with a carbon impact factor F value greater than 0.8 are retained and uploaded; the remaining parameters are temporarily discarded. This adjustment reduces the amount of uploaded data by approximately 60%, ensuring successful transmission of core parameters while avoiding the additional energy consumption caused by large-scale data retransmission.

[0061] Based on the above steps, this embodiment introduces a dynamic screening threshold mechanism, enabling the parameter screening process to flexibly adapt to changes in network environment and energy consumption. When network congestion or energy constraints occur, the transmission volume is proactively reduced by increasing the threshold, achieving dual savings in communication resources and energy, and further strengthening the low-carbon attributes of the cold chain logistics data sharing process.

[0062] Example 4: Based on Embodiment 1 above, this embodiment elaborates on the data security preprocessing steps before uploading the target model parameters to the server. Specifically, before uploading the target model parameters to the server, the method further includes: processing the identity identifier of the cold chain logistics temperature control data based on a hash algorithm to obtain a hash ID; and performing privacy alignment processing on the target model parameters based on the hash ID.

[0063] In this context, "identity identifier" refers to information that uniquely identifies the source of cold chain logistics data, such as the license plate number of the transport vehicle, the cold storage number, the waybill number, or the unique identification code of the consignee. A hash algorithm is a one-way cryptographic system that maps input data of arbitrary length to output data of fixed length (i.e., a hash value). This process is irreversible; the original input cannot be deduced from the hash value. Privacy alignment refers to the process in multi-party federated learning tasks where parties determine the correspondence of data samples or verify the legitimacy of data sources by comparing hash IDs without exchanging original identity information.

[0064] In this embodiment, considering that cold chain logistics involves multiple participants such as cargo owners, logistics companies, and warehousing providers, and that the data from each party often contains sensitive commercial information or private data, directly uploading model parameters with their original identifiers poses a risk of identity leakage. Therefore, before executing S104 to upload parameters, the identifier of the data is extracted and encrypted using a hash algorithm (such as SHA-256, MD5, etc.) to generate a meaningless hash ID. Subsequently, the client binds this hash ID to the target model parameter and uploads it. After receiving the data, the server can complete the aggregation and alignment of parameters by comparing the consistency of the hash IDs, without needing to know the specific identifier content.

[0065] It should be noted that this hash ID-based privacy alignment mechanism, while ensuring the trustworthiness of the data source, effectively severs the direct link between model parameters and the original identity, thus building a defense depth in the "data security" dimension. Even if data is intercepted during transmission, attackers cannot deduce specific vehicle or cargo owner information through the hash ID, thereby greatly improving the security of cold chain logistics data sharing.

[0066] As an example, a cold chain logistics alliance includes multiple transportation companies. When each company uploads temperature control data model parameters, the system automatically extracts the vehicle license plate number as an identifier and calculates its SHA-256 hash value as a hash ID. The server compares the hash IDs to confirm whether parameters from different companies belong to the same transportation batch or the same alliance member, thus achieving reliable aggregation of multi-party data without exposing the actual license plate information.

[0067] Based on the above steps, this embodiment introduces hash ID technology to desensitize identity information before data upload, solving the problem of identity privacy leakage when multiple parties participate in federated learning, and providing reliable security for low-carbon temperature control data sharing in cold chain logistics.

[0068] Example 5: Based on the above embodiment 1, this embodiment elaborates in detail the source of local model parameters and the global update closed-loop process of federated learning.

[0069] Specifically, before filtering the local model parameters based on the carbon impact factor, the method also includes: training the initial model locally based on cold chain logistics temperature control data to obtain local model parameters.

[0070] The initial model refers to the model held by the client at the start of the current training round. It can be the untrained original model or the global model updated in the previous round. Local training refers to the process of updating the model parameters locally on the client using the collected temperature control data.

[0071] In this embodiment, the client uses the acquired cold chain logistics temperature control data as a training sample set and inputs it into the initial model for forward propagation calculation. Specifically, a loss function is defined to measure the difference between the model's predicted values ​​and the true labels (such as the probability of temperature anomalies or energy consumption levels). The error is backpropagated using a gradient descent algorithm (such as stochastic gradient descent SGD) to calculate the gradient update of the model parameters, thereby obtaining the updated local model parameters.

[0072] It should be understood that the purpose of local training is to extract the feature patterns in the temperature control data and solidify them into the model parameters, which will then serve as the original objects for the filtering operation.

[0073] As an example, the client uses the temperature sequence data of the carriage collected in the past hour to train the local LSTM (Long Short-Term Memory) model. By minimizing the mean square error between the predicted temperature and the actual temperature, the weight matrix and bias vector of the model are updated to obtain the local model parameters for this round.

