A prefabricated dish intelligent preservation management method based on warehouse information

Abstract

The present application relates to the technical field of cold chain logistics, in particular to a kind of pre-cooked meal intelligent preservation management method based on warehousing information, comprising: obtaining the initial state information of pre-cooked meal, and initial state information is imported into warehousing information network, determine the scene factor of each pre-cooked meal under corresponding grid unit;With the scene factor corresponding to pre-cooked meal, extract the scene time series to be identified;Based on the change trend of scene time series, view the difference data set under the scene factor corresponding to each pre-cooked meal;Based on the difference data set of each pre-cooked meal, determine the environmental proportion reference density of each grid unit, and determine the reference time of pre-cooked meal under the corresponding scene factor, obtain the difference of reference time under each scene factor;Calculate the preservation evaluation effect, and determine the risk warning condition corresponding to each pre-cooked meal with the distribution of preservation evaluation effect under the corresponding warehousing information network;The preservation effect of pre-cooked meal is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cold chain logistics, in particular to a pre-prepared dish intelligent preservation management method based on warehouse information. BACKGROUND

[0002] Under the background of the rapid development of the pre-prepared dish industry, warehouse management, as a key link to ensure product quality, faces multiple technical challenges. Traditional warehouse management methods highly rely on manual experience, and through regular sampling inspection or fixed parameter threshold monitoring to identify the preservation condition of pre-prepared dishes. Generally, due to different processing methods, it is difficult for a single threshold alarm mechanism to capture the hidden risks under the synergistic action of multiple factors, resulting in reduced storage preservation effect.

[0003] For example, Chinese Patent Publication No. CN119624283A discloses a pork cold chain preservation control system and storage device. The control system includes a detection unit, a preservation analysis unit, and a cold chain transfer analysis unit. The preservation analysis unit determines the weight loss degree of the pork, the temperature uniformity and humidity uniformity of the placement area at a single time point, determines the temperature and humidity abnormal area, abnormal index and preservation index of the pork at a single time point according to the temperature data and humidity data, and determines the adjustment mode of the cold chain preservation storage device. The cold chain transfer analysis unit is used to construct a pork freezing evaluation model to adjust the placement area and placement transfer time of the pork. The storage device includes a vehicle body and a refrigerated box body, and the refrigerated box body includes an arc curve slide rail, a temperature and humidity adjusting assembly, a spacing adjusting assembly, a detection assembly, an air supply assembly, and a power supply assembly.

[0004] For example, Chinese Patent Publication No. CN117333094A discloses a green crisp plum cold chain logistics preservation control method and system based on a simulation model. Dynamic logistics data is collected; the dynamic logistics data is imported into a simulation model trained in advance to obtain dynamic candidate environment information, which includes dynamic green crisp plum environment information; fresh green crisp plum samples are identified from reference samples to obtain environment information and corresponding gas conditioning ratios of the fresh green crisp plum samples. The reference samples are obtained by recording the logistics storage space with fresh green crisp plum samples. The reference samples correspond to the same recording area as the dynamic logistics data. Green crisp plum logistics information is determined based on the environment information of the fresh green crisp plum samples, the corresponding gas conditioning ratios, and the dynamic green crisp plum environment information.

[0005] In the prior art, it is explained that in the cold chain preservation scenario, the humidity and temperature can be adjusted, and then the gas conditioning ratio and other methods need to be checked. However, the prior art only explains that the method can be used, without considering the synergistic effect of these parameters in the same scenario, resulting in the inability to identify the preservation effect of multiple factor abnormalities, and reducing the efficiency of risk warning and preservation effect in the preservation process. SUMMARY

[0006] In order to solve the above technical problems, the technical scheme adopted by the present application is: a prefabricated dish intelligent preservation management method based on warehouse information, comprising: S1, obtaining the initial state information of each prefabricated dish, and importing the initial state information into the warehouse information network, and determining the scene factor of each prefabricated dish under the corresponding grid unit.

[0007] S2, using the scene factor corresponding to the prefabricated dish, continuously monitoring the state information of the prefabricated dish, and extracting the scene time sequence to be identified; based on the change trend of the scene time sequence, checking the difference data set under the corresponding scene factor of each prefabricated dish.

[0008] S3, based on the difference data set of each prefabricated dish, decomposing the interaction of the prefabricated dish under different scene factors, determining the environmental matching reference density of each grid unit, and based on the environmental matching reference density, determining the reference time of each prefabricated dish under the corresponding scene factor, and obtaining the difference of the reference time under each scene factor.

[0009] S4, based on the environmental matching reference density and the difference of the reference time of the prefabricated dish under different scene factors, calculating the preservation evaluation effect, and determining the risk warning condition corresponding to each prefabricated dish according to the distribution of the preservation evaluation effect under the corresponding warehouse information network.

