Logistics storage warehousing management method based on big data

Abstract

The invention relates to the technical field of logistics storage management, and discloses a logistics storage warehousing management method based on big data, which comprises the following steps: step 1, acquiring order data in an order management system through Internet of Things equipment, acquiring cargo information by using an RF ID tag and bar code scanning equipment, acquiring storage environment parameters by using a temperature and humidity sensor and a light sensor, and storing the storage environment parameters in a database; a built-in sensor of a forklift and a built-in sensor of an automatic guide vehicle are used for collecting equipment state data, real-time image data of goods are collected through a high-definition camera, and the collected data comprise timestamps, data source identifiers and initial data values. According to the method, modeling is carried out on historical order quantity, warehousing frequency, cargo types and time sequence characteristics by utilizing a long-short-term memory network model, so that time dependence and seasonal change trend of data are accurately captured, and accurate prediction of future warehousing demands is realized; and the effects of improving the prediction accuracy and providing a reliable basis for goods allocation and resource scheduling are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics warehousing management, and in particular to a logistics warehousing warehousing management method based on big data. Background Art

[0002] With the rapid development of the global economy and the booming e-commerce, the logistics industry is facing increasingly complex market demands. Logistics warehousing is the core link in supply chain management, and efficient management is directly related to the operational efficiency and cost control of the logistics system. However, the traditional logistics warehousing warehousing management method has many problems and cannot meet the needs of the modern logistics industry for intelligent and refined management. The following are the problems of traditional warehousing management:

[0003] The demand for logistics warehousing is highly uncertain and is affected by order quantity, commodity type, seasonality and sudden demand fluctuations. Traditional forecasting methods usually rely on empirical methods or simple linear regression models, which fail to effectively mine the time series laws and seasonal change trends in historical data, resulting in low forecasting accuracy, lagging or excess resource allocation, and failure to adapt to dynamically changing market demand.

[0004] As the core link of warehouse management, cargo space allocation has a direct impact on warehouse space utilization and operating efficiency. Traditional cargo space allocation methods mostly use fixed rules or static partitioning strategies, lacking dynamic adjustment capabilities. Failure to combine cargo volume, weight, turnover frequency and the real-time status of storage cargo space leads to chaotic placement of cargo, serious waste of space, and increased cargo search and transportation time.

[0005] The warehousing process involves the collaborative work of multiple resources such as forklifts, automated guided vehicles, loading and unloading equipment, and labor. Traditional resource scheduling methods fail to fully consider resource operation time, cost, and actual operation status. They usually adopt fixed static scheduling schemes and cannot dynamically respond to problems such as peak warehousing, resource conflicts, or idle equipment, which reduces resource utilization, increases enterprise operating costs, and restricts the improvement of logistics efficiency.

[0006] Common abnormal problems in the warehousing process include damaged goods, missing goods, data deviation or cargo location conflicts. Traditional abnormal monitoring mainly relies on manual inspections or delayed feedback mechanisms, which have problems such as low detection accuracy and slow response speed. Once an abnormal situation occurs, it will lead to interruption of warehousing operations, loss or delay of goods, and even have a chain effect on the overall warehousing operation process, further expanding losses.

[0007] In traditional logistics and warehousing management, data isolation between different systems is serious. The systems fail to achieve real-time sharing and integration of data, and cannot build an intelligent management platform based on full-link data. This leads to delayed warehousing management information and extensive decision-making, which cannot meet the real-time and accuracy requirements of the modern logistics industry.

[0008] Therefore, those skilled in the art provide a logistics warehousing and storage management method based on big data to solve the above-mentioned problems. Summary of the invention

[0009] In view of the deficiencies in the prior art, the present invention provides a logistics warehousing warehousing management method based on big data to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: A logistics warehousing warehousing management method based on big data, comprising:

[0011] Step 1: Collect order data in the order management system through IoT devices, collect cargo information using RFID tags and barcode scanning devices, collect storage environment parameters using temperature and humidity sensors and light sensors, collect equipment status data using built-in sensors of forklifts and automatic guided vehicles, and collect real-time image data of cargo through high-definition cameras. The collected data includes timestamps, data source identifiers, and initial data values.

