A cold chain warehousing and inventory management system based on big data analytics

Abstract

This invention relates to the field of cold chain warehousing inventory management technology, and more particularly to a cold chain warehousing inventory management system based on big data analysis. The system includes: a correlation factor determination module, used to classify the set of energy consumption-related parameters of cold storage into uncorrelated parameter sets and correlated parameter sets using a correlation degree model, wherein the correlation degree model is constructed based on the Apriori algorithm; a data imputation module, used to generate imputation parameters from the uncorrelated parameter sets using Lagrange interpolation; an energy consumption prediction module, used to generate predicted cold storage energy consumption from the imputation parameters and correlated parameter sets using an energy consumption prediction model, wherein the energy consumption prediction model is constructed based on an architecture that fuses a temporal convolutional network with gated recurrent units and an attention mechanism; and a cold chain warehousing inventory management module, used to manage the cold chain warehousing inventory according to the predicted cold storage energy consumption. This invention achieves intelligent cold chain inventory management based on predicted cold storage energy consumption.

Description

Technical Field

[0001] This invention relates to the field of cold chain warehousing and inventory management technology, and in particular to a cold chain warehousing and inventory management system based on big data analysis. Background Technology

[0002] Cold chain warehouses, as a core link in the processing, storage, and transportation of cold chain food, are playing an increasingly prominent role. According to data released by the Cold Chain Alliance, cold storage capacity is showing a steady growth trend, which is expected to continue for some time to come.

[0003] However, the rapid development of the cold chain industry is also accompanied by challenges such as high energy consumption and high costs. Therefore, in-depth research into the energy consumption of cold chain warehouses and strengthening energy-efficient inventory management are of vital importance to improving economic efficiency.

[0004] Currently, an analysis of the energy consumption of a 30,000-ton cold chain warehouse reveals that the refrigeration system accounts for 83.3% of the total energy consumption. Of this, fans account for 13.3%, cooling energy accounts for 70.0%, lighting energy accounts for 2.3%, and other energy consumption accounts for 14.4%. Therefore, it is evident that the rational utilization of the refrigeration system's energy consumption is crucial for improving energy efficiency in cold chain warehouse inventory management.

[0005] Therefore, how to accurately predict the energy consumption of cold chain warehousing in order to realize the transformation of cold chain warehousing from passive refrigeration inventory management to intelligent cold control inventory management based on predicted energy consumption is a technical problem that needs to be solved. Summary of the Invention

[0006] To address this, the present invention provides a cold chain warehousing and inventory management system based on big data analysis. The system uses the Apriori algorithm to determine the real-time energy consumption-related parameter set of cold storage, and then completes the parameters that are not related to cold storage energy consumption and inputs them into an energy consumption prediction model based on a temporal convolutional fusion bidirectional memory network to generate a more accurate prediction of cold storage energy consumption, thereby realizing intelligent cold storage and inventory management based on energy consumption prediction.

[0007] To achieve the above objectives, this invention proposes a cold chain warehousing and inventory management system based on big data analysis, comprising:

[0008] The correlation factor determination module is used to classify the set of energy consumption related parameters of cold storage into a set of unrelated parameters and a set of related parameters through a correlation degree model. The correlation degree model is constructed based on an improved Apriori algorithm with a set dynamic threshold, which is dynamically adjusted according to the change in the state of cold storage and / or the feature value of the correlation rule set.

[0009] A data imputation module, which is connected to the correlation factor determination module, is used to generate imputation parameters from the uncorrelated parameter set using the Lagrange interpolation method;

[0010] An energy consumption prediction module, connected to the correlation factor determination module and the data imputation module, is used to generate a predicted cold storage energy consumption by using the imputation parameters and the correlation parameter set through an energy consumption prediction model. The energy consumption prediction model is constructed based on an architecture that integrates a temporal convolutional network with a gated recurrent unit and an attention mechanism. The temporal convolutional network is used to generate a temporal parameter sequence based on the imputation parameters and the correlation parameter set. The gated recurrent unit is used to generate the predicted cold storage energy consumption as a hidden state based on the temporal parameter sequence. The attention mechanism is used to control the updating of the hidden state by the gated recurrent unit.

[0011] The cold chain warehouse inventory management module is connected to the energy consumption prediction module and is used to manage the inventory optimization, cargo storage location and inbound / outbound strategies of the cold chain warehouse based on the predicted cold storage energy consumption.