[0010] The beneficial effects of the present application are: first, the present application decomposes the initial state information of the prefabricated dish into a three-layer architecture of environmental dynamic factor, product state factor and space-time position factor through the warehouse information network; then the abnormal value detection is performed on the related scene factors, the data content combination composed of the abnormal values is taken as the abnormal mode, the abnormal mode that can appear under the current scene is labeled, and the data representation of each grid unit in the warehouse information network under the multi-factor scene is completed.

[0011] Second, the present application locates the abnormal components in the current obtained data by dividing the time window, the trend item related to temperature and humidity, and the abnormal fluctuation item of gas composition, focuses on the specific performance of each data under the scene factor, identifies the form of data scoring under the combination of multiple scene factors, and adjusts the acquisition of the difference data set by using the data of the corresponding abnormal mode in the time window, so as to accurately locate the preservation abnormal area and analyze the causes.

[0012] Thirdly, the application extracts the relative situation and the abnormal situation of each grid unit under the corresponding value by using each scene factor contained in the differential data set to construct the environmental matching vector thereof and obtaining the reference density of each grid relative to the historical data in the form of density clustering, and defines the reference time thereof, identifies the difference of the stored strategy existing in the continuous time window and adjacent days under the value of the scene factor under different conditions, and views the problem of reduced preservation effect caused by the static configuration. Then, the effect of preservation in each grid unit is described by the preservation evaluation effect, and the single-point threshold and neighborhood early warning are performed according to the value, so as to improve the early warning efficiency and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0013] The application will be further described below in combination with the drawings and embodiments.

[0014] Figure 1 It is a flowchart of a pre-prepared meal intelligent preservation management method based on warehouse information.

[0015] Figure 2 It is a flowchart of step S1 of a pre-prepared meal intelligent preservation management method based on warehouse information.

[0016] Figure 3 It is a flowchart of step S2 of a pre-prepared meal intelligent preservation management method based on warehouse information.

[0017] Figure 4 It is a flowchart of step S3 of a pre-prepared meal intelligent preservation management method based on warehouse information.

[0018] Figure 5 It is a flowchart of step S4 of a pre-prepared meal intelligent preservation management method based on warehouse information. DETAILED DESCRIPTION

[0019] The embodiments of the application will be described in detail below. The embodiments described below are exemplary and are only used to explain the application, and cannot be understood as a limitation of the application. If the specific technology or condition is not indicated in the embodiments, the technology or condition described in the literature in the art or according to the product instruction is used.

[0020] Reference Figure 1 A pre-prepared meal intelligent preservation management method based on warehouse information, comprising: S1, obtaining initial state information of each pre-prepared meal, and importing the initial state information into a warehouse information network to determine scene factors of each pre-prepared meal under the corresponding grid unit.

[0021] S2, continuously monitor the state information of the prepared dishes using the corresponding scene factors of the prepared dishes, and extract the scene time series to be identified; based on the change trend of the scene time series, view the difference data set under each prepared dish corresponding scene factor.

[0022] S3, based on the difference data set of each prepared dish, decompose the interaction of the prepared dishes under different scene factors, determine the environmental matching reference density of each grid unit, and based on the environmental matching reference density, determine the reference time of each prepared dish under the corresponding scene factor, and obtain the difference of the reference time under each scene factor.

[0023] S4, based on the environmental matching reference density and the difference of the reference time of the prepared dishes under different scene factors, calculate the preservation evaluation effect, and determine the risk warning condition of each prepared dish based on the distribution of the preservation evaluation effect under the corresponding warehouse information network.

[0024] The initial state information above represents the information marked by the batch, type and taste of each prepared dish before entering the warehouse, to represent the relative situation of the prepared dishes produced under each batch, and then these information is used to search in the warehouse information to determine the specific information of the prepared dishes.

[0025] The scene factor can represent the state of the prepared dishes during preservation, which can be divided into three categories, such as environmental dynamic factor, product state factor and space-time position factor. The scene factors contained in each category are different, and these scene factors will represent the state of the prepared dishes after cutting processing, or the state of the prepared dishes after curing during preservation storage. Since the prepared dishes stored in the current scene are mainly preservation management of fish and other aquatic products, the management method will be more biased towards long-term or focus on meat quality preservation.

[0026] For example, the environmental dynamic factor includes: temperature and humidity, gas composition, light intensity, light intensity represents the relative intensity of ultraviolet inhibition, mainly explains the environmental parameters of the prepared dishes after production, which can affect the preservation time and quality of the prepared dishes, such as temperature >-15℃ accelerates fat oxidation; CO2>20% has good inhibition effect.

[0027] The product state factor will include: initial total number of colonies, water activity, pH value and fat content; these parameters represent the parameters required for product consumption, when some of these parameters are abnormal, the prepared dishes may deteriorate quickly, such as water activity Aw>0.85 easy to breed bacteria, pH<5.5 inhibit spoilage bacteria, etc.

