Intelligent Optimization Method and System for Low-Temperature Storage Location of Cell Banks

By establishing a dynamic thermal field model for liquid nitrogen storage devices and employing a multi-objective optimization method, the problem of thermal exposure differences caused by storage and retrieval operations in liquid nitrogen storage devices was solved, achieving efficient optimization of sample storage locations, reducing liquid nitrogen consumption, and shortening sample retrieval time.

CN123117433APending Publication Date: 2026-11-17SEMPRIS (BEIJING) CELL TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611263497.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in liquid nitrogen storage devices fail to quantify location-level dynamic thermal exposure differences driven by access operations, resulting in samples near high-frequency access locations experiencing high thermal exposure. Furthermore, existing management systems cannot predict future access demand patterns and cannot optimize sample storage locations under thermodynamic constraints.

Method used

By establishing a dynamic thermal field model of liquid nitrogen storage equipment, collecting temperature change data, predicting access frequency, and using a multi-objective combined optimization method to achieve Pareto optimal allocation between samples and storage locations, the storage location is optimized to minimize heat exposure and maximize access efficiency.

Benefits of technology

It achieved a reduction of liquid nitrogen consumption by more than 15%, a reduction of average sample retrieval time by more than 60%, and provided a forward-looking location optimization solution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117433A_ABST
    Figure CN123117433A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent optimization method and system for cryogenic storage locations in cell banks, relating to the field of intelligent management technology for biobanks. The method includes: collecting temperature change data of each storage location in a liquid nitrogen storage device during access operations, and establishing a cumulative heat exposure index for each storage location; predicting the access probability distribution of each sample based on sample attributes and associated project progress information; performing multi-objective combined optimization of the mapping relationship between samples and storage locations with the optimization objectives of maximizing access efficiency and minimizing cumulative heat exposure; selecting a target scheme from a set of Pareto optimal layout schemes and generating location adjustment instructions. This invention models the interior of the liquid nitrogen storage device as a dynamic thermal field driven by access behavior, which can reduce liquid nitrogen consumption by more than 15% and shorten sample retrieval time by more than 60%.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for biobanks, specifically to a method and system for intelligent optimization of the low-temperature storage location in cell banks. Background Technology

[0002] With the rapid development of cell therapy, precision medicine, and gene editing technologies, large-scale biobanks are playing an increasingly important supporting role in clinical research and industrialization. Liquid nitrogen storage devices, as the core infrastructure for long-term preservation of biological samples, require safe, efficient, and traceable storage management of tens to hundreds of thousands of cell samples in an ultra-low temperature environment approaching -196 degrees Celsius. However, with the continuous growth in the size of biobanks and the parallel advancement of related research projects, traditional storage location management methods are facing increasingly prominent efficiency and quality bottlenecks.

[0003] Existing biobank management systems primarily focus on sample identification tracking and environmental monitoring. Chinese patent application CN107590632A discloses a management system, method, and server for cell production, storage, and transportation. This system uses IoT terminals to add identification tags to cell samples and achieves information management of the entire cell lifecycle through production, storage, and transportation modules. However, this approach treats the interior of liquid nitrogen storage equipment as a uniformly heated isothermal container, and assigns sample storage locations sequentially based on the order of entry or available numbering, failing to consider the inherent differences between storage locations in terms of temperature stability, accessibility, and thermal impact on adjacent samples.

[0004] In the field of general warehousing and logistics, storage location optimization has become a mature research direction. US Patent Application Publication No. US20130226649A1 discloses a method for efficient storage and retrieval location allocation. This method allocates items to warehouse locations based on multiple constraints such as item size, weight, and picking frequency, using priority ranking rules. However, this method is geared towards static location allocation under ambient temperature conditions, does not address the thermodynamic disturbances of storage and retrieval operations on the surrounding storage environment, and does not establish a predictive model for storage and retrieval frequency to proactively guide location allocation.

[0005] In the field of cold chain warehousing, existing research has incorporated temperature zoning constraints into the objective function of storage optimization. Recent academic studies have proposed a fresh food storage optimization method integrating a particle swarm optimization-guided genetic-simulated annealing hybrid algorithm, constructing a multi-objective model that includes picking paths, temperature zone energy consumption, and first-in-first-out (FIFO) penalties. However, such methods are geared towards food cold chain scenarios ranging from 0 to -18 degrees Celsius, and their temperature constraints employ a static region partitioning approach, dividing the warehouse into two fixed temperature zones: a refrigerated zone and a frozen zone. This ignores the dynamic temperature fluctuations within the same temperature zone caused by differences in the frequency of storage and retrieval operations. This approach differs fundamentally from liquid nitrogen storage environments: each opening operation of a liquid nitrogen storage device causes hot air to penetrate downwards along the opening, creating a transient, non-uniform temperature gradient field inside the device. The temperature rise near the tank opening is significantly greater than at the deeper bottom layers, and the time required to return to steady state exhibits a complex nonlinear coupling relationship with the duration of the opening, the liquid level, and the sample filling density of adjacent storage locations.

[0006] Based on the above analysis, existing technologies have the following fundamental shortcomings in optimizing the allocation of storage locations for liquid nitrogen cryogenic storage: they assume the interior of liquid nitrogen storage devices to be a homogeneous and isothermal space, ignoring the quality differences between storage locations caused by the dynamic thermal field effect driven by access operations. Specifically, samples near high-frequency access locations experience a much higher cumulative heat exposure than samples in low-frequency regions. Existing management systems cannot quantify this location-level heat exposure difference, predict future access demand patterns for each sample, or perform forward-looking optimization allocation of sample storage locations under thermodynamic constraints. Summary of the Invention

[0007] To address the core technical bottleneck in existing large-scale biobank cryogenic storage management—which treats the storage space as a homogeneous isothermal body while ignoring location-level dynamic thermal exposure differences driven by access operations—this invention provides a method and system for intelligent optimization of cryogenic storage locations in cell banks. By modeling the interior of the liquid nitrogen storage device as a dynamic thermal field driven by access behavior and predicting access frequencies based on sample attributes and project schedule, under the constraint of thermal exposure budget, Pareto optimal allocation of the mapping relationship between samples and storage locations is achieved from the perspective of transient heat transfer mechanisms and multi-objective combined optimization principles.