[0012] Furthermore, the temporal convolutional network has at least two residual blocks, each residual block comprising an extended causal convolutional layer, a weight normalization layer, a LeakyReLU activation function layer, and a regularization layer;

[0013] The gated loop unit is used to generate the predicted cold storage energy consumption as a hidden state based on the time series parameter sequence.

[0014] The attention mechanism is used to control the updating of the hidden state by the gated loop unit.

[0015] Furthermore, the hidden state includes the previous hidden state, the current hidden state, and the transitive hidden state, and the attention mechanism includes a weight generation subunit and a transitive subunit;

[0016] The weight generation subunit is used to generate attention weights based on the current hidden state and the target vector through an attention mechanism.

[0017] The transmission subunit is used to generate the transmission hidden state based on the attention weight and the previous hidden state, and to transmit the transmission hidden state to the gated loop unit.

[0018] Furthermore, the energy consumption prediction module also includes an objective function unit;

[0019] The objective function unit is used to control the optimization prediction of the gated recurrent unit through an objective function, wherein the objective function is constructed based on cross-entropy and activation function.

[0020] In the above scheme, the similarity between the user's current hidden state and the target item is calculated through an attention mechanism to update the hidden state of the gated loop unit. Combined with the potential temporal parameter sequence features, it can effectively improve the performance and accuracy of cold storage energy consumption prediction tasks.

[0021] Furthermore, the correlation factor determination module includes a preprocessing unit, a correlation rule generation unit, and a judgment unit;

[0022] The preprocessing unit is used to process the set of energy consumption-related parameters of the cold storage and generate an interval representation.

[0023] The association rule generation unit is used to iterate the interval representation to generate an association rule set associated with the comprehensive energy consumption of the cold storage.

[0024] The judgment unit is used to classify the set of energy consumption related parameters of the cold storage according to the set of association rules to generate the set of unrelated parameters and the set of associated parameters.

[0025] Furthermore, the dynamic threshold includes a support threshold and a confidence threshold, the cold storage status change includes the change in external temperature of the cold storage and the change in cargo storage, the feature value of the association rule set is the average lift of the association rule set, and the judgment unit includes a support dynamic adjustment subunit and a confidence dynamic adjustment subunit.

[0026] The support dynamic adjustment subunit is used to set the support threshold according to the change in external temperature of the cold storage and the change in cargo storage.

[0027] The confidence dynamic adjustment subunit is used to set the confidence threshold based on the average lift of the association rule set.

[0028] Furthermore, the preprocessing unit includes a data cleaning subunit, a standardization subunit, and a clustering and grouping subunit;

[0029] The data cleaning subunit is used to clean the set of cold storage energy consumption related parameters and generate a cleaned cold storage energy consumption related dataset.

[0030] The standardization subunit is used to standardize the energy consumption related dataset of the cold storage cleaning to generate a standard cold storage energy consumption related dataset.

[0031] The clustering sub-unit is used to discretize the standard cold storage energy consumption related dataset using a clustering algorithm to generate the interval representation;

[0032] The set of energy-related parameters for cold storage includes internal and external environmental parameters of cold chain storage, switching parameters of cold chain storage, inventory parameters of cold chain storage, and refrigeration equipment parameters of cold chain storage.

[0033] Furthermore, the data filling module includes a historical dataset update unit and a filling unit;

[0034] The historical dataset update unit is used to update the historical dataset of the cold storage based on the operational stability of the cold storage.

[0035] The filling unit is used to determine the corresponding parameters in the cold storage historical dataset based on the unrelated parameter set, and to generate filling parameters by using the Lagrange interpolation method for the corresponding parameters.

[0036] The above scheme overcomes the problem that the correlation between various parameters in the set of energy consumption parameters for cold storage is not clear and is prone to change over time. It then achieves accurate mining of the current correlation and realizes dynamic and accurate correlation judgment of the set of energy consumption parameters for cold storage. As a result, the data input into the energy consumption prediction model of the temporal convolutional fusion bidirectional memory network has a high correlation and can accurately generate predicted cold storage energy consumption.

[0037] Furthermore, the energy consumption parameters of the cold storage include the category of inventory goods, and the cold chain warehousing inventory management module also includes an inventory optimization unit and a goods storage location planning unit;

[0038] The inventory optimization unit is used to adjust the regional inventory limit of cold chain storage based on the predicted cold storage energy consumption.

[0039] The cargo storage location planning unit is used to determine the recommended storage area based on the predicted cold storage energy consumption and the category of inventory goods.

[0040] Furthermore, the cold chain warehousing inventory management module includes an inbound / outbound strategy unit;

[0041] The inbound / outbound strategy unit is used to select the time period during which the predicted energy consumption of the cold storage is in the low power consumption range as the recommended inbound / outbound time period.