[0028] The space-time position factor represents the cold storage coordinates, stacking density, cold air flow field speed, and storage-related time at the corresponding position, to explain how some positions achieve frozen storage. The stacking density here represents the number of products stacked in a unit volume, and the cold air flow field speed represents the wind speed when the cold air circulates. It is used to explain whether there are corresponding situations such as high corner temperature of the cold storage, airflow dead angle caused by over-dense stacking, and local thawing caused by insufficient wind speed under different stacking conditions of the positions, resulting in a decrease in the frozen preservation effect of the current prepared food.

[0029] After obtaining the above-mentioned scene factors, part of the indicators therein are used as the scene factors for the current judgment, such as using the words directly expressing the current preservation state in temperature and humidity, gas composition as the main scene factors for subsequent judgment.

[0030] As shown in Figure 2 The implementation of step S1 also includes: S11, identifying the initial state information of the prepared food under each grid unit in the corresponding grid unit in the warehouse information network.

[0031] S12, performing data layering processing on the initial state information to obtain the environment dynamic factor, product state factor and space-time position factor of the corresponding grid unit in turn. At this time, the three types of data, environment dynamic factor, product state factor and space-time position factor, are mapped to the corresponding warehouse information network to obtain the data information contained in each grid unit.

[0032] S13, combining the environment dynamic factor, product state factor and space-time position factor according to the position of each grid unit in the warehouse information network to obtain the scene factor of each prepared food in the corresponding grid unit.

[0033] At this time, the scene factor of the prepared food stored in each grid unit is described by combining the environment dynamic factor, product state factor and space-time position factor. In addition to determining these values, it is also necessary to identify whether there are abnormal related problems between the scene factors.

[0034] Therefore, the implementation process of step S1 also includes: determining the standard deviation of each scene factor, marking the data in each scene factor that is more than three times the standard deviation as abnormal data, and combining the grid units corresponding to the abnormal data into an abnormal data set.

[0035] For the abnormal data set of each scene factor, the scene factors are spliced, and the values of the spliced scene factors in the abnormal data set are output as the abnormal mode corresponding to the scene factor.

[0036] When the scene factors are spliced here, all the scene factors currently recognized are combined into a vector or record form according to the position of the abnormality, and these data are combined into a feature matrix about abnormal data to facilitate subsequent processing; at this time, if the rare abnormal data is recognized, the average value in the abnormal data set is selected to replace it, and the average value is selected as the output scene factor, that is, the average value is used to replace the original abnormal value, and the group average is used to replace the individual abnormality to reduce the bias influence; prevent misleading data decision because of partial abnormal data.

[0037] As for the abnormal mode corresponding to the scene factor, it indicates that the current problem is a non-normal fluctuation form such as temperature sudden change, gas proportion imbalance, and the content of the problem is marked at the corresponding scene factor, and then combined with the position of the scene factor in the warehouse information network for output, which is convenient for subsequent monitoring of the scene factor to identify the trend of time change to identify the corresponding change form of the difference part in the time sequence.

[0038] The scene factor at this time can recognize the change of the environment related data through the pre-set temperature and humidity sensor, gas sensor, and then the camera or other equipment is used to monitor or locate the position related data of the prepared food storage, and the product state can be detected by periodically sampling the prepared food. The product state is viewed at a longer time interval compared with the previous two scene factors.

[0039] In an embodiment of the present application, step S2 monitors the prepared food information stored in each grid unit of the warehouse information network in the form of a continuous time period, and identifies the difference part under the value of multiple scene factors such as temperature, humidity, gas composition and wind speed. These parts will represent the different values of the current preservation measure in the continuous time period.

[0040] The scene time sequence above represents the value of the data represented by the scene factor in the time sequence, and the scene factor represents the data of the temperature, humidity, wind speed and light intensity of the prepared food stored with the change of the time sequence. The part of the product state factor associated with the scene factor is not identified at this time. This product state represents the overall preservation of the product. At this time, the relevant data in the environmental dynamic factor and the space-time position factor are analyzed. The space-time position factor is mainly used to describe the change trend in different grid units. This factor is used to identify the change trend in the position of the current grid unit in combination with the environmental dynamic factor, and finally identify which area may be normal and which area may be abnormal in preservation.

[0041] As Figure 3As shown, step S2 further includes, when acquiring the difference data set, S21, using the basic length of the time window, starting from the left end of the scene time sequence, dividing the time growing part, and counting the data and window sliding step length under each time window.

[0042] S22, decomposing the scene time sequence to obtain the trend item corresponding to the temperature and humidity and the abnormal fluctuation item corresponding to the gas composition under each time window.

[0043] S23, based on the position of the grid unit, counting the trend item and the abnormal fluctuation item on different grid units, and giving the abnormal pattern matching degree of the scene time sequence.

[0044] S24, differentiating the data according to the abnormal pattern matching degree of the scene time sequence, and acquiring the difference data set under the scene factor corresponding to the prepared food.