[0012] Step 2: The multi-source data collected in step 1, including order data, cargo information, equipment status, environmental parameters and image data, are transmitted using the MQTT protocol. The transmitted data is integrated through the Kafka message queue and sorted by timestamp. The integrated data is stored in the Hadoop distributed file system, and the data source information is annotated for subsequent cleaning and standardization processing;

[0013] Step 3: For the integrated data stored in step 2, the interquartile range method is used to eliminate outliers in the data, and the missing values ​​are filled by the K nearest neighbor method according to the data source information. The cleaned data is standardized according to the maximum and minimum value ranges, and the standardized data is updated to the Hadoop distributed file system to form a unified standardized data set for prediction and optimization;

[0014] Step 4: extract the order quantity, cargo type, warehousing frequency, time characteristics and storage environment parameters from the standardized data set formed in step 3 as input features, and use the long short-term memory network model to model the above input features, including order quantity prediction and warehousing frequency prediction based on time series, combine the prediction results with cargo attribute information, and generate a prediction data set for cargo space allocation and resource scheduling;

[0015] Step 5: Based on the forecast data set generated in step 4, combined with the current space status of the storage location and the volume and weight information of the goods, by establishing the goods-storage location allocation rules, the goods are allocated to the storage locations that meet the space and carrying conditions according to the goods attributes, and the occupancy status and remaining space of each storage location are recorded for subsequent resource scheduling and dynamic adjustment;

[0016] Step 6: Based on the result of the cargo space allocation in step 5 and the warehousing frequency and order quantity in the forecast data set, combined with the current status data of forklifts, automatic guided vehicles and warehouse operators, establish optimization rules for resource scheduling, and generate a scheduling plan including resource call sequence and allocation plan according to the principle of minimizing operation time and cost, and pass the scheduling plan to the execution module to control equipment operation and personnel operation;

[0017] Step 7: Combined with the scheduling plan executed in step 6, the real-time status and environmental parameters of the goods during the warehousing process are monitored, and the rule-based anomaly detection method is used to identify data deviations. Combined with the real-time images of the goods collected by the high-definition camera, the damage and omission of the goods are detected through image recognition technology, and early warning information is generated. The early warning information is compared with the scheduling plan and then transmitted to the feedback system to adjust the warehousing operation process;

[0018] Step 8. Based on the feedback information generated in step 7 and the resource scheduling plan in step 6, combined with the forecast data in step 4 and the cargo space allocation status in step 5, a real-time visual interface is constructed to display the warehousing progress, cargo space occupancy and exception handling results, and a dynamic adjustment interface is provided to achieve real-time scheduling optimization and cargo space allocation adjustment.

[0019] Preferably, in step 3, the interquartile range method is used to eliminate outliers, and the calculation formula is:

[0020] Outlier range = [Q1-1.5·IQR, Q3+1.5·IQR],

[0021] Among them, Q1 and Q3 are the first and third quartiles, and IQR is the interquartile range;

[0022] The IQR calculation formula is:

[0023] IQR=Q3-Q1,

[0024] The values ​​in the data that exceed the above range are considered as outliers and removed.

[0025] Preferably, the missing values ​​in step 3 are filled using the K nearest neighbor method, and the weight calculation formula is:

[0026] Among them, w i For sample x i The weight of

[0027] d(x i , x target ) is the sample x i With target missing samples x target The distance is calculated using the Euclidean distance formula:

[0028] Among them, m is the number of feature dimensions, x ij For sample x i The j-th eigenvalue of

[0029] x target，j is the target point x target The eigenvalue in the jth dimension,

[0030] j is the index of the feature, ranging from 1 to m.

[0031] Preferably, when the long short-term memory network model in step 4 predicts the time series, the input is the time series feature x t , the model structure includes input gate, forget gate, output gate and cell state update formula, where:

[0032] The input gate formula is: t =σ(W xi x t +W hi h t-1 +b i ),

[0033] Among them, W xi is the input x t AND input gate i t The weight matrix, W hi is the hidden state h at the previous moment t-1 AND input gate i t The weight matrix, b i is the bias term of the input gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function;

[0034] The forget gate formula is: t =σ(W xf x t +W hf h t-1 +b f ),

[0035] Among them, W xf is the input x t With the forget gate f t The weight matrix, W hf is the hidden state h at the previous moment t-1With the forget gate f t The weight matrix, b f is the bias term of the forget gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function;

[0036] The output gate formula is: t =σ(W xo x t +W ho h t-1 +b o ),

[0037] Among them, W xo It is input W xo AND output gate o t The weight matrix, W ho is the hidden state h at the previous moment t-1 AND output gate o t The weight matrix, b o The bias term of the output gate, σ is the Sigma activation function;

[0038] The cell state update formula is:

[0039] c t =f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ),

[0040] Among them, f t ⊙c t-1 is the forget gate for the unit state c at the previous moment t-1 The weighted update of

[0041] i t ⊙tanh(W xc x t +W hc h t-1 +b c )The input gate controls the new candidate state generated by combining the current input with the hidden state.