[0042] The above solution enables dynamic management of cold chain warehousing inventory limits, goods storage locations, and inbound / outbound strategies by predicting cold storage energy consumption.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] The Apriori algorithm is used to determine the set of parameters related to cold storage energy consumption in real time. Then, by supplementing parameters that are not related to cold storage energy consumption and inputting them into an energy consumption prediction model based on a temporal convolutional fusion bidirectional memory network, a more accurate prediction of cold storage energy consumption is generated, thereby realizing intelligent cold storage inventory management based on energy consumption prediction.

[0045] By using an attention mechanism to calculate the similarity between the user's current hidden state and the target item to update the hidden state of the gated loop unit, and combining the potential temporal parameter sequence features, the performance and accuracy of cold storage energy consumption prediction tasks can be effectively improved.

[0046] It overcomes the problem that the correlation between multiple parameters in the set of energy consumption parameters for cold storage is not clear and is prone to change over time. It then realizes the accurate mining of the current correlation and achieves dynamic and accurate correlation judgment of the set of energy consumption parameters for cold storage. As a result, the data input into the energy consumption prediction model of the temporal convolutional fusion bidirectional memory network has a high correlation and can accurately generate predicted cold storage energy consumption.

[0047] It enables dynamic management of cold chain warehouse inventory limits, goods storage locations, and inbound / outbound strategies by predicting cold storage energy consumption. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the structure of a cold chain warehousing and inventory management system based on big data analysis according to an embodiment of the present invention;

[0049] Figure 2 This is a flowchart illustrating the cold chain warehousing and inventory management system based on big data analysis, according to an embodiment of the present invention.

[0050] Figure 3 This is a schematic diagram of the residual block structure of the energy consumption prediction model of the cold chain warehouse inventory management system based on big data analysis, according to an embodiment of the present invention.

[0051] Figure 4 This is a detailed structural diagram of the bidirectional long short-term memory network unit and attention mechanism of the energy consumption prediction model of the cold chain warehousing and inventory management system based on big data analysis, according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0053] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0054] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0055] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0056] like Figures 1 to 4 As shown, this invention provides a cold chain warehousing and inventory management system based on big data analysis. It uses the Apriori algorithm to determine the set of parameters related to cold storage energy consumption in real time. Then, by supplementing parameters that are not related to cold storage energy consumption and inputting them into an energy consumption prediction model based on a temporal convolutional fusion bidirectional memory network, a more accurate prediction of cold storage energy consumption is generated, thereby realizing intelligent cold storage inventory management based on energy consumption prediction.

[0057] like Figures 1 to 4 As shown, this embodiment proposes a cold chain warehousing and inventory management system based on big data analysis, including:

[0058] The correlation factor determination module is used to classify the set of energy consumption related parameters of cold storage into a set of unrelated parameters and a set of related parameters through a correlation degree model. The correlation degree model is constructed based on an improved Apriori algorithm with a set dynamic threshold, which is dynamically adjusted according to the change in the state of cold storage and / or the feature value of the correlation rule set.

[0059] A data imputation module, which is connected to the correlation factor determination module, is used to generate imputation parameters from the uncorrelated parameter set using the Lagrange interpolation method;

[0060] An energy consumption prediction module, connected to the correlation factor determination module and the data imputation module, is used to generate a predicted cold storage energy consumption by using the imputation parameters and the correlation parameter set through an energy consumption prediction model. The energy consumption prediction model is constructed based on an architecture that integrates a temporal convolutional network with a gated recurrent unit and an attention mechanism. The temporal convolutional network is used to generate a temporal parameter sequence based on the imputation parameters and the correlation parameter set. The gated recurrent unit is used to generate the predicted cold storage energy consumption as a hidden state based on the temporal parameter sequence. The attention mechanism is used to control the updating of the hidden state by the gated recurrent unit.

[0061] The cold chain warehouse inventory management module is connected to the energy consumption prediction module and is used to manage the inventory optimization, cargo storage location and inbound / outbound strategies of the cold chain warehouse based on the predicted cold storage energy consumption.

[0062] Understandably, cold storage energy consumption prediction based on TCN-GRU (Temporal Convolutional Fusion Bidirectional Memory Network) uses the energy consumption of cold storage in the previous period and other related parameters to predict the energy consumption of cold storage in the current period. This requires a strong correlation / causal relationship between the input and the predicted values. However, this correlation may change over time. For example, when the outdoor temperature is low, the correlation between the average opening time and inventory of cold storage and the energy consumption of cold storage is weak. Therefore, these values ​​should be screened out and replaced with data using Lagrange interpolation to ensure that the causal relationship between the input and output variables of the model is clear and the control strategy for cold chain warehousing inventory management is accurate.