[0045] Preferably, in the current step, the part of the scene factor related to the environmental dynamic factor is processed to identify the change trend of the temperature and humidity and the gas composition, and the ultraviolet light intensity during the storage of the prepared food is a long-time stable state and is only used as an identifier of the prepared food. Then, whether the trends generated at different positions are the same is counted according to the related positions in the scene factor, such as the cold storage coordinates, and whether these generated trends have abnormal patterns is determined, such as the case that the low wind speed at some positions causes abnormal temperature change, and the case that the imbalance of the gas composition corresponding to the air conditioning proportion at some positions causes abnormal temperature change. The possible abnormal parts are identified for the combined abnormal pattern, and whether it is related to the position corresponding to the current scene factor is calculated to determine the storage condition of the prepared food at different positions. Relative to the abnormal pattern identified in step S1, the content of the abnormal pattern matching degree calculation in step S2 is to compare the trend item and the abnormal fluctuation item of each grid unit in the current time window with the abnormal pattern identified in step S1, and to quantify the similarity. The purpose is to determine whether the freshness-keeping state of each region in the current warehouse environment appears to be consistent with the known abnormal pattern, so as to accurately locate the freshness-keeping abnormal region and analyze the cause.

[0046] At this time, the freshness-keeping state of which grid unit is mainly located to be highly matched with the known abnormal pattern, and in which time window is the abnormal pattern likely to appear, so as to finally complete the trend identification in the scene. The difference data set identified thereafter is obtained by differentiating the scene time sequence and screening in combination with the abnormal pattern matching degree, and the core purpose is to extract the dynamic characteristics reflecting the change of the freshness-keeping state in the time sequence and focus on the fluctuation part related to the known abnormal pattern.

[0047] Preferably, the trend item corresponding to the temperature and humidity described above is obtained by linear regression or Holt-Winters method, and the data can represent the change of temperature and humidity in the corresponding time series; the abnormal fluctuation item corresponding to the gas composition represents the residual error of the gas composition corresponding to the gas mixing ratio in the time series, and the residual error is used to judge the fluctuation part exceeding the expected value, and then the trend item and the abnormal fluctuation item corresponding data are calculated with the data labeled as abnormal mode in the scene factor to calculate the matching degree, for example, using the calculation method of Pearson correlation coefficient, or converting the corresponding data into feature vectors to calculate the similarity by cosine similarity, and the similarity calculated with the data labeled as abnormal mode is used as the abnormal mode matching degree at this time. Finally, the data with abnormal mode matching degree greater than 0.8 is used as the differential data to output the differential data set.

[0048] Preferably, when counting the data in each time window and the window sliding step, the processing method further includes: taking the left endpoint of each time window as the starting point, accumulating the data in each time window, determining the time delay of each data point relative to the starting point of the time window, and counting the average time delay and the maximum time delay of the data in the time window; according to the average time delay and the maximum time delay in the time window, the window sliding step is adjusted. When adjusting the window sliding step, if any one of the average time delay and the maximum time delay is greater than a preset time delay threshold, the current window sliding step is multiplied by a screening coefficient, and the attenuation coefficient is less than 1. The attenuation coefficient can be set according to the average value of historical data to improve the resolution of the time series; if it is less than the preset time delay threshold, the current window sliding step can be maintained, or appropriately increased according to the current demand to reduce the data calculation amount. The time delay threshold set at this time is set according to the average time delay and the maximum time delay respectively, and the preset time delay threshold can be set based on the average value of historical data in the last 100 data processing batches, or the moving average value set in the adjacent batch is selected as the preset time delay threshold at this time. It should be noted that the time delay represents the difference between the data arrival time and the window starting time, which is used to represent the position of the corresponding data in the time window.

[0049] The data amount in the current time window is counted, and based on the data amount change rate of the adjacent time window, it is judged that there is a key point in the adjacent time window relative to the current time window, and the key points are divided into each time window according to the position of each key point.

[0050] The data amount change rate calculated at this time is calculated in the current time window and the next time window, and then when there is an absolute value of the data amount change rate greater than a preset change rate threshold, the end position of the current time window is marked as a key point, and then the time window is divided according to the position of the key point, and at this time, the time window division is continuously iterated until all time windows of the scene time sequence are divided. Through dynamic monitoring of the time delay and the data amount change, the window division strategy can adapt to the characteristics of the data stream, such as burst traffic, periodic fluctuations, and the like.

[0051] Preferably, the time window represents a data segment of a fixed time length intercepted from the scene time sequence, such as a five-minute window; the window sliding step represents the time interval of each movement of the time window, and at this time, the maximum value of the window sliding step is less than half of the time window, so as to obtain a time sequence in the form of sliding as much as possible. The basic length of the time window is used to represent the basic window size and the sliding step of the time sequence in processing, so as to facilitate subsequent dynamic adjustment of the window size and the sliding step.