[0008] The technical solution of this invention is as follows: A smart optimization method for the cryogenic storage location of cell banks includes the following steps: Temperature change data of each storage location in the liquid nitrogen storage device during access operations are collected. The cumulative heat exposure index of each storage location is established by integrating the temperature response of all access operations experienced by each storage location within a preset statistical period. Based on the attribute information of each sample and the progress information of the projects associated with each sample, the access probability distribution of each sample within a preset time window is predicted. With maximizing the access efficiency of each sample and minimizing the cumulative heat exposure of each storage location as optimization objectives, and with constraints on the temperature sensitivity level of the samples and the aggregation constraints of samples in the same batch as constraints, a multi-objective combination optimization is performed on the mapping relationship between samples and storage locations to obtain a Pareto optimal layout scheme set. From the Pareto optimal layout scheme set, a target scheme that meets preset preferences is selected, and a sample position adjustment instruction is generated.

[0009] This invention also provides a smart optimization system for the low-temperature storage location of cell banks, comprising: The thermal exposure modeling module collects temperature change data of each storage location in the liquid nitrogen storage device during access operations. It integrates the temperature response of each storage location during all access operations within a preset statistical period to establish a cumulative thermal exposure index for each location. The access frequency prediction module predicts the access probability distribution of each sample within a preset time window based on the attribute information of each sample and the progress information of the projects associated with each sample. The location optimization module optimizes the mapping relationship between samples and storage locations by maximizing the access efficiency of each sample and minimizing the cumulative thermal exposure of each storage location, using constraints such as the temperature sensitivity level of the samples and the aggregation constraints of samples in the same batch, to obtain a Pareto optimal layout scheme set. The scheme output module selects a target scheme that meets preset preferences from the Pareto optimal layout scheme set and generates a sample location adjustment instruction.

[0010] The beneficial effects of this invention are as follows: First, this invention, by establishing a cumulative heat exposure index for each storage location, transforms the internal structure of liquid nitrogen storage devices from a homogeneous isothermal assumption to a dynamic thermal field model driven by storage and retrieval behavior, thus providing each storage location with a quantifiable heat exposure budget. The mechanism lies in the fact that during each opening operation of the liquid nitrogen storage device, the process of hot air intruding downwards along the opening follows transient heat transfer laws. Different spatial locations experience varying degrees of temperature rise and recovery time constants due to differences in distance from the opening and liquid level height. The integral accumulation of this temperature response curve over time reflects the overall degree of thermal shock experienced by each location. Compared to the prior art that treats the storage space as a homogeneous container, this invention reveals heat exposure differences of up to an order of magnitude between locations, providing a physical basis for subsequent differentiated location allocation.

[0011] Secondly, this invention upgrades from historical statistics to forward-looking prediction by predicting the access probability distribution based on sample attributes and project progress information. The mechanism lies in the fact that sample access behavior in a biobank is not uniformly random, but closely coupled with the progress nodes of related research projects. Project initiation and completion phases are typically accompanied by concentrated sample access operations, forming sudden access spikes. Conventional time-series statistical methods cannot capture this non-stationary mutation pattern associated with project events. This invention uses project progress node information as a driving factor for the prediction model, enabling access frequency prediction to anticipate upcoming access peaks. Compared to simple frequency statistics based solely on historical access records in existing technologies, this invention significantly improves prediction accuracy, allowing location optimization to pre-adjust sample layout before access peaks arrive.

[0012] Third, this invention achieves a Pareto optimal balance between maximizing access efficiency and minimizing heat exposure by embedding heat exposure constraints into a multi-objective combined optimization constraint system. The mechanism lies in the inherent contradiction between these two objectives: access efficiency requires placing high-frequency samples in convenient locations near equipment openings, while minimizing heat exposure requires placing temperature-sensitive samples in deeper locations away from openings. This invention, through a multi-objective Pareto optimization framework, can find a set of non-dominated solutions without sacrificing any objective, allowing operators to select the optimal solution based on current management preferences. This multi-objective Pareto optimization paradigm has not been adopted in existing biobank management systems and cold chain storage optimization schemes. The synergistic effect of these three technical features is manifested in the following ways: heat exposure modeling provides the physical basis for location differentiation, and access frequency prediction provides the temporal basis for sample priority. Both together provide high-quality input for multi-objective optimization, making the optimization results both physically reasonable and forward-looking. The combined effect of the three exceeds the sum of their individual effects. Individual thermal exposure modeling can only achieve static temperature zoning management, and individual frequency prediction can only achieve a proximity allocation strategy without considering temperature. However, the fusion of the two with multi-objective optimization can achieve a forward-looking Pareto optimal layout under dynamic thermal field constraints, which can reduce liquid nitrogen consumption by more than 15% and shorten the average sample retrieval time by more than 60%. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the intelligent optimization method for low-temperature storage location of cell banks provided in this embodiment of the invention.

[0014] Figure 2 This is a schematic diagram of the architecture of the intelligent optimization system for low-temperature storage location of cell banks provided in an embodiment of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0016] Reference Figure 1 The intelligent optimization method for low-temperature storage location of cell banks provided in this embodiment of the invention includes the following steps S1 to S4.

[0017] Step S1: Location thermal exposure modeling. The core objective of Step S1 is to transform the thermal environment of each storage location in the liquid nitrogen storage device from a qualitative description into a quantitative indicator, accumulating thermal exposure indicators.

[0018] In the implementation scenario of this invention, the liquid nitrogen storage device is typically a large gas-phase liquid nitrogen tank or a liquid-phase liquid nitrogen tank, with multiple layers of cryopreservation racks inside. Each layer of the cryopreservation rack contains several cryopreservation box positions, and the entire device can hold thousands to tens of thousands of cryopreservation tube samples. When an operator needs to access a specific sample, they need to open the top cover of the device, lift the corresponding cryopreservation rack above the operating platform, complete the sample retrieval or placement operation, and then return the cryopreservation rack and close the cover. During this process, ambient air at room temperature enters the device from below through the opening, creating a transient temperature gradient field.

[0019] In one specific embodiment, the internal space of the liquid nitrogen storage device is discretized as follows: Each storage location has one storage location. The spatial coordinates are the vertical depth of this position from the device opening. and horizontal offset distance This is used to characterize the temperature changes at each storage location during each opening operation. Temperature data can be collected by embedding temperature sensors at several representative locations inside the device, or by establishing a temperature response prediction model for each storage location through finite element thermal simulation.