[0042] W xc is the input x t and the weight matrix of the cell state, x t is the hidden state h at the previous moment t-1 and the weight matrix of the cell state, b c is the bias term for updating the unit state, tanh is the activation function, and ⊙ represents the element-wise product.

[0043] Preferably, the prediction model in step 4 optimizes the loss function by minimizing the mean square error, and the formula is:

[0044] Among them, L represents the loss function, n is the number of samples, and y i is the actual value, is the predicted value and i is the index variable.

[0045] Preferably, the distribution rule of the goods in step 5 is based on the volume V of the goods. i , weight W i And the remaining space S j , the cargo space allocation must meet the following constraints:

[0046]

[0047] Among them, X ij is the decision variable for whether cargo i is allocated to cargo location j, n is the total number of cargoes, and m is the total number of cargo locations.

[0048] Preferably, the allocation priority of the goods in the cargo space allocation is based on the frequency of goods entering the warehouse F i The distance F between the cargo space and the warehouse entrance i Calculation, the priority calculation formula is:

[0049] Among them, P ij The priority of assigning item o to storage location j.

[0050] Preferably, in the resource scheduling plan in step 6, the scheduling priority of the forklift and the automatic guided vehicle is based on the operation time T k and resource cost C k Calculation, the optimization goal is to minimize the following objective function:

[0051]

[0052] Among them, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k.

[0053] Preferably, in the scheduling plan, resource scheduling is optimized by a genetic algorithm, and the fitness function of the genetic algorithm is:

[0054] Among them, P k is the penalty cost of resource k, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k.

[0055] Preferably, the rule-based anomaly detection in step 7 is combined with image recognition technology, wherein the image recognition uses the YOLOv5 model to detect the state of the goods, outputs the detection frame coordinates and category probabilities, and the anomaly detection range includes damaged goods, omissions and data deviations. The range of data deviations is determined by the following formula: x∈[μ-3σ,μ+3σ],

[0056] Among them, μ is the data mean and σ is the data standard deviation.

[0057] The present invention provides a logistics warehousing warehousing management method based on big data. It has the following beneficial effects:

[0058] 1. The present invention uses a long short-term memory network model to model historical order volume, warehousing frequency, cargo types and time series characteristics, accurately captures the time dependence and seasonal change trend of the data, and realizes accurate prediction of future warehousing demand, thereby improving the prediction accuracy and providing a reliable basis for cargo space allocation and resource scheduling.

[0059] 2. The present invention combines a mixed integer programming model to dynamically adjust the remaining space status of cargo spaces and establish optimal matching rules between goods and cargo spaces, thereby maximizing the utilization of storage space and dynamically optimizing cargo space allocation, thereby reducing space waste and improving storage management efficiency.

[0060] 3. The present invention constructs resource scheduling optimization rules, combines genetic algorithms to generate optimal scheduling plans, dynamically adjusts the calling order of forklifts, automatic guided vehicles and human resources, realizes efficient coordination of resource utilization and minimization of operation time and cost, and achieves the effect of reducing operating costs and improving equipment and manpower efficiency.

[0061] 4. The present invention combines image recognition technology with a rule engine to monitor the status of goods and the storage environment in real time, identify damage, omissions and data deviation anomalies of goods, and generate real-time warning information, so as to achieve rapid response and adjust the warehousing operation process, thereby improving the safety and stability of the warehousing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0063] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0064] The present invention is described in detail below in conjunction with the accompanying drawings:

[0065] Example:

[0066] Please see attached Figure 1 The embodiment of the present invention provides a logistics warehousing warehousing management method based on big data, including:

[0067] Step 1: Collect order data in the order management system through IoT devices, collect cargo information using RFID tags and barcode scanning devices, collect storage environment parameters using temperature and humidity sensors and light sensors, collect equipment status data using built-in sensors of forklifts and automatic guided vehicles, and collect real-time image data of cargo through high-definition cameras. The collected data includes timestamps, data source identifiers, and initial data values.