[0074] Based on the above steps, this embodiment transforms the temperature control data features scattered across various nodes into shareable model parameters through a local training process, providing the necessary data foundation for subsequent carbon impact factor screening and ensuring the integrity of the federated learning process.

[0075] Furthermore, in order to achieve continuous model optimization and knowledge sharing, the method also includes: receiving global model parameters issued by the server, which are obtained by the server aggregating target model parameters uploaded by multiple clients; and updating the initial model based on the global model parameters.

[0076] The global model parameters refer to the unified model parameters generated by the server after integrating the model parameters contributed by all participants. Aggregation processing refers to the process by which the server merges the parameters uploaded by each client according to a preset strategy (such as weighted average, median aggregation, etc.).

[0077] In this embodiment, after each client uploads the filtered target model parameters to the server, the server does not simply add them up, but performs an aggregation operation. Since this solution has already removed redundant parameters associated with low value or high carbon emissions in the preceding steps, the server receives only high-value target model parameters. The server can use the FedAvg (Federated Average) algorithm to perform a weighted average of the target model parameters based on the amount of data or carbon impact factor weight of each client, generating new global model parameters. Subsequently, the server distributes the global model parameters to each client. After receiving these parameters, the client replaces its local initial model parameters with them, thus completing one round of global model update iteration.

[0078] It should be understood that this update mechanism enables clients to share each other's knowledge while keeping the original data within the domain.

[0079] It should be noted that because the uploaded parameters are "refined" after filtering, the computational resources and communication bandwidth required by the server during aggregation are significantly reduced. This further reduces energy consumption from the global perspective of federated learning, aligning with the overall goal of low-carbon development in cold chain logistics. Furthermore, the filtering mechanism eliminates low-quality parameters that may contain noise or anomalies, resulting in better robustness and convergence efficiency in the aggregated global model.

[0080] As an example, in a cold chain logistics alliance, the server collects target model parameters from 100 transport vehicles. Because each vehicle has undergone carbon impact factor screening to remove approximately 30% of redundant parameters, the server's data processing workload is reduced by 30%, improving the accuracy of the aggregated global model in predicting temperature control energy consumption. After receiving this global model, the client updates its local model weights and prepares for the next round of training.

[0081] Based on the above steps, this embodiment demonstrates the feasibility of the invention under a federated learning architecture by constructing a complete closed loop of "local training - filtering and uploading - global aggregation - model distribution". Furthermore, the participation of filtered parameters in aggregation not only reduces the computational and communication energy consumption of the server but also improves the convergence quality of the global model, achieving a synergistic optimization that is both low-carbon and highly efficient.

[0082] Example 6: This embodiment provides a low-carbon temperature control data sharing system for cold chain logistics based on federated learning, with the following architecture: Figure 2 As shown, the system includes three federated learning-based cold chain logistics low-carbon temperature control data sharing devices 201 and a server 202. The three devices 201 correspond to three types of edge nodes in the cold chain logistics scenario: refrigerated truck onboard temperature control terminals, cold storage temperature control monitoring gateways, and transit warehouse temperature control data acquisition nodes. Each device 201 establishes a bidirectional communication link with the server 202. Specifically, each device 201 uploads target model parameters to the server 202. Based on locally collected cold chain temperature control data (such as real-time temperature inside the compartment, refrigeration unit energy consumption, temperature control equipment operating time, and ambient humidity), each device 201 completes federated learning model training and low-carbon screening locally before uploading lightweight model parameters, temperature control data features, and carbon impact factor data to the server 202, rather than the original full-volume temperature control data, thereby reducing data transmission volume and communication energy consumption. The server 202... The server 202 sends global model parameters to device 201. Specifically, after receiving the target model parameters uploaded by all devices 201, the server 202 performs a global federated aggregation operation to weight and aggregate the low-carbon temperature control model parameters of each device 201, generating globally optimized low-carbon temperature control shared model parameters. These global model parameters are then transmitted to each device 201 via the distribution link. After receiving the global model parameters, each device 201 updates its local federated learning model, thereby optimizing its local cold chain temperature control strategy. This achieves cross-node collaborative sharing of temperature control data and low-carbon optimization. Throughout the process, the bidirectional data interaction between device 201 and server 202 only transmits model parameters and lightweight feature data, ensuring the privacy and security of cold chain temperature control data while avoiding the high energy consumption problem caused by full data transmission. This effectively meets the dual needs of low-carbon temperature control and data sharing in cold chain logistics scenarios.