[0020] Storage location during each lid-opening and retrieval operation Temperature from the boiling point of liquid nitrogen (Approximately -196 degrees Celsius) The temperature rises to its peak temperature and then gradually recovers. This process is modeled as a transient heat transfer process, and the data is stored in [location missing]. In the The temperature response function during the second lid-opening operation is: , in: For storage location In the During the second opening operation, at all times The temperature value is a scalar quantity in degrees Celsius, calculated by this formula, and characterizes the temperature response of a specific location over time during a specific operation. is the reference value for the boiling point temperature of liquid nitrogen, is a scalar quantity with a value of -196 degrees Celsius, and is a known physical constant; For storage location In the During the second opening operation, relative to the reference temperature The peak temperature rise is a scalar value, ranging from 0 to 80 degrees Celsius (i.e., the absolute temperature rises to -116 degrees Celsius), and is measured in degrees Celsius. It is determined by the spatial coordinates of the storage location and the opening operation parameters, and reflects the maximum extent of heat intrusion at that location during this operation. It is a time variable, a scalar, and its unit is seconds (s). For the first The opening time of the next opening operation is a scalar quantity in seconds, and is determined by the access operation log. For storage location The thermal recovery time constant is a scalar value ranging from 30 to 600 seconds, measured in seconds. It is determined by the spatial coordinates of the storage location and the thermophysical properties of the surrounding medium. Physically, it represents the characteristic time required for the temperature to recover from its peak value to a steady state. The larger the value, the longer it takes for the location to return to a steady state after the cover is opened, meaning the weaker the thermal buffering capacity.

[0021] Peak temperature rise The calculation needs to consider the spatial coordinate parameters of the storage location. In this embodiment, Determined by the following relationship: , in: For the opening of the equipment (i.e. , The maximum temperature rise under the reference opening time is a scalar value, ranging from 50 to 80 degrees Celsius, and is determined by actual measurement calibration or thermal simulation. It reflects the upper limit of the intensity of thermal intrusion at the opening of the equipment. The temperature decay coefficient in the vertical depth direction is a scalar quantity, ranging from 0.01 to 0.10 cm. The unit is cm. The height of the liquid nitrogen level is determined by the natural convection characteristics of the gas inside the equipment and the liquid nitrogen level. The larger the value, the faster the temperature decays along the depth direction, meaning the better the equipment's insulation performance; For storage location The vertical depth from the equipment opening is a scalar value, ranging from 0 to 120 cm, and is determined by the equipment structure and the position of the cryogenic rack. The temperature attenuation coefficient in the horizontal offset direction is a scalar value ranging from 0.005 to 0.05 cm. The unit is cm. The dimensions are determined by the inner diameter of the equipment and the horizontal layout of the cryogenic storage racks; For storage location The horizontal offset distance from the central axis of the equipment is a scalar value, ranging from 0 to 30 cm, and the unit is cm; For the first The duration of each lid opening operation is a scalar value, ranging from 10 to 300 seconds, and is calculated from the difference between the lid opening and closing times in the operation log. The opening time is a reference value, a scalar value, and is set to 60 seconds for calibration. The standard opening time used at that time; The effect of opening time on the index is a scalar quantity, ranging from 0.3 to 0.8, dimensionless, and determined by fitting measured data. A value less than 1 indicates that the peak temperature rise increases sublinearly with the increase of the opening time, meaning that the marginal thermal effect of extending the opening time decreases. The unit is Celsius. Right side of the equation: The unit is Celsius. middle The unit is cm , The unit is cm; the product is dimensionless, and the exponential function is dimensionless; similarly... Dimensionless; Since both the numerator and denominator are 's', the ratio is dimensionless, and it remains dimensionless after exponentiation. Therefore, the unit on the right side of the equation is degrees Celsius, consistent with the left side.

[0022] After obtaining the temperature response of each storage location for each opening operation, step S1 further calculates the temperature response of each storage location within a preset statistical period. The cumulative heat exposure value within a given period. The calculation of the cumulative heat exposure value involves integrating the temperature response curve exceeding a preset temperature threshold over time. Specifically, this is set within the statistical period. Internal storage location They experienced Secondary direct or indirect temperature disturbances, then storage location Cumulative heat exposure value for: , in: For storage location In the statistical period The cumulative heat exposure value within the storage area is a scalar value ranging from 0 to positive infinity, with units of degrees Celsius multiplied by seconds (°C·s). It is calculated using this formula and characterizes the overall degree of thermal shock experienced by the storage location within the statistical period. The larger the value, the more severe the deviation of the sample at that location from the steady-state low-temperature environment within the statistical period; For storage location In the statistical period The total number of temperature disturbances experienced within the device is a positive integer scalar, including direct access operations and indirect temperature disturbances caused by access to adjacent locations (see the calculation of passive heat exposure increments below). is the temperature threshold offset, which is a scalar with a value range of 5 to 20 degrees Celsius. It is dimensionless and corrected to units of degrees Celsius. It is determined by the sample library management strategy. The purpose of setting this threshold is to filter out temperature fluctuations that are too small and will not have a substantial impact on sample activity, and only to include significant temperature deviations in the cumulative heat exposure. This indicates the maximum value operator, which returns zero when the temperature deviation is below a threshold, ensuring that only positive deviations are counted in the integral; the upper limit of integration. This indicates the moment when the integration stops at 5 times the time constant after the lid is opened, at which point the temperature has recovered to less than 1% of the steady state. The unit is °C·s. (The text within the integral sign is incomplete and cannot be translated.) The parameter is the temperature difference, in °C; the integral variable. The unit is seconds (s); summation does not change the unit. Therefore The unit is °C·s, consistent with the definition. In actual operation, when an operator performs a storage or retrieval operation on a cryogenic rack within a liquid nitrogen storage device, not only are the storage locations on that rack directly exposed to the intrusion of hot air, but the storage locations on adjacent cryogenic racks also experience a certain degree of temperature disturbance due to heat conduction and convection effects. This invention refers to this temperature rise indirectly caused by storage or retrieval operations at adjacent locations as passive heat exposure increment.