[0063] Specifically, the Lagrange interpolation method fills in and replaces the unrelated parameter set based on the closest normal historical data.

[0064] Furthermore, such as Figure 3 As shown, the temporal convolutional network has at least two residual blocks, each residual block comprising an extended causal convolutional layer, a weight normalization layer, a LeakyReLU activation function layer, and a regularization layer.

[0065] Specifically, such as Figure 3 As shown, the Temporal Convolutional Network (TCN) unit includes: causal convolution to ensure the causality of the temporal parameter sequence, employing a one-dimensional convolutional layer with a kernel size of 9, where each output point is derived from the input of the previous 9 time steps; dilated convolution to expand the receptive field through skip-interval sampling and capture long-period features; and residual connections, preferably 4, to alleviate the gradient vanishing problem and accelerate model convergence. Therefore, the Temporal Convolutional Network (TCN) unit is suitable for the periodic prediction of energy consumption in different cold storage facilities.

[0066] Further, see Figure 4The hidden state includes the previous hidden state, the current hidden state, and the transitive hidden state; the attention mechanism includes a weight generation subunit and a transitive subunit.

[0067] The weight generation subunit is used to generate attention weights based on the current hidden state and the target vector through an attention mechanism.

[0068] The transmission subunit is used to generate the transmission hidden state based on the attention weight and the previous hidden state, and to transmit the transmission hidden state to the gated loop unit.

[0069] See Figure 4 The weight generation subunit is specifically:

[0070]

[0071] In the formula, This represents the output value of the attention mechanism. The softmax activation function controls the target vector at time t. Should it be updated? Let represent the current hidden state, and W represent the attention weight.

[0072] See Figure 4 The specific transmission subunit is:

[0073]

[0074] In the formula, These represent the previous hidden state and the transitive hidden state, respectively. Let W represent the input embedding at time t, where ReLU is the ReLU activation function, and W represents the attention weights. For learnable weight vectors, All of these are learnable bias terms.

[0075] See Figure 4 The process of generating the next hidden state for the next time step based on the passed hidden state and the output value of the attention mechanism is as follows:

[0076]

[0077] In the formula, These represent the next hidden state, the previous hidden state, and the passed hidden state, respectively. This represents the output value of the attention mechanism.

[0078] Specifically, see Figure 4The gated recurrent unit (GRU) in the energy consumption prediction model is an improved recurrent neural network structure. Its core design uses update gates and reset gates to model and predict the dynamic features of time series. The expression of the gated recurrent unit is as follows:

[0079]

[0080] In the formula, These represent the current hidden state, the previous hidden state, and the passed hidden state, respectively. Let represent the update gate, the input embedding, and the reset gate at time t, respectively. ELU is the ELU (Exponential Linear Unit) activation function. All of these are learnable weight vectors. All are learnable bias terms. Specifically, the previous hidden state and the passed hidden state are updated according to the update gate, and the sum of these is used to obtain the current hidden state. The value of the transitive hidden state is updated through the transitive subunit. The reset gate is used to determine whether the input embedding and the previous hidden state need to be updated to generate the transitive hidden state. Specifically, it will be as follows: Figure 3 The output of the temporal convolutional network (TCN) shown is used as Figure 4 The input is embedded into the gated recurrent unit (GRU), and then into the gated recurrent unit and the attention mechanism.

[0081] Furthermore, the energy consumption prediction module also includes an objective function unit;

[0082] The objective function unit is used to control the optimization prediction of the gated recurrent unit through an objective function, wherein the objective function is constructed based on cross-entropy, specifically as follows:

[0083]

[0084] In the formula, L is the objective function, y represents the label value, p represents the predicted value, F is the output of the weight generation sub-unit, W is the learnable weight vector, and b is the learnable bias term.

[0085] In the above scheme, the similarity between the user's current hidden state and the target item is calculated through an attention mechanism to update the hidden state of the gated loop unit. Combined with the potential temporal parameter sequence features, it can effectively improve the performance and accuracy of cold storage energy consumption prediction tasks.

[0086] Furthermore, such as Figure 2 As shown, the correlation factor determination module includes a preprocessing unit, a correlation rule generation unit, and a judgment unit;

[0087] The preprocessing unit is used to process the set of energy consumption-related parameters of the cold storage and generate an interval representation.