[0052] Preferably, when the scene time sequence is decomposed into the trend item and the abnormal fluctuation item, the mapping relationship between the trend item of temperature and humidity and the abnormal fluctuation item of gas composition needs to be determined, the preset mapping relationship in the scene time sequence is iterated, and the trend item and the abnormal fluctuation item on different grid units are counted according to the value range under each mapping relationship; that is, the data counted include but are not limited to the slope, the median, the mean value and the fluctuation degree relative to the mean value of temperature and humidity, and the fluctuation degree is represented as the absolute average value of each data point and the mean value thereof; and the average value, the maximum value and the conditional probability of imbalance of the proportion of gas composition; after these data are counted, the similarity between the part corresponding to these values and the data in the scene factor identified as corresponding to the abnormal mode is calculated, and then the difference data set to be processed subsequently is extracted according to the similarity.

[0053] When the difference data set is output in step S24, the implementation manner further includes: if there is intersection of the difference data sets, the time window corresponding to the difference data set is queried, the position difference of the scene factor contained in the time window is used to identify the index value of each difference data set, and the difference between the left end statistic and the right end statistic of the index value in the time window is used as the data output by the difference data set.

[0054] When multiple difference datasets have overlaps in the time dimension, i.e., their time windows intersect, the system automatically identifies these overlapping time windows and performs independent analysis for each overlapping window. Within each overlapping time window, the system compares the difference data of different grid cells for each scenario factor. These scenario factors can include temperature, humidity, gas composition, etc. Through comparison, the system can identify which scenario factors have significant differences in which locations and quantify these differences as specific indicator values, where the indicator value is the statistical value when the trend term and abnormal fluctuation term on different grid cells are counted. For each identified indicator value, the system calculates its statistical value at the start and end points of the overlapping window. Then, the system calculates the difference between the two statistical values as the change amount of the indicator value within the time window. This change amount is the difference data output by the overlapping window, which reflects the degree of change in scenario factors caused by spatial location differences within a specific time window. In this way, the system can provide more detailed data support to help users better understand the dynamic changes in the warehouse environment.

[0055] Preferably, the difference used by the difference dataset includes, but is not limited to, first-order difference, high-order difference, etc., to obtain difference datasets in multiple dimensions.

[0056] In an embodiment of the present application, when processing the difference dataset matching the abnormal pattern in step S3, the interaction of each prepared dish in the abnormal and stable states is determined by viewing the environmental proportion at different locations and the reference time, where the environmental proportion represents the values of temperature, humidity, gas, and corresponding light intensity under corresponding conditions, and the duration of these environmental proportions is considered as the reference time. The environmental proportion reference density can reflect the change law under dynamic environment, making the warehouse environment control more flexible and adaptive. The reference time determination can predict the shelf life of the prepared dish under the current environment, providing a basis for inventory management and sales planning.

[0057] Suppose when processing the difference dataset of a certain cold storage, it is first decomposed into temperature difference sequence and humidity difference sequence. Then, the average value of temperature is calculated as 2°C and the average value of humidity is calculated as 60% using statistical methods, as the reference density of environmental proportion. Next, through time series analysis, it is predicted that the temperature will rise to 5°C within the next 24 hours, exceeding the safety threshold of 4°C for prepared dishes. Therefore, the reference time of the prepared dish is determined to be 24 hours, prompting the management personnel to take measures such as cooling within 24 hours to extend its shelf life.

[0058] As Figure 4As shown, the implementation of step S3 includes: S31, based on the difference data set of each prepared dish, identifying the target scene factor under different interaction conditions, decomposing the difference data set according to the target scene factor, and generating independent difference sequences of each target scene factor, including but not limited to humidity difference sequence, temperature difference sequence and gas difference sequence. At this time, the proportion of carbon dioxide, oxygen and nitrogen under the current refrigeration is mainly considered to store roasted fish prepared dishes and the like. The gas difference sequence takes gas concentration as its ordinate and time point as its abscissa, and displays the concentrations of carbon dioxide, oxygen and nitrogen in the gas difference sequence.

[0059] S32, based on the independent difference sequence of the target scene factor, describing the environmental matching vector of each grid unit under the warehouse information network.

[0060] S33, density query is performed on the environmental matching vector of each grid unit under the warehouse information network to obtain the environmental matching reference density of each grid unit.

[0061] S34, time series prediction is performed based on the environmental matching vector and the environmental matching reference density of each grid unit to set the reference time of each grid unit.

[0062] Preferably, the environmental matching reference density can be based on the clustering analysis of the environmental matching vectors in the current data and historical data, and the environment matching clusters with similar preservation effects are identified, and then the value of the cluster center is selected as the environmental matching reference density. Through clustering analysis, environment change patterns with similar preservation effects can be identified, which means that several typical ideal environment condition combinations can be extracted from a large amount of environment monitoring data, and then these reference values are compared with the current values. When the current parameters deviate from the center of the cluster, a warning mechanism is triggered, that is, the environment conditions are adjusted to complete the storage management of the prepared dishes. The clustering processing can improve the personalized processing form for different prepared dishes, so that the prepared dishes can maintain an ideal state as much as possible during storage to improve the preservation effect.