[0023] Specifically, when the target storage location a certain adjacent storage location Target location while being accessed Therefore, the increase in passive heat exposure is borne for: , in: For storage location Due to adjacent positions The The passive heat exposure increment suffered by each access operation is a scalar quantity in °C·s, calculated by this formula, and characterizes the degree of indirect thermal impact of adjacent access operations on the target location. The thermal coupling attenuation proportionality coefficient is a scalar quantity, ranging from 0.05 to 0.30, dimensionless, and determined by the thermal insulation structure between the cryogenic racks inside the equipment and the properties of the filling medium. The larger the size, the worse the thermal insulation performance between cryopreservation racks, and the stronger the impact of adjacent operations on the target location. Adjacent storage locations In the The peak temperature rise during this operation is a scalar quantity in degrees Celsius, defined as in the aforementioned formula; For target storage location The thermal recovery time constant is defined as in the aforementioned formula; For target storage location and adjacent storage locations The spatial distance between the two storage locations is a scalar, ranging from 1 to 50 cm, and is calculated from the relative positions of the two storage locations on the cryopreservation rack. The equivalent thermal propagation characteristic length of the filling medium inside the equipment is a scalar quantity, ranging from 5 to 20 cm, and is determined by a combination of the thermal diffusivity of the gaseous or liquid nitrogen inside the equipment and the thermal conductivity of the cryogenic rack material. The larger the value, the wider the spatial propagation range of the thermal disturbance. The unit should be ℃·s. Right side of the equation: Dimensionless The unit is ℃. The unit is s. The units for both the numerator and denominator are cm, the ratio is dimensionless, and the exponential function is dimensionless. Therefore, the overall unit on the right is °C·s, consistent with the left side.

[0024] Passive heat exposure increment Will be overlaid to the target storage location Cumulative heat exposure value In actual calculations, for each cryopreservation rack involved in the opening operation, the system traverses all storage locations within a preset space range (e.g., 30cm) around the cryopreservation rack, calculates and adds up its passive heat exposure increment.

[0025] Through step S1, a cumulative heat exposure value is obtained for each storage location in the liquid nitrogen storage device. This value comprehensively reflects the overall thermal shock experienced by the location during the statistical period due to direct access operations and indirect thermal effects from adjacent locations. Locations with higher values ​​will be marked as "high heat exposure locations" in the subsequent optimization allocation step S3, and are not suitable for placing temperature-sensitive samples.

[0026] Step S2: Access frequency prediction. The core objective of step S2 is to predict the access probability distribution of each sample within a future preset time window based on the sample's attribute information and the progress information of related projects, so as to provide demand-side input for the optimized allocation in step S3.

[0027] In one specific embodiment, the liquid nitrogen storage device stores... Samples, each sample ( The sample has the following attributes: sample type (e.g., primary cells, iPSCs, tumor tissue, etc.), temperature sensitivity level (divided into three levels: high sensitivity, medium sensitivity, and low sensitivity), batch number, entry time, and one or more associated research project numbers. The system obtains the above information from the sample bank management database, and also obtains the progress schedule of each research project, including the current stage of the project, the expected time node for entering the next stage, and the expected duration of each stage.

[0028] For each sample Its future preset time window The access probability distribution (e.g., over the next 30 days) is first given by the baseline prediction model. Based on the sample's historical access records, the baseline prediction model uses a weighted moving average method within a sliding window to calculate the baseline access probability for each day within the window period. In this embodiment, the baseline access probability is determined by statistically analyzing access records over the past 90 days and assigning exponentially decaying weights based on time proximity. This method is a mature approach in the field of time series statistics and will not be described in detail here.

[0029] However, baseline predictions based solely on historical records suffer from a systematic lag when faced with sudden access demands caused by changes in project schedules. When a related project transitions from the preparation phase to the experimental execution phase, the access frequency of its related samples may surge to 5 to 10 times the normal level within a few days, and the baseline prediction model cannot anticipate such sudden changes. To address this issue, this invention introduces an event-triggered impulse response superposition mechanism.

[0030] Specifically, the system continuously monitors the progress status of the projects associated with each sample. When the system detects a sample... A related project When entering a preset access trigger phase (such as "experiment start," "batch testing," or "project completion"), an event-triggered impulse response is superimposed on the baseline value of the sample's predicted access probability. The superimposed predicted access probability value... for: , in: For the sample In the future The predicted access probability value is a scalar, ranging from 0 to 1, dimensionless (probability value), calculated by this formula, and represents the sample at time t. The possibility of being accessed; For the sample At any moment The baseline access probability is a scalar, ranging from 0 to 1, dimensionless, and determined by a sliding window weighted moving average method based on historical records. For the sample The set of associated items detected entering the access trigger phase within the prediction time window; For the index variable of the associated project, iterate through Each triggering item in; For the project The amplitude of the impulse response generated after entering the triggering phase is a scalar, ranging from 0.1 to 0.9, dimensionless, and determined by the statistical average of access intensity historically triggered by similar phases in this project. The larger the value, the stronger the pull effect of the project phase on the sample access frequency; For the project The start time of entering the triggering phase is a scalar, in days (with the predicted start date as the origin), obtained from the project schedule or calculated based on the current project progress and the expected time nodes; For the project The pulse decay time constant during the trigger phase is a scalar value ranging from 3 to 30 days, determined by the historical average duration of similar phases in this project. The larger the value, the longer the peak access duration and the slower the pulse decay during that phase. For indicator functions, when The value is 1 if the triggering phase begins, and 0 otherwise, ensuring that the pulse response only takes effect after the triggering phase has started. It is a natural exponential function that describes the process by which the impulse response gradually decays over time. Dimensionless. Right side of the equation: Dimensionless; Dimensionless Both the numerator and denominator are in the range of days; the ratio is dimensionless; the exponential function is dimensionless. Dimensionless. Therefore, the right side is dimensionless overall, consistent with the left side.

[0031] In the prediction time window Inside, the system processes each sample Daily access probability prediction Aggregate the samples to obtain the total access probability of each sample within the window period. : , in For the sample Total access probability (dimensionless aggregate value) within the prediction time window. The preset time window length ranges from 7 to 30 days. The larger the value, the more times the sample will be accessed in the future, and in step S3 it should be assigned to a location with higher accessibility.

[0032] Through step S2, the system obtained all Access probability distribution of samples within the prediction time window The temperature sensitivity level and batch number of each sample, as well as other attribute information, will serve as the demand-side input for step S3, multi-objective optimization.

[0033] Step S3: Multi-objective constraint optimization. The core objective of step S3 is to construct a multi-objective combined optimization model based on the location-level thermal exposure index established in step S1 and the sample-level access probability distribution predicted in step S2, and solve for the Pareto optimal mapping relationship between samples and storage locations.

[0034] Let the sample set be denoted as The set of storage locations is denoted as ,in The total number of samples, The total number of storage locations, Define decision variables. ,in Indicates the sample Assign to storage location , This indicates that no allocation will be made.

[0035] The multi-objective optimization model in this embodiment includes two optimization objectives and two sets of constraints.