[0068] Step 2: The multi-source data collected in step 1, including order data, cargo information, equipment status, environmental parameters and image data, are transmitted using the MQTT protocol. The transmitted data is integrated through the Kafka message queue and sorted by timestamp. The integrated data is stored in the Hadoop distributed file system, and the data source information is annotated for subsequent cleaning and standardization processing;

[0069] Step 3: For the integrated data stored in step 2, the interquartile range method is used to eliminate outliers in the data, and the missing values ​​are filled by the K nearest neighbor method according to the data source information. The cleaned data is standardized according to the maximum and minimum value ranges, and the standardized data is updated to the Hadoop distributed file system to form a unified standardized data set for prediction and optimization;

[0070] Step 4: extract the order quantity, cargo type, warehousing frequency, time characteristics and storage environment parameters from the standardized data set formed in step 3 as input features, and use the long short-term memory network model to model the above input features, including order quantity prediction and warehousing frequency prediction based on time series, combine the prediction results with cargo attribute information, and generate a prediction data set for cargo space allocation and resource scheduling;

[0071] Step 5: Based on the forecast data set generated in step 4, combined with the current space status of the storage location and the volume and weight information of the goods, by establishing the goods-storage location allocation rules, the goods are allocated to the storage locations that meet the space and carrying conditions according to the goods attributes, and the occupancy status and remaining space of each storage location are recorded for subsequent resource scheduling and dynamic adjustment;

[0072] Step 6: Based on the result of the cargo space allocation in step 5 and the warehousing frequency and order quantity in the forecast data set, combined with the current status data of forklifts, automatic guided vehicles and warehouse operators, establish optimization rules for resource scheduling, and generate a scheduling plan including resource call sequence and allocation plan according to the principle of minimizing operation time and cost, and pass the scheduling plan to the execution module to control equipment operation and personnel operation;

[0073] Step 7: Combined with the scheduling plan executed in step 6, the real-time status and environmental parameters of the goods during the warehousing process are monitored, and the rule-based anomaly detection method is used to identify data deviations. Combined with the real-time images of the goods collected by the high-definition camera, the damage and omission of the goods are detected through image recognition technology, and early warning information is generated. The early warning information is compared with the scheduling plan and then transmitted to the feedback system to adjust the warehousing operation process;

[0074] Step 8. Based on the feedback information generated in step 7 and the resource scheduling plan in step 6, combined with the forecast data in step 4 and the cargo space allocation status in step 5, a real-time visual interface is constructed to display the warehousing progress, cargo space occupancy and exception handling results, and a dynamic adjustment interface is provided to achieve real-time scheduling optimization and cargo space allocation adjustment.

[0075] Step 1 Benefits: Collect multi-source data through IoT devices, RFID, barcode scanners, sensors and cameras to ensure real-time data of the entire process, including order information, cargo status, storage environment and equipment operation status. The data sources are comprehensive and time-synchronized, providing high-quality basic data for subsequent data analysis, prediction and optimization;

[0076] Benefits of step 2: Use MQTT protocol and Kafka message queue to achieve efficient and low-latency data transmission and integration, ensure the temporal consistency of data through timestamp sorting, and use Hadoop distributed file system to ensure high reliability and scalability of data storage. Label the data source information to provide traceability basis for subsequent data cleaning and analysis.

[0077] Benefits of step 3: Outliers are eliminated through the interquartile range method to improve data reliability; missing data are filled through the K nearest neighbor method to ensure data integrity; maximum-minimum value standardization is used to unify the scale of data to form a high-quality, standardized unified data set, providing ideal data input for model training and optimization;

[0078] Step 4 Benefits: Use long short-term memory networks to model historical data, achieve high-precision predictions of order volume and warehousing frequency, and generate a forecast data set based on cargo attribute information, which can predict future warehousing needs in advance, provide a scientific basis for cargo space allocation and resource scheduling, and effectively reduce resource waste and shortages caused by inaccurate predictions;

[0079] Step 5 Benefits: Combine forecast data and cargo attributes to dynamically optimize cargo-cargo location matching, allocate cargo locations based on space status, volume, and carrying capacity, record occupancy status and remaining space, significantly improve storage space utilization, and avoid space waste and cargo search difficulties caused by unreasonable cargo location arrangement;

[0080] Step 6 Benefits: Use resource status information and optimization rules to dynamically adjust the scheduling plans of forklifts, automatic guided vehicles, and manual labor, generate the optimal scheduling plan through genetic algorithms, achieve resource collaboration and minimize operation time and cost, improve operation efficiency, reduce operating costs, and avoid resource conflicts or idleness;