[0088] The association rule generation unit is used to iterate the interval representation to generate an association rule set associated with the comprehensive energy consumption of the cold storage.

[0089] The judgment unit is used to classify the set of energy consumption related parameters of the cold storage according to the set of association rules to generate the set of unrelated parameters and the set of associated parameters.

[0090] Specifically, the Apriori algorithm is scalable and performs well, particularly suitable for single-dimensional attributes. Its process for mining association rules is as follows: Iteration: Find all itemsets in the transaction database that are greater than or equal to the user-specified minimum support threshold, i.e., frequent itemsets; Using these frequent itemsets, mine strong association rules that meet the user's needs, i.e., the set of association rules whose support and confidence are greater than or equal to the corresponding thresholds. Let D be the set of data transactions, and A and B be itemsets, where A is an interval representation and B is the comprehensive energy consumption of the cold storage. The association rule set contains multiple association rules. An association rule can be represented as follows: if itemset A appears, then itemset B may also appear, and it has support s and confidence c in the transaction set D. Here, support s refers to the percentage of itemsets A and B contained in transaction set D, while confidence c is the percentage of itemsets A and B contained in transaction set D.

[0091] The itemset A, representing the intervals, includes the level values ​​of multiple parameters in the cold storage energy consumption-related parameter set, as well as combinations of these level values. For example, itemset A includes temperature intervals one to three within the cold chain storage, humidity intervals one to three within the cold chain storage, and temperature intervals one to three outside the cold chain storage, as well as combinations of these elements, such as temperature interval one within the cold chain storage + humidity interval two within the cold chain storage + temperature interval one outside the cold chain storage. The interval ranges are, for example, temperature interval one within the cold chain storage being a low-temperature range [-19.5, -19.2] degrees Celsius, temperature interval two within the cold chain storage being a medium-temperature range (-19.2, -18.8] degrees Celsius, and temperature interval three within the cold chain storage being a high-temperature range (-18.8, -18.4] degrees Celsius.

[0092] It should be noted that the comprehensive energy consumption of the cold storage is the sum of refrigeration power consumption (REC) and direct power consumption (DEC). Refrigeration power consumption includes the power consumption of core equipment such as compressors and condensers, while direct power consumption includes the power consumption of auxiliary equipment such as lighting, air coolers, defrosting, and air curtains.

[0093] Furthermore, such as Figure 2As shown, the dynamic threshold includes a support threshold and a confidence threshold; the cold storage status change includes changes in the external temperature of the cold storage and changes in cargo storage; the feature value of the association rule set is the average lift of the association rule set; and the judgment unit includes a support dynamic adjustment subunit and a confidence dynamic adjustment subunit.

[0094] The support dynamic adjustment subunit is used to set the support threshold according to the change in external temperature of the cold storage and the change in cargo storage.

[0095] The confidence dynamic adjustment subunit is used to set the confidence threshold based on the average lift of the association rule set.

[0096] Specifically, the support dynamic adjustment subunit is represented as follows:

[0097]

[0098] In the formula, Indicates the support threshold. These represent the temperature change value and the temperature change threshold, respectively, and S represents the amount of goods stored.

[0099] Specifically, the average lift is:

[0100]

[0101] In the formula, The average lifting degree of the association rule set. The process of improving the degree of support is as follows: X represents the interval, Y represents the comprehensive energy consumption of the cold storage, Support represents the degree of support, and N is the number of rules in the rule set.

[0102] Understandably, the support (A→B) reflects the co-occurrence frequency of itemset A representing the interval and itemset B representing the comprehensive energy consumption of the cold storage. Itemets with low support may be occasional but critical anomalous patterns, such as a surge in defrosting energy consumption due to low temperature ranges or high humidity. Therefore, when the average lift is greater than the high lift threshold, the confidence threshold is increased by a first predetermined percentage to avoid misjudging weak associations due to the prevalence of high-frequency combinations; when the average lift is greater than the high lift threshold, the confidence threshold is decreased by a second predetermined percentage to retain low-frequency but high-value anomalous association rules. Preferably, the high support threshold is 20%, the low support threshold is 1.8%, the first predetermined percentage is 5%, and the second predetermined percentage is 3%.

[0103] Furthermore, the judgment unit is used to set parameters that exist at least twice in the association rule set as the association parameter set, otherwise they are irrelevant parameters. For example, if the internal and external environmental parameters of the cold chain warehouse simultaneously satisfy the following rules: the range of the internal and external environmental parameters of the cold chain warehouse represents from one to the comprehensive energy consumption of the cold storage, the range of the internal and external environmental parameters of the cold chain warehouse represents two, and the humidity range of the cold chain warehouse represents from one to the comprehensive energy consumption of the cold storage, then the internal and external environmental parameters of the cold chain warehouse are set as the association parameter set.