[0036] The first optimization objective is to maximize access efficiency. Access efficiency is measured by the weighted sum of access convenience for all samples, where the weights are the access probabilities of each sample. Convenience is characterized by the reciprocal of the standardized distance from the storage location to the device opening. Access efficiency objective function. for: , in: Let be the objective function value for access efficiency, be a scalar with a range from 0 to the upper positive bound, and be dimensionless. It is calculated using this formula. A larger size means that high-frequency samples are more concentrated in shallow, convenient locations, resulting in higher overall access efficiency. The optimization objective is to maximize this efficiency. ; The decision variable matrix has 1 dimension. ,element ; For the sample The total access probability is defined in the same way as in step S2; For storage location The vertical depth from the equipment opening is defined in the same way as in step S1; This is the maximum vertical depth from the opening among all storage locations in the device. It is a scalar value in cm and is determined by the device's structural parameters. For storage location The standardized convenience index has a value range of 0 to 1, is dimensionless, and has a location convenience value close to 1 near the opening and a location convenience value close to 0 at the bottom or deep level.

[0037] The second optimization objective is to minimize cumulative heat exposure. For each candidate layout scheme, after allocating specific samples to each storage location, it is necessary to estimate the expected cumulative heat exposure of each location in the next statistical period. The samples... Access probability With assigned position Combining the thermal exposure response characteristics, a thermal exposure objective function is defined. for: , in: The objective function value for heat exposure is denoted as , which is a scalar quantity, and its unit is °C·s (inherited). The unit (of which) is calculated using this formula, A larger value means that temperature-sensitive samples are assigned to locations with severe heat exposure; the optimization objective is to minimize this value. ; For the sample The temperature sensitivity weight coefficient is a scalar with values ​​of {1.0, 2.0, 5.0}, corresponding to low, medium, and high sensitivity levels, respectively. It is dimensionless and determined by the temperature sensitivity level in the sample attributes. High-sensitivity samples have a larger weight, making the optimizer more inclined to place them in low-heat exposure locations. For storage location The cumulative heat exposure value is defined in the same way as in step S1. The unit should be with Consistent. Dimensionless Dimensionless The unit is ℃·s. Dimensionless, therefore The unit is ℃·s.

[0038] The constraints consist of two sets. The first set is the temperature sensitivity level constraint: samples with a high temperature sensitivity level can only be assigned when their cumulative heat exposure value is below a preset threshold. The storage location, that is: , in This is a collection of samples with high temperature sensitivity. The upper limit of the acceptable cumulative thermal exposure threshold for highly sensitive samples is set by the sample library administrator based on the sample type.

[0039] The second group is the constraint on the clustering of samples within the same batch: the maximum spatial span of samples within the same batch in the storage layout shall not exceed the preset upper limit of the batch clustering radius. .set up For batch Given the sample set in the dataset, then: , in For storage location and The spatial distance between them is defined as in the aforementioned formula; This represents the upper limit of the batch aggregation radius, a scalar value ranging from 10 to 30 cm, determined by the sample bank management strategy. Smaller sizes require samples from the same batch to be stored more closely together, which helps to shorten the opening time during batch operations, but excessively small sizes... This will severely limit the feasible domain for optimization.

[0040] In addition, each sample is assigned to exactly one storage location ( At most one sample can be placed in each storage location. ).

[0041] The aforementioned multi-objective combinatorial optimization problem is solved using a genetic algorithm based on non-dominated sorting. After initializing a random population, the algorithm searches the feasible region using crossover and mutation operators, selects individuals from the population using non-dominated sorting and crowding distance, and ultimately converges to the Pareto front. This method is a mature technique in the field of multi-objective optimization, with a population size of 200 to 500, an iteration count of 500 to 2000 generations, a crossover probability of 0.8 to 0.95, and a mutation probability of 0.01 to 0.05.

[0042] In the Pareto optimization process, this invention further introduces a Pareto front curvature detection and adaptive weight adjustment strategy. Specifically, after each iteration, the system adjusts the non-dominated solutions on the current Pareto front according to... Value sorting, calculating local curvature between adjacent solutions. : , in: For the Pareto frontier The estimated local curvature at each non-dominated solution is a scalar, ranging from 0 to positive values, dimensionless, and calculated using this formula. The larger the value, the greater the curvature of the Pareto front at that point, which usually corresponds to a region where the trade-off between the two optimization objectives undergoes a qualitative change; Between adjacent non-dominated solutions The difference between the target values; Between adjacent non-dominated solutions The difference between the target values. The denominator uses the standard form of the curvature formula to ensure numerical stability.

[0043] when Exceeding the preset curvature threshold (The value ranges from 0.1 to 1.0, determined by the optimization accuracy requirements) When the Pareto front is at the th... There exists a non-convex region near each solution. In this non-convex region, the system increases the sampling density, and the corresponding region... The proportion of genetic algorithm population within the value range is increased to 30% to 50% of the total population, while an adaptive weight switching strategy is adopted: temporarily switching to weighted summation method in non-convex regions, with weights... and Traverse between 0 and 1 with a step size of 0.05 to obtain non-convex Pareto solutions that cannot be reached by convex combination methods.

[0044] The non-convex region discovered by curvature detection has a clear physical meaning: the layout scheme corresponding to this region is exactly a subset of the schemes where both the heat exposure constraint and the batch aggregation constraint tend to be compact. That is, in these schemes, temperature-sensitive samples are arranged in the deep position with the lowest heat exposure value, while samples in the same batch are tightly aggregated to minimize the number of times the lid is opened during batch operations. The simultaneous satisfaction of the two constraints makes the Pareto front exhibit a non-convex bend in this region.

[0045] Furthermore, this invention extracts joint constraint parameters from a subset of solutions in the non-convex region, achieving an upgrade from a data-driven black box to a physically constrained gray box. Specifically, the system traverses all non-dominated solution schemes in the non-convex region and extracts two physical parameters for each scheme: thermal isolation distance. and batch aggregation radius Thermal isolation distance is defined as the distance between all temperature-sensitive samples (high sensitivity level) and all high-frequency access samples in the scheme. Minimum spatial distance between values ​​located in the top 20% of all samples: , in For the high-frequency access sample set (total access probability) (The subset of samples that are in the top 20% of the total sample). Batch cluster radius. Defined as the average of the maximum spatial span of the storage space occupied by each batch of samples.