[0081] Step 7 Benefits: Based on the rule engine and image recognition technology, it can realize real-time monitoring of cargo status and environmental parameters, quickly detect abnormalities such as cargo damage, omissions and data deviations, generate early warning information and feedback to adjust the operation process, reduce the risk of abnormal spread, and improve the safety and stability of the warehousing system;

[0082] Benefits of Step 8: By building a real-time visual interface, the warehousing progress, cargo space occupancy, resource scheduling and exception handling information are clearly displayed, and a dynamic adjustment interface is provided to achieve management transparency and operational controllability. The real-time feedback mechanism facilitates managers to make quick decisions and improves system response speed and overall collaboration efficiency.

[0083] In summary, the various steps of the present invention are closely linked to form a full-process intelligent logistics warehousing and storage management system from data collection, processing, prediction to optimization decision-making and feedback. Through efficient data collection and cleaning, data quality is ensured, and with the help of prediction models and optimization algorithms, the scientificity and flexibility of cargo space allocation and resource scheduling are improved. Real-time monitoring and visual interfaces enhance the transparency and adaptability of the system. On the whole, this method significantly improves the efficiency and intelligence level of logistics warehousing and storage management, while reducing operating costs and abnormal risks, providing efficient, reliable and scalable solutions for the modern logistics industry.

[0084] In step 3, the interquartile range method is used to eliminate outliers. The calculation formula is:

[0085] Outlier range = [Q1-1.5·IQR, Q3+1.5·IQR],

[0086] Among them, Q1 and Q3 are the first and third quartiles, and IQR is the interquartile range;

[0087] The IQR calculation formula is:

[0088] IQR=Q3-Q1,

[0089] The values ​​in the data that exceed the above range are considered as outliers and removed.

[0090] By eliminating outliers through the interquartile range method, we can effectively detect and eliminate outliers in the data, avoid the interference of abnormal data on subsequent analysis and modeling processes, and thus improve the reliability and stability of model training and optimization results.

[0091] The interquartile range method does not require assumptions about data distribution and is applicable to different types of data sets. In particular, when the data is non-normally distributed or significantly skewed, it can accurately assess the range of data distribution and identify anomalies.

[0092] After removing outliers, the quality and consistency of the data are ensured, the deviations introduced by abnormal data in subsequent standardization processing, feature extraction and model training are reduced, and the accuracy of data analysis and prediction is improved.

[0093] The interquartile range method only requires the calculation of the median, quartiles and simple linear operations of the data. It has low computational complexity and is easy to implement efficiently in large-scale data processing processes. It is suitable for anomaly detection needs in big data scenarios.

[0094] In step 3, the missing values ​​are filled using the K nearest neighbor method, and the weight calculation formula is:

[0095]

[0096] Among them, w i For sample x i The weight of

[0097] d(x i , x target ) is the sample x i With target missing samples x target The distance is calculated using the Euclidean distance formula:

[0098] Among them, m is the number of feature dimensions, x ij For sample x i The j-th eigenvalue of

[0099] x target，j is the target point x target The eigenvalue in the jth dimension,

[0100] j is the index of the feature, ranging from 1 to m.

[0101] The K nearest neighbor method is used to fill in missing values. By referring to samples with similar characteristics to the target missing samples, the missing information can be compensated while maintaining the overall distribution characteristics of the data, effectively improving the integrity of the data and ensuring the continuity and consistency of the data.

[0102] By calculating the Euclidean distance between samples to measure similarity, the K nearest neighbor method can dynamically determine the k neighboring samples closest to the target sample, fill in missing values ​​based on weight distribution, ensure that the filling results are close to the characteristics of the target sample, and avoid the deviation introduced by simple interpolation methods.

[0103] The K-nearest neighbor method can handle missing value filling of multi-dimensional features, especially in high-dimensional feature space. It can help improve the accuracy of the filling results by comprehensively considering the similarities of features in each dimension and performing weighted filling. It is suitable for processing complex multi-source data.

[0104] The complete data set after filling can provide high-quality input for subsequent feature extraction, standardization and predictive modeling, reduce modeling errors caused by missing values, and improve the accuracy and robustness of the predictive model.