[0083] Example 7: The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as the federated learning-based cold chain logistics low-carbon temperature control data sharing device, includes at least one of the hardware structures and software modules corresponding to each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0084] This application embodiment can divide the federated learning-based cold chain logistics low-carbon temperature control data sharing device into functional units according to the above method example. For example, each function can be divided into separate functional units, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0085] When using integrated units, Figure 3 A possible structural schematic diagram of the federated learning-based cold chain logistics low-carbon temperature control data sharing device (referred to as federated learning-based cold chain logistics low-carbon temperature control data sharing device 30) involved in the above embodiments is shown. The federated learning-based cold chain logistics low-carbon temperature control data sharing device 30 includes a processing unit 301 and a communication unit 302, and may also include a storage unit 303. Figure 3 The schematic diagram shown can be used to illustrate the structure of the federated learning-based low-carbon temperature control data sharing device for cold chain logistics involved in the above embodiments.

[0086] when Figure 3 The schematic diagram shown illustrates the structure of the federated learning-based low-carbon temperature control data sharing device for cold chain logistics involved in the above embodiments. The processing unit 301 is used to control and manage the operation of the federated learning-based low-carbon temperature control data sharing device for cold chain logistics. The communication unit 302 is used for the federated learning-based low-carbon temperature control data sharing device to communicate with other devices. The storage unit 303 is used to store the program code and data of the federated learning-based low-carbon temperature control data sharing device for cold chain logistics.

[0087] For example, communication unit 302 is used to acquire temperature control data for cold chain logistics; The processing unit 301 is used to determine the carbon impact factor based on the data characteristics of cold chain logistics temperature control data; to filter and process the local model parameters based on the carbon impact factor to obtain the target model parameters; and to upload the target model parameters to the server.

[0088] In one possible implementation, the data features include at least one of a first feature and a second feature, wherein the first feature is used to characterize temperature deviation and the second feature is used to characterize energy consumption ratio; in another possible implementation, the processing unit 301 is further used to determine the carbon impact factor based on the data features of the cold chain logistics temperature control data, including: calculating the carbon impact factor based on the correlation between the first feature or the second feature and carbon emissions.

[0089] In one possible implementation, the processing unit 301 is further configured to determine the local model parameters as the target model parameters in response to the carbon influence factor meeting the preset screening conditions.

[0090] In one possible implementation, the communication unit 302 is further configured to acquire the temperature fluctuation rate and refrigeration duration of the cold chain logistics temperature control data; in another possible implementation, the processing unit 301 is further configured to calculate the carbon impact factor based on the temperature fluctuation rate, refrigeration duration, and correlation.

[0091] In one possible implementation, the processing unit 301 is further configured to dynamically determine a screening threshold based on at least one of historical energy consumption data or current network status; the screening threshold is used to determine whether the carbon impact factor meets the preset screening conditions.

[0092] In one possible implementation, the processing unit 301 is further configured to process the identity identifier of the cold chain logistics temperature control data based on a hash algorithm to obtain a hash ID; and to perform privacy alignment processing on the target model parameters based on the hash ID.

[0093] In one possible implementation, the data features include at least one of temperature fluctuation rate, cooling time, temperature deviation, and energy consumption ratio.

[0094] In one possible implementation, the processing unit 301 is also used to perform local training on the initial model based on cold chain logistics temperature control data to obtain local model parameters.

[0095] In one possible implementation, the processing unit 301 is further configured to receive global model parameters sent by the server, which are obtained by the server aggregating target model parameters uploaded by multiple clients; and update the initial model based on the global model parameters.

[0096] The processing unit 301 can be a processor or a controller, and the communication unit 302 can be a communication interface, transceiver, transceiver circuit, transceiver device, etc. The term "communication interface" is a general term and may include one or more interfaces. The storage unit 303 can be a memory. When the federated learning-based cold chain logistics low-carbon temperature control data sharing device 30 is a chip, the processing unit 301 can be a processor or a controller, and the communication unit 302 can be an input interface and / or an output interface, pins, or circuits, etc. The storage unit 303 can be a storage unit within the chip (e.g., a register, cache, etc.) or a storage unit located outside the chip (e.g., read-only memory (ROM), random access memory (RAM, etc.).

[0097] The communication unit can also be called a transceiver unit. The antenna and control circuit with transceiver functions in the federated learning-based cold chain logistics low-carbon temperature control data sharing device 30 can be considered as the communication unit 302 of the federated learning-based cold chain logistics low-carbon temperature control data sharing device 30, and the processor with processing functions can be considered as the processing unit 301 of the federated learning-based cold chain logistics low-carbon temperature control data sharing device 30. Optionally, the device in the communication unit 302 used to implement the receiving function can be considered as a communication unit, which is used to execute the receiving steps in the embodiments of this application. The communication unit can be a receiver, a receiver circuit, etc. The device in the communication unit 302 used to implement the transmitting function can be considered as a transmitting unit, which is used to execute the transmitting steps in the embodiments of this application. The transmitting unit can be a transmitter, a transmitter, a transmitting circuit, etc.