[0046] The system performs statistics on all schemes in the non-convex region. and The distribution, take The 25th percentile is used as the lower limit constraint for thermal isolation distance. ,Pick The 75th percentile is used as the upper limit constraint for batch aggregation radius. Then and As additional constraints, these are fed back into the optimization model, and the multi-objective genetic algorithm is re-executed for secondary optimization. The feasible region of the secondary optimization is further refined by physical constraints, ensuring that the solutions on the final Pareto front all satisfy the physical rationality of thermal isolation and batch aggregation, thus exhibiting stronger generalization ability and practical feasibility.

[0047] Step S4: Solution Output and Execution. Step S4 selects a target solution that meets the current management preferences from the Pareto optimal layout solution set obtained after the secondary optimization in Step S3, and generates an executable sample position adjustment instruction.

[0048] In one specific embodiment, non-dominated solutions on the Pareto front are sorted according to user-preset preference weights. Users can set the relative importance of access efficiency and heat exposure control through a management interface; for example, prioritizing access efficiency during routine sample library operations and prioritizing heat exposure control during large-scale new sample intake. Based on the user-defined preference weights, the system uses a weighted ideal point distance method to select the solution from the Pareto front that best matches the user's preferences as the target solution.

[0049] After selecting a target solution, the system compares the sample-location mapping in the target solution with the current actual layout to identify a list of samples whose locations need adjustment. For each sample requiring adjustment, the system generates a location adjustment instruction, including the sample number, current storage location number, target storage location number, and suggested execution time. The system also calculates the optimal execution order for location adjustments, prioritizing adjustments to samples with the highest heat exposure risk, and merging multiple adjustment operations on the same cryogenic rack into a single lid-opening operation to reduce additional heat intrusion.

[0050] In addition, step S4 includes a benefit assessment. The system calculates three benefit indicators based on the layout before and after the adjustment: the expected change in liquid nitrogen consumption (estimated by comparing the product of the total number of openings and the duration of each opening before and after the adjustment), the expected change in average sample retrieval time (estimated by comparing the weighted average access path length of high-frequency samples before and after the adjustment), and the expected reduction in cumulative heat exposure (estimated by comparing the sum of the cumulative heat exposure values ​​of all highly sensitive samples before and after the adjustment). The three indicators are then summarized to generate a benefit assessment report for the sample bank management personnel to use as a reference for decision-making.

[0051] For newly added samples, the system automatically performs step S2 to predict the access probability and step S3 to allocate the location. It selects the optimal location from the available locations in the current layout to achieve intelligent location allocation for newly added samples.

[0052] In a complete operating cycle, the system re-executes the entire process of steps S1 to S4 according to the preset update cycle (1 to 7 days) to adapt to the dynamic environmental changes brought about by changes in the liquid level of the liquid nitrogen storage equipment, fluctuations in sample inventory, and adjustments to the project schedule, thereby achieving continuous optimization of the storage location layout.

[0053] Reference Figure 2 The intelligent optimization system for low-temperature storage location of cell banks provided in this embodiment of the invention includes a heat exposure modeling module, an access frequency prediction module, a location optimization module, and a scheme output module. The four modules interact with each other through a data bus.

[0054] The thermal exposure modeling module performs the function of step S1 in the aforementioned method embodiment. This module connects to the temperature acquisition subsystem of the liquid nitrogen storage device via a data interface to obtain temperature change data at each storage location within the device; it also connects to the operation log database of the sample bank management system via a data interface to obtain time records of each opening operation. The thermal exposure modeling module internally includes a temperature response calculation unit and a thermal exposure integration unit. The temperature response calculation unit calculates the temperature response based on the spatial coordinate parameters (vertical depth) of each storage location. and horizontal offset distance And the duration of each lid-opening operation. Calculate the peak temperature rise of each storage location during each operation. and temperature response function The heat exposure integration unit receives the output from the temperature response calculation unit, performs time integration on all temperature response curves within a preset statistical period, and obtains the cumulative heat exposure value at each storage location. In addition, the thermal exposure integration unit also includes a passive thermal exposure calculation subunit, which calculates the passive thermal exposure increment caused by access operations at adjacent locations based on the spatial distance between adjacent storage locations and the equivalent thermal conductivity parameter. This data is then superimposed on the cumulative heat exposure value at the target location. The output of the heat exposure modeling module is the complete... Vector of cumulative heat exposure values ​​for each storage location The vector is transmitted to the position optimization module as a supply-side input.

[0055] The access frequency prediction module performs the function of step S2 in the aforementioned method embodiment. This module connects to the sample attribute database and project progress management database of the sample library management system via a data interface to obtain the attribute information, historical access records, and progress node information of associated projects for each sample. The access frequency prediction module internally includes a baseline prediction unit and an event-triggered overlay unit. The baseline prediction unit calculates the baseline access probability of each sample on each day within a preset time window based on the historical access records of each sample using a sliding window weighted moving average method. The event-triggered overlay unit continuously monitors the progress status of associated projects for each sample. When it detects that a project has entered a preset access trigger stage, it generates an impulse response function based on the historical access intensity statistics and average duration of that stage, and overlays it onto the baseline prediction value of the corresponding sample. The output of the access frequency prediction module is all... Total access probability vector of samples The temperature sensitivity level and batch number information of each sample are transmitted to the location optimization module as demand-side input.

[0056] The location optimization module performs the function of step S3 in the aforementioned method embodiment. This module receives the cumulative heat exposure value vector output by the heat exposure modeling module. Access probability vector output by the access frequency prediction module Based on sample attribute information, a multi-objective combined optimization model is constructed. The location optimization module internally includes an objective function construction unit, a constraint management unit, a multi-objective genetic algorithm solution unit, and a Pareto front analysis unit. The objective function construction unit is based on the access efficiency objective function. and heat exposure objective function The definition of the optimization objective is constructed. The constraint management unit manages temperature sensitivity level constraints, batch aggregation constraints, and joint constraint parameters fed back from Pareto front analysis. The multi-objective genetic algorithm solution unit implements population initialization, crossover and mutation operations, and non-dominated sorting selection. The Pareto front analysis unit detects the local curvature of the Pareto front during the optimization process, executes an adaptive weight adjustment strategy in non-convex regions, and extracts the lower limit of thermal isolation distance and the upper limit of batch aggregation radius parameters from the non-convex region schemes, feeding them back to the constraint management unit for secondary optimization. The output of the position optimization module is a set of Pareto optimal layout schemes.