[0063] The environmental matching vector can include the slope, median, mean, and fluctuation degree of the mean of the temperature and humidity in the differential sequence, the proportion of the change of the gas component, and the conditional probability in the case of imbalance, to represent the value of each grid cell in the occurrence of temperature and humidity mutation and gas proportion imbalance; then the values are normalized to eliminate the dimension of the data statistics, and the environmental matching vector in the historical data is clustered by combining the data between single points, and the cosine similarity between the environmental matching vectors is continuously clustered by hierarchical clustering or K-Means clustering, until the cluster cluster formed presents a stable form, and then the value of the environmental matching vector at the center of the cluster cluster in the historical data is used as the environmental matching reference density at this time.

[0064] Preferably, in obtaining the environmental matching vector, the ratio of the independent differential sequence of each grid cell is obtained, and the mean, fluctuation degree, and extreme value of each grid cell are used as the value of each environmental matching vector. At this time, the humidity differential sequence, temperature differential sequence, and gas differential sequence are used as the three elements of the environmental matching vector, and the mean, fluctuation degree, and extreme value of the humidity differential sequence, temperature differential sequence, and gas differential sequence are calculated respectively; then the values are normalized and used as the environmental feature vector.

[0065] In identifying the reference time, the time interval from the start time of the time window to the time when the environmental matching vector value exceeds the environmental matching reference density is calculated as the reference time, the time length of the reference time is judged, and the specific reference value of the environmental matching vector that exceeds the environmental matching reference density is judged. At the same time, it is also necessary to identify the multivariate trend regression of the environmental matching vector, identify the trend slope, and judge the time length of the current environmental matching vector exceeding the environmental matching reference density to explain the reference time. The multivariate linear regression is calculated by using the humidity differential sequence, temperature differential sequence, and gas differential sequence as the independent variable, and the characteristic value of the environmental matching vector as the dependent variable, to obtain the trend of the multivariate linear regression. Then, the trend obtained is used to predict the value of the environmental matching vector at the next time and whether it exceeds the environmental matching reference density, so as to obtain the reference time.

[0066] Therefore, the implementation mode of step S34 further includes: based on the time period when any characteristic value of the environmental matching vector exceeds the characteristic value of the environmental matching reference density, the reference time of each grid cell in the corresponding time window is viewed.

[0067] The reference time difference of the reference time of the current day and the reference time of other days corresponding to the same time window in the same scene time sequence position is determined, the reference time of all time windows is traversed, and the reference time difference is summed as the difference of the reference time of each scene factor.

[0068] If the current reference time conflicts due to the overlap of the time window, the reference time corresponding to the time point is removed from the overlapping part to determine the reference time under the corresponding time window, and then when the reference time difference is correlated with the scene factor, the existing reference time difference and the scene factor are mapped, and the sum of the reference time differences under the same scene factor is taken as the difference of the reference time under the scene factor.

[0069] The reference time obtained at this time is the time after judging that the current prepared food can maintain normal preservation in the preservation state, and checking whether the normal preservation time exists difference in the same time window of different days. If there is difference, it may indicate that the current cold storage prepared food mode has a problem, and the control strategy needs to be adjusted, and the conditions leading to the decrease of the preservation effect of the current prepared food are found out according to the difference in time, and the warning control is performed according to the conditions to prevent the problem of decrease of the preservation degree of the prepared food in cold storage.

[0070] In an embodiment of the present application, step S4 displays the environment matching and reference time of each grid unit, and performs multi-alarm on each grid unit. The preservation evaluation effect is based on the value of the environment matching reference density and the time length of the duration and distribution of the reference time under the current time sequence to evaluate whether it is in a normal preservation state. Then, the preservation evaluation effect is evaluated according to the time length. The preservation evaluation effect at this time will be used as a semantic evaluation parameter to indicate whether the current prepared food can achieve the expected preservation time under the corresponding environment matching reference density and reference time. Finally, according to the different values, it is indicated that the part should be warned of risk, and the preservation management of the prepared food is finally completed.

[0071] Preferably, when calculating the preservation evaluation effect, the difference proportion of the current environment matching vector and the environment matching reference density is weighted and summed, and then 1 is subtracted from the weighted sum to represent the value of the preservation effect. The greater the value, the lower the risk, and the closer to the normal preservation state. The smaller the value, the greater the risk, and the current selected scene factor needs to be triggered in time, and the multi-regions associated with each scene factor are cooperatively triggered to complete the preservation storage of each prepared food in the warehouse.

[0072] Preferably, the weight of the preservation evaluation effect is set based on the current scene factor affecting the weight of the preservation effect, for example, the deviation of the initial colony total number, water activity, pH value and fat content corresponding to the scene factor measured relative to the standard value is taken as the weight, and the standard value is set according to the value of the preservation requirement. This standard value can be set to different values based on different prepared dishes. Since the deviation degree at this time contains multiple values, the values need to be normalized, and the cosine similarity relative to the standard value is calculated. That is, the initial colony total number, water activity, pH value and fat content in the scene factor at this time are input to obtain the deviation degree; or the difference ratio of the current environmental matching vector is additionally set as a superscript of the sensitivity coefficient, and the sensitivity coefficient is set according to the accuracy of the current scene factor in evaluating the preservation evaluation effect.