[0098] Figure 3 If the integrated units in the process are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. Storage media for storing computer software products include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0099] Figure 3 The units in the process can also be called modules; for example, a processing unit can be called a processing module.

[0100] This application also provides a hardware structure diagram of a federated learning-based cold chain logistics low-carbon temperature control data sharing device (referred to as federated learning-based cold chain logistics low-carbon temperature control data sharing device 40), see [link to relevant documentation]. Figure 4 The federated learning-based cold chain logistics low-carbon temperature control data sharing device 40 includes a processor 401, and optionally, a memory 402 connected to the processor 401.

[0101] In the first possible implementation, see Figure 4 The federated learning-based cold chain logistics low-carbon temperature control data sharing device 40 also includes a transceiver 403. The processor 401, memory 402, and transceiver 403 are connected via a bus. The transceiver 403 is used to communicate with other devices or communication networks. Optionally, the transceiver 403 may include a transmitter and a receiver. The device in the transceiver 403 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of this application. The device in the transceiver 403 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of this application.

[0102] Based on the first possible implementation method Figure 4 The schematic diagram shown can be used to illustrate the structure of the federated learning-based low-carbon temperature control data sharing device for cold chain logistics involved in the above embodiments.

[0103] in, Figure 4 This can also be illustrated by the system chip in a federated learning-based cold chain logistics low-carbon temperature control data sharing device. In this case, the actions performed by the aforementioned federated learning-based cold chain logistics low-carbon temperature control data sharing device can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0104] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0105] The processor in this application may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a separate semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a SoC (System-on-a-Chip), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), PLDs (programmable logic devices), or logic circuits that implement dedicated logic operations.

[0106] The memory in the embodiments of this application may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto.

[0107] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0108] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A method for sharing low-carbon temperature control data in cold chain logistics based on federated learning, characterized in that, include: Obtain temperature control data for cold chain logistics; Based on the data characteristics of the cold chain logistics temperature control data, the carbon impact factor is determined; Based on the carbon impact factor, the local model parameters are filtered to obtain the target model parameters; Upload the target model parameters to the server.

2. The method according to claim 1, characterized in that, The data features include at least one of a first feature and a second feature, wherein the first feature is used to characterize the temperature deviation and the second feature is used to characterize the energy consumption ratio. The determination of carbon impact factors based on the data characteristics of the cold chain logistics temperature control data includes: The carbon impact factor is calculated based on the correlation between the first feature or the second feature and carbon emissions.

3. The method according to claim 1, characterized in that, The process of filtering local model parameters based on the carbon impact factor to obtain target model parameters includes: In response to the carbon impact factor meeting the preset screening conditions, the local model parameters are determined as the target model parameters.

4. The method according to claim 2, characterized in that, The calculation of the carbon impact factor based on the correlation between the first feature or the second feature and carbon emissions includes: The temperature fluctuation rate and refrigeration duration of the cold chain logistics temperature control data are obtained. The carbon impact factor is calculated based on the temperature fluctuation rate, the cooling duration, and the correlation.

5. The method according to claim 2, characterized in that, The method further includes: The screening threshold is dynamically determined based on at least one of historical energy consumption data or current network status. The screening threshold is used to determine whether the carbon impact factor meets the preset screening conditions.

6. The method according to claim 3, characterized in that, Before uploading the target model parameters to the server, the method further includes: The identity identifier of the cold chain logistics temperature control data is processed based on a hash algorithm to obtain a hash ID; Privacy alignment is performed on the target model parameters based on the hash ID.

7. The method according to claim 1, characterized in that, The data characteristics include at least one of temperature fluctuation rate, cooling time, temperature deviation, and energy consumption ratio.

8. The method according to claim 1, characterized in that, Before performing the local model parameter screening process based on the carbon impact factor, the method further includes: The initial model is trained locally based on the cold chain logistics temperature control data to obtain the local model parameters.

9. The method according to claim 8, characterized in that, The method further includes: Receive global model parameters sent by the server, which are obtained by the server aggregating target model parameters uploaded by multiple clients; The initial model is updated based on the global model parameters.

10. A low-carbon temperature control data sharing device for cold chain logistics based on federated learning, characterized in that, The device includes: a communication unit and a processing unit; The communication unit is used to acquire temperature control data for cold chain logistics; The processing unit is configured to determine the carbon impact factor based on the data characteristics of the cold chain logistics temperature control data; filter the local model parameters based on the carbon impact factor to obtain the target model parameters; and control the communication unit to upload the target model parameters to the server.

Images