[0057] The scheme output module performs the function of step S4 in the aforementioned method embodiment. This module receives the Pareto optimal layout scheme set output by the location optimization module and selects the target scheme based on the preference weights set by the user through the management interface. The scheme output module internally includes a scheme selection unit, an adjustment instruction generation unit, and a benefit evaluation unit. The scheme selection unit selects the target scheme from the Pareto front using the weighted ideal point distance method. The adjustment instruction generation unit compares the target scheme with the current actual layout, generates a list of sample location adjustment instructions, and calculates the optimal execution order. The benefit evaluation unit calculates the expected change in liquid nitrogen consumption, the expected change in average sample retrieval time, and the expected reduction in cumulative heat exposure, generating a benefit evaluation report. The output of the scheme output module is transmitted to the sample library management system and the liquid nitrogen storage device control system via a data interface, realizing the automated execution of location adjustment instructions.

[0058] In one specific deployment embodiment, the four modules described above are deployed on the sample library management server and automatically execute the entire process optimization as scheduled tasks according to a preset update cycle (1 to 7 days). Temperature data is transmitted to the server via wired or wireless data links through the temperature sensor array built into the liquid nitrogen storage device. Operation logs and project progress information are obtained through the database interface of the sample library management system. The optimization results are presented on the management interface in the form of a list of location adjustment suggestions and a benefit assessment report. Operators can choose to execute with one click or confirm the execution step by step.

[0059] The technical effects of the present invention are verified through specific comparative experiments.

[0060] The experimental environment was set up as a medium-sized biobank, equipped with four large gas-phase liquid nitrogen tanks. Each tank had 12 cryopreservation racks, with each rack containing nine cryopreservation box positions. Each cryopreservation box could hold 81 cryopreservation tube samples, for a total capacity of 34,992 storage positions. Approximately 25,000 samples were stored during the experiment, linking 42 ongoing projects.

[0061] Example 1 uses the method of the present invention to optimize the layout of the storage location. The system collects 30 days of temperature change data and operation logs to perform thermal exposure modeling in step S1, and performs access frequency prediction in step S2 with a 30-day prediction time window. The population size of the genetic algorithm in step S3 is set to 300, and the number of iterations is set to 1000 generations.

[0062] Comparative Example 1 uses the traditional method of allocating positions sequentially according to the order of entry into the warehouse, without any position optimization.

[0063] Comparative Example 2 uses an ABC classification method based on access frequency, dividing samples into three categories: A (high frequency), B (medium frequency), and C (low frequency) according to their historical access frequency. Category A samples are assigned to the upper layer of the liquid nitrogen tank, Category B to the middle layer, and Category C to the lower layer. This method only considers access convenience and does not take into account temperature field differences.

[0064] Comparative Example 3 employs a static temperature zone division method from the cold chain storage field, dividing the liquid nitrogen tank into an upper high-temperature zone (0 to 40 cm from the opening) and a lower low-temperature zone (below 40 cm from the opening). Temperature-sensitive samples are fixedly assigned to the low-temperature zone. This method only considers the static temperature zone and does not account for the dynamic impact of storage and retrieval activities on temperature.

[0065] After 60 days of continuous operation, the performance metrics of each solution are compared as follows: Regarding the average monthly consumption of liquid nitrogen, Example 1 reduced it by 18.2% compared to Comparative Example 1. This is because the present invention centrally arranges high-frequency samples in the same cryopreservation rack area, allowing multiple high-frequency samples to be accessed in a single opening operation, significantly reducing the total number of openings and thus reducing liquid nitrogen evaporation loss caused by hot air intrusion. While Comparative Example 2 also arranged high-frequency samples on the upper layer, it did not consider the constraint of batch aggregation, resulting in samples from the same project being scattered in different liquid nitrogen tanks, requiring multiple openings during batch operations.

[0066] Regarding the average sample retrieval time, Example 1 reduced it by 64.3% compared to Comparative Example 1. This is because the present invention, through the access probability prediction in step S2, pre-positions samples expected to be frequently accessed in convenient locations, eliminating the need for operators to search for target samples layer by layer in the deep cryogenic storage rack. While Comparative Example 3 placed temperature-sensitive samples in deeper locations, it did not optimize the allocation of non-sensitive samples, resulting in a scattered distribution of non-sensitive high-frequency samples that affected retrieval efficiency.

[0067] Regarding the cumulative heat exposure of highly sensitive samples, the average cumulative heat exposure value of highly sensitive samples in Example 1 was reduced by 71.5% compared to Comparative Example 1 and by 83.2% compared to Comparative Example 2. This is because Comparative Example 2, while placing high-frequency samples in the upper layer, did not spatially isolate temperature-sensitive samples from high-frequency samples, resulting in severe passive heat exposure for temperature-sensitive samples in adjacent positions in the upper layer. In contrast, this invention, through the thermal isolation distance constraint in step S3, ensures sufficient spatial distance between temperature-sensitive samples and high-frequency samples, effectively reducing passive heat exposure. Although the static temperature zone division in Comparative Example 3 places temperature-sensitive samples in the lower low-temperature zone, this method does not consider the differences in passive heat exposure caused by access from adjacent positions within the same low-temperature zone, thus limiting its actual protection effect.

[0068] Based on the above comparison results, it can be seen that the present invention achieves significantly better performance than existing technical solutions in three dimensions: liquid nitrogen consumption, sample retrieval efficiency, and sample preservation quality through the synergistic effect of location-level thermal exposure modeling (step S1), project progress-aware access frequency prediction (step S2), and multi-objective optimization under thermodynamic constraints (step S3).

[0069] The core innovation of this invention lies in transforming the understanding of the interior of liquid nitrogen storage devices from a traditional homogeneous isothermal space to a dynamic thermal field model driven by access behavior, and on this basis, constructing a closed-loop collaborative system of "thermal exposure modeling - access frequency prediction - multi-objective constraint optimization".

[0070] The mechanism of this collaborative system can be understood from the following three levels. At the physical level, each opening operation of the liquid nitrogen storage device follows transient heat transfer laws. Different spatial locations experience differentiated temperature disturbances due to differences in distance from the opening and liquid level. Integrating the temperature response over time across multiple operations quantifies the cumulative heat exposure differences at each location. This quantification reveals a qualitative difference of up to an order of magnitude between locations, providing an indispensable physical basis for differentiated location allocation. At the predictive level, the access behavior of biological samples is closely coupled with the progress nodes of related research projects. Access impulses generated by project phase transitions cause conventional time-series prediction methods to systematically fail. However, modeling project progress events as impulse responses and superimposing them on baseline predictions can significantly improve prediction accuracy. This improved accuracy enables proactive adjustments to location optimization rather than reactive responses. At the optimization mechanism level, there is an inherent contradiction between the two objectives of maximizing access efficiency and minimizing heat exposure. The multi-objective Pareto optimization framework can find a set of non-dominated solutions without sacrificing either objective. Moreover, the non-convex region of the Pareto front corresponds precisely to the collaborative solution of thermal isolation and batch clustering. The secondary optimization strategy of extracting physical constraints from this non-convex region and feeding them back to the optimization model upgrades the final solution from a purely data-driven model to a physically constrained gray-box model, which has stronger generalization ability.