[0105] When the long short-term memory network model in step 4 predicts the time series, the input is the time series feature x t , the model structure includes input gate, forget gate, output gate and cell state update formula, where: input gate formula is: i t =σ(W xi x t +W hi h t-1 +b i ),

[0106] Among them, W xi is the input x t AND input gate i t The weight matrix, W hi is the hidden state h at the previous moment t-1 AND input gate i t The weight matrix, b i is the bias term of the input gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function;

[0107] The forget gate formula is: t =σ(W xf x t +W hf h t-1 +b f ),

[0108] Among them, W xf is the input x tWith the forget gate f t The weight matrix, W hf is the hidden state h at the previous moment t-1 With the forget gate f t The weight matrix, b f is the bias term of the forget gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function;

[0109] The output gate formula is: t =σ(W xo x t +W ho h t-1 +b o ),

[0110] Among them, W xo It is input W xo AND output gate o t The weight matrix, W ho is the hidden state h at the previous moment t-1 AND output gate o t The weight matrix, b o The bias term of the output gate, σ is the Sigma activation function;

[0111] The cell state update formula is:

[0112] c t =f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ),

[0113] Among them, f t ⊙c t-1 is the forget gate for the unit state c at the previous moment t-1 The weighted update of

[0114] i t ⊙tanh(W xc x t +W hc h t-1 +b c )The input gate controls the new candidate state generated by combining the current input with the hidden state.

[0115] W xc is the input x t and the weight matrix of the cell state, x t is the hidden state h at the previous moment t-1and the weight matrix of the cell state, b c is the bias term for updating the unit state, tanh is the activation function, and ⊙ represents the element-wise product.

[0116] The dynamic modeling of time series features is achieved through the input gate, forget gate, output gate and cell state update formula of the long short-term memory network model. The long short-term memory network can accurately capture the long-term dependence and short-term fluctuations of historical data, avoiding information forgetting and gradient disappearance problems. The stability and applicability of the model in nonlinear time series prediction provide strong support for scenarios such as order demand prediction and warehousing frequency prediction in logistics warehousing. Through the long short-term memory network model, the prediction results are more accurate and reliable, providing a scientific basis for subsequent cargo space allocation and resource scheduling optimization.

[0117] The prediction model in step 4 optimizes the loss function by minimizing the mean square error, which is:

[0118]

[0119] Among them, L represents the loss function, n is the number of samples, and y i is the actual value, is the predicted value and i is the index variable.

[0120] Minimizing the mean square error has significant advantages in amplifying the impact of samples with large prediction errors, uniformly handling positive and negative errors, and optimizing the smoothness of the process. It is particularly suitable for continuous variable prediction problems in logistics warehousing and storage management. The simple and efficient calculation method ensures the rapid convergence of the optimization process and the stability of the prediction results, laying a reliable foundation for the application of the model in order demand forecasting and storage frequency forecasting, and providing accurate input for scientific decision-making in subsequent cargo space allocation and resource scheduling.

[0121] The distribution rule of goods in step 5 is based on the volume V of the goods i , weight W i And the remaining space S j , the cargo space allocation must meet the following constraints:

[0122]

[0123]

[0124] Among them, X ij is the decision variable for whether cargo i is allocated to cargo location j, n is the total number of cargoes, and m is the total number of cargo locations;

[0125] The allocation priority of goods in the cargo space allocation is based on the frequency of goods entering the warehouse F i The distance F between the cargo space and the warehouse entrance i Calculation, the priority calculation formula is:

[0126] Among them, P ij The priority of assigning item o to storage location j.

[0127] High-frequency goods are allocated to the cargo space near the warehouse entrance first, optimizing the cargo handling path and operation efficiency. The allocation rules that comprehensively consider the cargo attributes and storage dynamic parameters make the solution flexible and adaptable, which can meet the needs of complex logistics and warehousing scenarios, significantly improve the efficiency, safety and resource utilization of cargo allocation, and provide a scientific solution for the intelligent storage and warehousing management.

[0128] In the resource scheduling plan in step 6, the scheduling priority of forklifts and AGVs is based on the operation time T k and resource cost C k Calculation, the optimization goal is to minimize the following objective function:

[0129]

[0130] Among them, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k;

[0131] In the scheduling plan, the resource scheduling is optimized by genetic algorithm, and the fitness function of the genetic algorithm is:

[0132] Among them, P k is the penalty cost of resource k, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k.