[0104] Furthermore, the preprocessing unit includes a data cleaning subunit, a standardization subunit, and a clustering and grouping subunit;

[0105] The data cleaning subunit is used to clean the set of cold storage energy consumption related parameters and generate a cleaned cold storage energy consumption related dataset.

[0106] The standardization subunit is used to standardize the energy consumption related dataset of the cold storage cleaning to generate a standard cold storage energy consumption related dataset.

[0107] The clustering sub-unit is used to discretize the standard cold storage energy consumption related dataset using a clustering algorithm to generate the interval representation;

[0108] The set of energy-related parameters for cold storage includes internal and external environmental parameters of cold chain storage, switching parameters of cold chain storage, inventory parameters of cold chain storage, and refrigeration equipment parameters of cold chain storage. The interval represents the level of each parameter in the set of energy-related parameters for cold storage.

[0109] Specifically, the internal and external environmental parameters include the temperature inside the cold chain warehouse, the humidity inside the cold chain warehouse, the temperature outside the cold chain warehouse, and the humidity outside the cold chain warehouse. The cold chain warehouse switch parameters are the average duration of door opening and closing on the current day. The cold chain warehouse inventory parameters include the frequency of goods entering and leaving the warehouse, the number of times the cold storage door is opened and closed, and the inventory quantity. The cold chain warehouse refrigeration equipment parameters include the compressor running time, refrigerant pressure, and equipment aging coefficient.

[0110] Specifically, data cleaning involves removing data that is significantly distorted in terms of energy consumption parameters related to the cold storage, such as occasional compressor readings showing pump pressure or dry pressure close to 0. Since the original data volume is large and this type of data constitutes a small proportion with minimal impact, it is therefore removed.

[0111] Specifically, the standardization process involves standardizing the maximum and minimum values ​​to eliminate dimensional differences in the data.

[0112] Specifically, the clustering algorithm is the K-means clustering algorithm, and the number of clusters is set to 3 levels.

[0113] Furthermore, the data filling module includes a historical dataset update unit and a filling unit;

[0114] The historical dataset update unit is used to update the historical dataset of the cold storage based on the operational stability of the cold storage.

[0115] The filling unit is used to determine the corresponding parameters in the cold storage historical dataset based on the unrelated parameter set, and to generate filling parameters by using the Lagrange interpolation method for the corresponding parameters.

[0116] Specifically, the historical dataset update unit is used to calculate the statistical correlation coefficient between historical data and the comprehensive energy consumption of the cold storage, and stores historical data with correlation coefficients greater than a lower limit for periodically updating the cold storage historical dataset. It is understandable that using only the statistical correlation coefficient to determine the correlation between data would result in insufficient data input to the energy consumption prediction model, making accurate energy consumption prediction impossible. Therefore, the reliability of the supplemented data is ensured solely through the statistical correlation coefficient.

[0117] Specifically, the filling unit generates filling parameters by interpolating multiple corresponding parameters.

[0118] The above solution overcomes the problem that the correlation between various parameters in the cold storage energy consumption parameter set is not clear and is prone to change over time. It then achieves accurate mining of the current correlation and realizes dynamic and accurate correlation judgment of the cold storage energy consumption parameter set. As a result, the data input into the energy consumption prediction model constructed by TCN and GRU has a high correlation and can accurately generate predicted cold storage energy consumption.

[0119] Furthermore, the energy consumption parameters of the cold storage include the category of inventory goods, and the cold chain warehousing inventory management module also includes an inventory optimization unit and a goods storage location planning unit;

[0120] The inventory optimization unit is used to adjust the regional inventory limit of cold chain storage based on the predicted cold storage energy consumption.

[0121] The cargo storage location planning unit is used to determine the recommended storage area based on the predicted cold storage energy consumption and the category of inventory goods.

[0122] Furthermore, the cold chain warehousing inventory management module includes an inbound / outbound strategy unit;

[0123] The inbound / outbound strategy unit is used to select the time period during which the predicted energy consumption of the cold storage is in the low power consumption range as the recommended inbound / outbound time period.

[0124] Specifically, the inventory optimization unit calculates the inventory capacity mapping based on the energy consumption forecast for the next 24 hours. When the capacity mapping exceeds the safety threshold, it triggers a reduction in the regional inventory limit.