[0071] The synergistic effect of the above three levels is characterized by nonlinear growth: individual thermal exposure modeling can only achieve static temperature zone division, individual frequency prediction can only achieve proximity allocation without considering temperature, while the integration of the three can achieve forward-looking Pareto optimal layout under dynamic thermal field constraints. This synergistic effect exceeds the sum of the effects of the individual application of each technology.

[0072] The above description is merely a preferred embodiment of the present invention and is not intended to limit the 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 scope of the claims of the present invention.

Claims

1. A method for intelligent optimization of the low-temperature storage location in a cell bank, characterized in that, Includes the following steps: Temperature change data of each storage location in the liquid nitrogen storage device during storage and retrieval operations are collected. The cumulative heat exposure index of each storage location is established by integrating the temperature response of all storage and retrieval operations experienced by each storage location within a preset statistical period. Based on the attribute information of each sample and the progress information of the projects associated with each sample, predict the access probability distribution of each sample within a preset time window. With the optimization objectives of maximizing the access efficiency of each sample and minimizing the cumulative heat exposure of each storage location, and with the constraints of the temperature sensitivity level of the samples and the aggregation constraints of samples in the same batch, a multi-objective combination optimization is performed on the mapping relationship between samples and storage locations to obtain the Pareto optimal layout scheme set. Select a target scheme that meets the preset preference from the Pareto optimal layout scheme set, and generate a sample position adjustment instruction.

2. The method according to claim 1, characterized in that, The method for establishing the cumulative heat exposure index for each storage location includes: modeling each opening operation of the liquid nitrogen storage device as a transient heat transfer process; calculating the peak temperature rise and recovery time constant of each storage location in a single opening operation based on the spatial distance from the opening and the difference in liquid level at each storage location; and integrating the temperature rise curves corresponding to all opening operations experienced by each storage location within the preset statistical period to obtain the cumulative heat exposure value of each storage location.

3. The method according to claim 2, characterized in that, The calculation of the peak temperature rise of each storage location also includes: taking the heat intrusion caused by the samples being accessed in the adjacent storage locations of the target storage location as an additional heat source item; calculating the passive heat exposure increment borne by the target storage location due to adjacent access operations based on the spatial distance between the target storage location and the adjacent storage location and the equivalent thermal conductivity of the filling medium between them; and adding the passive heat exposure increment to the cumulative heat exposure value of the target storage location.

4. The method according to claim 3, characterized in that, The method for predicting the access probability distribution of each sample within a preset time window includes: obtaining the progress node information of the project associated with each sample; when the associated project is detected to enter a preset access triggering stage, the access probability of the corresponding sample is superimposed on the baseline prediction value based on historical data, and the amplitude of the event triggering pulse response is determined by the historical access intensity statistics of the project stage, and the decay rate is determined by the average duration of the project stage.

5. The method according to claim 4, characterized in that, The multi-objective combined optimization also includes: detecting the local curvature of the Pareto front during the Pareto optimization process; when the local curvature exceeds a preset curvature threshold, it is determined that there is a non-convex region; increasing the sampling density within the target space corresponding to the non-convex region and using an adaptive weight strategy to adjust the optimization search direction; wherein the physical meaning of the non-convex region is a subset of schemes where the thermal exposure constraint and the batch aggregation constraint are simultaneously compact.

6. The method according to claim 5, characterized in that, Also includes: Based on the subset of schemes in the non-convex region, the minimum spatial distance between temperature-sensitive samples and high-frequency access samples in each scheme is extracted as the thermal isolation distance, and the maximum spatial span of the storage locations occupied by samples in the same batch is extracted as the batch aggregation radius. A lower limit constraint on the thermal isolation distance and an upper limit constraint on the batch aggregation radius are established, and the joint constraint is fed back to the multi-objective combined optimization as an additional constraint condition for secondary optimization.

7. The method according to claim 1, characterized in that, The preset statistical period is 7 to 90 days. The temperature baseline used in the calculation of the cumulative heat exposure index is the boiling point of liquid nitrogen. Only the period when the temperature rises above the preset temperature threshold is included in the cumulative heat exposure value. The preset temperature threshold is 5 to 20 degrees Celsius above the boiling point of liquid nitrogen.

8. The method according to claim 1, characterized in that, The preset time window is 7 to 30 days from now. The prediction model uses a sliding window method to perform incremental training with a preset update cycle of 1 to 7 days.

9. The method according to claim 1, characterized in that, After the step of selecting the target scheme from the Pareto optimal layout scheme set, the method further includes: calculating the expected liquid nitrogen consumption, expected average sample retrieval time, and expected cumulative heat exposure reduction ratio according to the layout schemes before and after adjustment, and generating a benefit assessment report.

10. A cell bank low-temperature storage location intelligent optimization system, used to implement the method described in any one of claims 1-9, characterized in that, include: The thermal exposure modeling module is used to collect temperature change data of each storage location in the liquid nitrogen storage device during the storage and retrieval operation. Based on the temperature response of each storage location during all storage and retrieval operations within a preset statistical period, the module performs time integration to establish the cumulative thermal exposure index of each storage location. The access frequency prediction module is used to predict the access probability distribution of each sample within a preset time window based on the attribute information of each sample and the progress information of the project associated with each sample. The location optimization module is used to optimize the mapping relationship between samples and storage locations by taking the maximization of the access efficiency of each sample and the minimization of the cumulative heat exposure of each storage location as optimization objectives, and taking the temperature sensitivity level constraint of the sample and the aggregation constraint of the same batch of samples as constraints, to obtain the Pareto optimal layout scheme set. The scheme output module is used to select a target scheme that meets preset preferences from the Pareto optimal layout scheme set and generate a sample position adjustment instruction.

Citation Information

Patent Citations

  • Management system for cell production, storage and transportation, method thereof and sever

    CN107590632A

  • Methods for Assigning Items to Slots for Efficient Access of the Items for Storage and Delivery

    US20130226649A1