[0133] This method comprehensively considers resource operation time, cost and penalty cost, which can effectively reduce logistics warehousing operation costs and improve equipment utilization efficiency. At the same time, it adapts to dynamic scenario requirements, ensures scheduling flexibility and stability, and provides efficient and reliable resource scheduling strategies for complex logistics warehousing scenarios.

[0134] In step 7, rule-based anomaly detection is combined with image recognition technology. The image recognition uses the YOLOv5 model to detect the status of goods, output the detection box coordinates and category probability, and the anomaly detection range includes damaged goods, omissions, and data deviations. The range of data deviations is determined by the following formula:

[0135] x∈[μ-3σ,μ+3σ],

[0136] Among them, μ is the data mean and σ is the data standard deviation.

[0137] This method has both high precision and real-time performance, can quickly detect and feedback abnormal information, and dynamically adjust the operation process to reduce the impact of abnormal spread. The combination of rule-based statistical methods and image recognition technology makes anomaly detection more comprehensive and robust, significantly improves the safety and efficiency of logistics warehousing systems, and provides reliable support for the intelligent management of complex warehousing scenarios.

[0138] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A logistics warehousing and storage management method based on big data, characterized in that: include: Step 1: Collect order data in the order management system through IoT devices, collect cargo information using RFID tags and barcode scanning devices, collect storage environment parameters using temperature and humidity sensors and light sensors, collect equipment status data using built-in sensors of forklifts and automatic guided vehicles, and collect real-time image data of cargo through high-definition cameras. The collected data includes timestamps, data source identifiers, and initial data values. Step 2: The multi-source data collected in step 1, including order data, cargo information, equipment status, environmental parameters and image data, are transmitted using the MQTT protocol. The transmitted data is integrated through the Kafka message queue and sorted by timestamp. The integrated data is stored in the Hadoop distributed file system, and the data source information is annotated for subsequent cleaning and standardization processing; Step 3: For the integrated data stored in step 2, the interquartile range method is used to eliminate outliers in the data, and the missing values ​​are filled by the K nearest neighbor method according to the data source information. The cleaned data is standardized according to the maximum and minimum value ranges, and the standardized data is updated to the Hadoop distributed file system to form a unified standardized data set for prediction and optimization; Step 4: extract the order quantity, cargo type, warehousing frequency, time characteristics and storage environment parameters from the standardized data set formed in step 3 as input features, and use the long short-term memory network model to model the above input features, including order quantity prediction and warehousing frequency prediction based on time series, combine the prediction results with cargo attribute information, and generate a prediction data set for cargo space allocation and resource scheduling; Step 5: Based on the forecast data set generated in step 4, combined with the current space status of the storage location and the volume and weight information of the goods, by establishing the goods-storage location allocation rules, the goods are allocated to the storage locations that meet the space and carrying conditions according to the goods attributes, and the occupancy status and remaining space of each storage location are recorded for subsequent resource scheduling and dynamic adjustment; Step 6: Based on the result of the cargo space allocation in step 5 and the warehousing frequency and order quantity in the forecast data set, combined with the current status data of forklifts, automatic guided vehicles and warehouse operators, establish optimization rules for resource scheduling, and generate a scheduling plan including resource call sequence and allocation plan according to the principle of minimizing operation time and cost, and pass the scheduling plan to the execution module to control equipment operation and personnel operation; Step 7: Combined with the scheduling plan executed in step 6, the real-time status and environmental parameters of the goods during the warehousing process are monitored, and the rule-based anomaly detection method is used to identify data deviations. Combined with the real-time images of the goods collected by the high-definition camera, the damage and omission of the goods are detected through image recognition technology, and early warning information is generated. The early warning information is compared with the scheduling plan and then transmitted to the feedback system to adjust the warehousing operation process; Step 8. Based on the feedback information generated in step 7 and the resource scheduling plan in step 6, combined with the forecast data in step 4 and the cargo space allocation status in step 5, a real-time visual interface is constructed to display the warehousing progress, cargo space occupancy and exception handling results, and a dynamic adjustment interface is provided to achieve real-time scheduling optimization and cargo space allocation adjustment.

2. According to a big data-based logistics warehousing management method according to claim 1, it is characterized in that: In step 3, the interquartile range method is used to eliminate outliers, and the calculation formula is: Outlier range = [Q1-1.5·IQR, Q3+1.5·IQR], Among them, Q1 and Q3 are the first and third quartiles, and IQR is the interquartile range; The IQR calculation formula is: IQR=Q3-Q1, The values ​​in the data that exceed the above range are considered as outliers and removed.