[0125] Specifically, the cargo storage location planning unit calculates the matching degree with candidate areas based on the cargo's category (frozen / refrigerated), temperature sensitivity (e.g., storage temperature ±2℃), and shelf life:

[0126]

[0127] In the formula, Score represents the matching degree. These represent the ideal storage temperature and the current actual temperature of cargo i, respectively. Indicates the upper limit of storage temperature. These represent predicted energy consumption and reference energy consumption, respectively.

[0128] The cargo storage location planning unit prioritizes the allocation of high-value goods (such as medicines and seafood) to the core area with stable temperature control; ordinary goods are dynamically adjusted to the peripheral area and a flexible inventory limit is set; cross-contamination of odors is avoided, such as separating seafood and dairy products, and high energy consumption periods are dynamically avoided: if the predicted energy consumption of a certain area exceeds the limit in the next 2 hours, the allocation of new goods is suspended.

[0129] The inbound / outbound strategy unit is used to determine recommended low-energy-consumption periods based on the predicted hourly energy consumption values ​​for the next 24 hours.

[0130] This embodiment uses IoT sensors (temperature and humidity, number of access control openings and closings) to collect data in real time and integrates it into the cold chain warehousing business system, such as WMS (warehouse management system) or ERP (enterprise resource planning).

[0131] Specifically, after format conversion of the data, the Apriori function was called, setting the initial support threshold to 7.5% and the initial confidence threshold to 40%, to analyze the typical door opening frequency of cold storage facilities. This revealed the relationship between weekday type and cold storage temperature, laying the foundation for linking cold storage door opening behavior with energy consumption analysis. Cold storage door opening behavior determines the operating mode of the air cooler, thus affecting the energy consumption level of the cold storage to a certain extent. To establish a reasonable energy consumption prediction model, this paper combines the door opening behavior of the cold storage facility over a period of time, the operating status of the air cooler, and the temperature inside the cold storage. Continuous door opening for 140 seconds with the air cooler off resulted in a 0.64℃ increase in cold storage temperature; single door openings of 10 to 20 seconds resulted in temperature fluctuations of no more than 0.1℃. In fact, if the cold storage facility frequently opens its doors briefly over a period of time, the temperature fluctuations will increase accordingly.

[0132] The above solution enables dynamic management of cold chain warehousing inventory limits, goods storage locations, and inbound / outbound strategies by predicting cold storage energy consumption.

[0133] In this embodiment, the Apriori algorithm is used to determine the real-time cold storage energy consumption-related parameter set. Then, parameters unrelated to cold storage energy consumption are supplemented and input into an energy consumption prediction model built based on TCN and GRU, generating a more accurate prediction of cold storage energy consumption. This enables intelligent cold chain warehouse inventory management based on predicted energy consumption. An attention mechanism is used to calculate the similarity between the user's current hidden state and the target item to update the hidden state of the gated loop unit. Combined with potential temporal parameter sequence features, this effectively improves the performance and accuracy of the cold storage energy consumption prediction task. It overcomes the problem of unclear and easily changing relationships among various parameters in the cold storage energy consumption-related parameter set, thus accurately mining current relationships and achieving dynamic and accurate correlation judgment of the cold storage energy consumption-related parameter set. This results in higher correlation of the data input to the energy consumption prediction model of the temporal convolutional fusion bidirectional memory network, enabling accurate generation of predicted cold storage energy consumption. This allows for dynamic management of cold chain warehouse inventory limits, goods storage locations, and inbound / outbound strategies based on predicted cold storage energy consumption.

[0134] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0135] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention; various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cold chain warehousing and inventory management system based on big data analytics, characterized in that, include: The correlation factor determination module is used to classify the set of energy consumption related parameters of cold storage into a set of unrelated parameters and a set of related parameters through a correlation degree model. The correlation degree model is constructed based on an improved Apriori algorithm with a set dynamic threshold, which is dynamically adjusted according to the change in the state of cold storage and / or the feature value of the correlation rule set. The data imputation module, which is connected to the correlation factor determination module, is used to determine the corresponding parameters in the cold storage historical dataset based on the uncorrelated parameter set, and generate imputation parameters by Lagrange interpolation of the corresponding parameters. An energy consumption prediction module, connected to the correlation factor determination module and the data imputation module, is used to generate a predicted cold storage energy consumption by using the imputation parameters and the correlation parameter set through an energy consumption prediction model. The energy consumption prediction model is constructed based on an architecture that integrates a temporal convolutional network with a gated recurrent unit and an attention mechanism. The temporal convolutional network is used to generate a temporal parameter sequence based on the imputation parameters and the correlation parameter set. The gated recurrent unit is used to generate the predicted cold storage energy consumption as a hidden state based on the temporal parameter sequence. The attention mechanism is used to control the updating of the hidden state by the gated recurrent unit. The cold chain warehouse inventory management module is connected to the energy consumption prediction module and is used to manage the inventory optimization, cargo storage location and inbound / outbound strategies of the cold chain warehouse based on the predicted cold storage energy consumption.