[0073] As shown in Figure 5 The implementation mode of step S4 includes: S41, based on the difference of the reference time, screening the grid cells located in the same reference time difference, and calculating the preservation evaluation effect of each grid cell.

[0074] S42, clustering based on the preservation evaluation effect value of each grid cell, clustering each grid cell to form a risk clustering cluster. At this time, the clustering is performed according to the distance between each grid cell, and then each risk clustering cluster after clustering needs to meet the minimum sample number, for example, 500 meters or less is taken as the neighborhood radius when the risk clustering cluster is clustered, and 5 is taken as the minimum neighborhood sample number of each risk clustering cluster. DBSCAN clustering method is used for clustering.

[0075] S43, based on the preservation evaluation effect value of each grid cell in the risk clustering cluster, triggering the single-point risk warning condition and the neighborhood risk warning condition on the grid cell, and outputting the risk warning condition of the prepared dish under different scene factors.

[0076] At this time, the risk warning condition is selected based on the value of the preservation evaluation effect value under different scene factors, for example, 0.6 and 0.8 are taken as the division value, and the grid cells in the risk clustering cluster with a preservation evaluation effect value less than 0.6, 0.6 to 0.8 and greater than 0.8 are taken as high-risk areas, medium-risk areas and low-risk areas in turn. At this time, 0.6 and 0.8 are only used to illustrate the interval of the division based on the preservation evaluation effect value, and the value can also be set according to the average value used to divide the high-risk area, the medium-risk area and the low-risk area in the historical data.

[0077] Preferably, the single-point risk warning condition indicates that there is only one scenario factor value in a certain grid cell or only one grid cell in multiple grid cells is abnormal, and then the abnormal part of the corresponding data of temperature, gas composition, humidity, etc. based on the grid cell is filtered from the database to obtain the risk warning condition at the location, and then the related grid cell is alarmed to facilitate subsequent adjustment of the storage mode of the prepared food. The neighborhood risk warning condition indicates that the adjacent grid cells have abnormal data at the same time, and the corresponding risk warning condition is extracted from the database when the adjacent grid cells have synchronous decline in the freshness evaluation effect value. If the adjacent grid cells still have partial data abnormality but not at the same time after processing the simultaneously abnormal data components, the single-point risk warning condition is used for separate processing, and then the two risk warning conditions are used as the output risk warning condition.

[0078] At this time, the output risk warning condition also needs to be output according to the position corresponding to the time window, and the implementation mode of step S43 further includes: according to the risk warning condition of the prepared food under different scenario factors, the time window corresponding to the risk warning condition is matched with the reference time in the time window to output the matched risk warning condition.

[0079] At this time, the output risk warning condition also needs to be output according to the position corresponding to the time window, and the implementation mode of step S43 further includes: according to the risk warning condition of the prepared food under different scenario factors, the time window corresponding to the risk warning condition is matched with the reference time in the time window to output the matched risk warning condition.

[0080] Such processing can also update the freshness evaluation effect value of different grid cells in time according to the data in the time window to dynamically evolve the threshold value, and can also bind the current value with a specific time period to reduce the false alarm rate in actual alarm, timely respond to the existing risk situation, and finally complete the preservation of the prepared food.

[0081] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application, which are still covered by the protection scope of the present application.

Claims

1. An intelligent fresh-keeping management method for pre-prepared dishes based on storage information, characterized in that: include: S1, obtaining the initial state information of each pre-prepared dish, and importing the initial state information into the warehouse information network to determine the scene factor of each pre-prepared dish in the corresponding grid unit; S2, using the scene factors corresponding to the pre-prepared dishes, continuously monitors the status information of the pre-prepared dishes and extracts the time series of the scenes to be identified; Based on the changing trend of the scene time series, view the differential data set under the corresponding scene factors of each pre-prepared dish; the scene time series represents the value of the data represented by the scene factor in the time series, and the scene factor represents the data of the temperature, humidity, wind speed, and light intensity stored in the pre-prepared dish that changes with the time series; S3: Based on the differential datasets of each pre-prepared dish, the interaction between the pre-prepared dishes under different scenario factors is decomposed to determine the reference density of the environmental ratio of each grid cell. Based on the reference density of the environmental ratio, the benchmark time of each pre-prepared dish under the corresponding scenario factor is measured to obtain the difference in the benchmark time under each scenario factor; S4: Based on the differences in environmental ratio reference density and benchmark time of pre-prepared dishes under different scenario factors, calculate the freshness evaluation effect, and determine the corresponding risk warning conditions for each pre-prepared dish based on the distribution of the freshness evaluation effect in the corresponding warehouse information network; The implementation of step S3 includes: S31, based on the differential data sets of each pre-prepared dish, identifying target scenario factors under different interaction situations, decomposing the differential data sets according to the target scenario factors, and generating independent differential sequences of each target scenario factor; S32, based on the independent difference sequence of the target scene factor, describes the environmental ratio vector of each grid cell in the warehouse information network; S33, performing a density query based on the environmental ratio vector of each grid unit in the warehouse information network to obtain the environmental ratio reference density of each grid unit; performing cluster analysis based on the environmental ratio vectors in the current data and historical data to identify environmental ratio clusters with similar preservation effects, and selecting the value of the cluster center as the environmental ratio reference density; S34, performing time series prediction based on the environmental ratio vector and the environmental ratio reference density of each grid cell, setting a reference time for each grid cell; calculating the time interval from the start of the time window to the time when the environmental ratio vector value exceeds the environmental ratio reference density as the reference time; The implementation of step S34 also includes: Based on the time period when any eigenvalue in the environmental ratio vector exceeds the eigenvalue of the environmental ratio reference density, check the benchmark time of each grid cell in the corresponding time window; Determine the benchmark time difference between the benchmark time corresponding to the current day and other days in the same time window at the same scene time series position, traverse the benchmark time of all time windows, and sum the benchmark time differences as the difference of the benchmark time under each scene factor.

2. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 1, characterized in that: The implementation of step S1 further includes: S11, using the corresponding grid units in the warehouse information network, identifying the initial state information corresponding to the pre-prepared dishes in each grid unit; S12, performing data layering processing on the initial state information to obtain the environmental dynamic factors, product state factors and spatiotemporal location factors under the corresponding grid units in sequence; S13, combining the environmental dynamic factor, the product status factor, and the spatiotemporal location factor according to the position of each grid unit in the warehouse information network to obtain the scene factor of each pre-prepared dish in the corresponding grid unit.

3. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 1, characterized in that: The implementation process of step S1 also includes: Determine the standard deviation of each scenario factor, mark the data exceeding three times the standard deviation in each scenario factor as abnormal data, and combine the grid cells corresponding to the abnormal data into an abnormal data set; For the abnormal data set of each scenario factor, each scenario factor is spliced ​​together, and the abnormal pattern corresponding to the scenario factor is output based on the value of each scenario factor in the abnormal data set after splicing.

4. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 3 is characterized in that: The implementation of step S2 further includes: S21, using the basic length of the time window, starts from the left end of the scene time series and splits it towards the part with increasing time, and counts the data under each time window and the window sliding step; S22, decomposing the scene time series to obtain the trend item corresponding to the temperature and humidity and the abnormal fluctuation item corresponding to the gas composition in each time window; S23, based on the location of the grid cells, counts the trend items and abnormal fluctuation items on different grid cells, and gives the abnormal pattern matching degree of the time series of the given scenario; The abnormal pattern matching degree is calculated by comparing the trend item and abnormal fluctuation item of each grid cell in the current time window with the abnormal pattern identified in step S1 to quantify their similarity; S24, performing data differentiation based on the abnormal pattern matching degree of the scene time series to obtain a differential data set under the scene factors corresponding to the pre-prepared dishes; using data with an abnormal pattern matching degree greater than a set threshold as the data for differentiation to output its differential data set.

5. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 4 is characterized in that: When counting the data in each time window and the window sliding step, the processing method also includes: Taking the left endpoint of each time window as the starting point, accumulate data for each time window, determine the latency of each data point relative to the start point of the time window, and calculate the average and maximum latency of the data within the time window. Adjust the window sliding step size based on the average and maximum latency within the time window. Count the data volume in the current time window, and based on the data volume change rate of adjacent time windows, determine whether there are key points in the adjacent time windows relative to the current time window, and divide each time window according to the location of each key point.

6. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 4 is characterized in that: The implementation of step S24 further includes: If there is an intersection of differential data sets, the time window corresponding to the differential data set is used for query. The index value of each differential data set is identified by the position difference of the scene factors contained in the time window, and the difference between the index value statistics at the left end and the right end of the time window is used as the data output by the differential data set.

7. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 1, characterized in that: Therefore, the implementation of step S4 includes: S41, based on the difference in the benchmark time, screen the grid cells with the same benchmark time difference and calculate the preservation evaluation effect of each grid cell; S42, clustering the grid units based on the preservation evaluation effect value of each grid unit to form risk clusters; S43, based on the freshness assessment effect value of each grid unit in the risk cluster, triggers the single-point risk warning condition and the neighborhood risk warning condition on the grid unit, and outputs the risk warning conditions of pre-prepared dishes under different scenario factors.

8. The method for intelligent fresh-keeping management of pre-prepared dishes based on storage information according to claim 7, characterized in that: The implementation of step S43 further includes: According to the risk warning conditions of pre-prepared dishes under different scenario factors, for the time window corresponding to the risk warning conditions, the data of each risk warning condition is matched with the reference time in the time window, and the matched risk warning conditions are output.

Images