3. According to a method for logistics warehousing and storage management based on big data according to claim 1, it is characterized in that: In step 3, the missing values ​​are filled using the K nearest neighbor method, and the weight calculation formula is: Among them, w i For sample x i The weight of d(x i , x target ) is the sample x i With target missing samples x target The distance is calculated using the Euclidean distance formula: Among them, m is the number of feature dimensions, x ij For sample x i The j-th eigenvalue of x target，j is the target point x target The eigenvalue in the jth dimension, j is the index of the feature, ranging from 1 to m.

4. According to a method for logistics warehousing and storage management based on big data as claimed in claim 1, it is characterized in that: When the long short-term memory network model in step 4 predicts the time series, the input is the time series feature x t , the model structure includes input gate, forget gate, output gate and cell state update formula, where: The input gate formula is: t =σ(W xi x t +W hi h t-1 +b i ), Among them, W xi is the input x t AND input gate i t The weight matrix, W hi is the hidden state h at the previous moment t-1 AND input gate i t The weight matrix, b i is the bias term of the input gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function; The forget gate formula is: t =σ(W xf x t +W hf h t-1 +b f ), Among them, W xf is the input x t With the forget gate f t The weight matrix, W hf is the hidden state h at the previous moment t-1 With the forget gate f t The weight matrix, b f is the bias term of the forget gate, h t-1 is the hidden state of the previous time step, σ is the Sigmoid activation function; The output gate formula is: t =σ(W xo x t +W ho h t-1 +b o ), Among them, W xo It is input W xo AND output gate o t The weight matrix, W ho is the hidden state h at the previous moment t-1 AND output gate o t The weight matrix, b o The bias term of the output gate, σ is the Sigmoid activation function; The cell state update formula is: c t =f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ), Among them, f t ⊙c t-1 is the forget gate for the unit state c at the previous moment t-1 The weighted update of i t ⊙tanh(W xc x t +W hc h t-1 +b c )The input gate controls the new candidate state generated by combining the current input with the hidden state. W xc is the input x t and the weight matrix of the cell state, x t is the hidden state h at the previous moment t-1 and the weight matrix of the cell state, b c is the bias term for updating the unit state, tanh is the activation function, and ⊙ represents the element-wise product.

5. The method for logistics warehousing and storage management based on big data according to claim 1 is characterized in that: The prediction model in step 4 optimizes the loss function by minimizing the mean square error, and the formula is: Among them, L represents the loss function, n is the number of samples, and y i is the actual value, is the predicted value and i is the index variable.

6. The method for logistics warehousing and storage management based on big data according to claim 1 is characterized in that: The distribution rule of the goods in step 5 is based on the volume V of the goods. i , weight W i And the remaining space S j , the cargo space allocation must meet the following constraints: Among them, X ij is the decision variable for whether cargo i is allocated to cargo location j, n is the total number of cargoes, and m is the total number of cargo locations.

7. A logistics warehousing and storage management method based on big data according to claim 6, characterized in that: The allocation priority of goods in the cargo space allocation is based on the frequency of goods entering the warehouse F i The distance F between the cargo space and the warehouse entrance i Calculation, the priority calculation formula is: Among them, P ij The priority of assigning item o to storage location j.

8. The method for logistics warehousing and storage management based on big data according to claim 1 is characterized in that: In the resource scheduling plan in step 6, the scheduling priority of forklifts and automated guided vehicles is based on the operation time T k and resource cost C k Calculation, the optimization goal is to minimize the following objective function: Among them, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k.

9. A logistics warehousing and storage management method based on big data according to claim 8, characterized in that: In the scheduling plan, the resource scheduling is optimized by genetic algorithm, and the fitness function of the genetic algorithm is: Among them, P k is the penalty cost of resource k, K is the total number of resources, T k is the operation time of resource k, C k is the usage cost of resource k.

10. The method for logistics warehousing and storage management based on big data according to claim 1, characterized in that: In step 7, rule-based anomaly detection is combined with image recognition technology, wherein image recognition uses the YOLOv5 model to detect the state of goods, outputs the detection frame coordinates and category probabilities, and the anomaly detection range includes damaged goods, omissions, and data deviations. The range of data deviations is determined by the following formula: x∈[μ-3σ,μ+3σ], Among them, μ is the data mean and σ is the data standard deviation.

Images