2. The cold chain warehousing and inventory management system based on big data analysis according to claim 1, characterized in that, The temporal convolutional network has at least two residual blocks, each residual block comprising an extended causal convolutional layer, a weight normalization layer, a LeakyReLU activation function layer, and a regularization layer.

3. The cold chain warehousing and inventory management system based on big data analysis according to claim 1, characterized in that, The hidden state includes the previous hidden state, the current hidden state, and the transitive hidden state; the attention mechanism includes a weight generation subunit and a transitive subunit. The weight generation subunit is used to generate attention weights based on the current hidden state and the target vector through an attention mechanism. The transmission subunit is used to generate the transmission hidden state based on the attention weight and the previous hidden state, and to transmit the transmission hidden state to the gated loop unit.

4. The cold chain warehousing and inventory management system based on big data analysis according to claim 1, characterized in that, The energy consumption prediction module also includes an objective function unit; The objective function unit is used to control the optimization prediction of the gated recurrent unit through an objective function, wherein the objective function is constructed based on cross-entropy and activation function.

5. The cold chain warehousing and inventory management system based on big data analysis according to claim 1, characterized in that, The correlation factor determination module includes a preprocessing unit, a correlation rule generation unit, and a judgment unit; The preprocessing unit is used to process the set of energy consumption-related parameters of the cold storage and generate an interval representation. The association rule generation unit is used to iterate the interval representation to generate an association rule set associated with the comprehensive energy consumption of the cold storage. The judgment unit is used to classify the set of energy consumption related parameters of the cold storage according to the set of association rules to generate the set of unrelated parameters and the set of associated parameters.

6. The cold chain warehousing and inventory management system based on big data analysis according to claim 5, characterized in that, The dynamic threshold includes a support threshold and a confidence threshold; the cold storage status change includes changes in the external temperature of the cold storage and changes in cargo storage; the feature value of the association rule set is the average lift of the association rule set; and the judgment unit includes a support dynamic adjustment subunit and a confidence dynamic adjustment subunit. The support dynamic adjustment subunit is used to set the support threshold according to the change in external temperature of the cold storage and the change in cargo storage. The confidence dynamic adjustment subunit is used to set the confidence threshold based on the average lift of the association rule set.

7. The cold chain warehousing and inventory management system based on big data analysis according to claim 5, characterized in that, The preprocessing unit includes a data cleaning subunit, a standardization subunit, and a clustering and grouping subunit; The data cleaning subunit is used to clean the set of cold storage energy consumption related parameters and generate a cleaned cold storage energy consumption related dataset. The standardization subunit is used to standardize the energy consumption related dataset of the cold storage cleaning to generate a standard cold storage energy consumption related dataset. The clustering sub-unit is used to discretize the standard cold storage energy consumption related dataset using a clustering algorithm to generate the interval representation; The set of energy-related parameters for cold storage includes internal and external environmental parameters of cold chain storage, switching parameters of cold chain storage, inventory parameters of cold chain storage, and refrigeration equipment parameters of cold chain storage.

8. The cold chain warehousing and inventory management system based on big data analysis according to claim 1, characterized in that, The data filling module also includes a historical dataset update unit; The historical dataset update unit is used to update the historical dataset of the cold storage based on the operational stability of the cold storage.

9. The cold chain warehousing and inventory management system based on big data analysis according to any one of claims 1 to 8, characterized in that, The energy consumption parameters of the cold storage include the categories of inventory goods, and the cold chain warehousing inventory management module also includes an inventory optimization unit and a goods storage location planning unit. The inventory optimization unit is used to adjust the regional inventory limit of cold chain storage based on the predicted cold storage energy consumption. The cargo storage location planning unit is used to determine the recommended storage area based on the predicted cold storage energy consumption and the category of inventory goods.

10. The cold chain warehousing and inventory management system based on big data analysis according to any one of claims 1 to 8, characterized in that, The cold chain warehousing inventory management module includes an inbound / outbound strategy unit; The inbound / outbound strategy unit is used to select the time period during which the predicted energy consumption of the cold storage is in the low power consumption range as the recommended inbound / outbound time period.

Images