A digital operation system suitable for cakes

Abstract

The application discloses a kind of digital operation system suitable for pastry, it is related to food processing digital management technical field, the system includes the intelligent production terminal of central factory, the store sales terminal of each store containing environmental sensor and electronic price tag, and cloud operation platform, in cloud operation platform, data convergence module integrates sales, inventory, environment and external data, demand prediction module exports each item forecast, production scheduling module generates production work order accordingly, shelf life dynamic management module according to the standard shelf life of each pastry item preset and the environmental parameters collected by the environmental sensor, dynamically calculate the actual remaining saleable length of each batch of pastry, dynamic pricing module obtains the length and inventory level linkage, generates real-time discount rate and pushes to electronic price tag, and intelligent allocation module generates allocation instruction according to inventory surplus and inventory gap, to reduce pastry loss, improve the purpose of operating efficiency and profit margin.

Description

Technical Field

[0001] This invention relates to the field of digital management technology for food processing, specifically a digital operation system applicable to pastries. Background Technology

[0002] Pastries are characterized by short shelf life, fluctuating demand, numerous varieties, and relatively sensitive production processes. Under the traditional operating model, store production planning and inventory management rely heavily on manual experience, making it difficult to accurately respond to sudden changes in demand caused by external factors such as weather, holidays, and commercial events. This often leads to the dilemma of stockouts for best-selling items and large-scale losses of slow-moving items. At the same time, the quality of pastries is greatly affected by the temperature and humidity of the storage environment. Fixed shelf life on labels cannot reflect the true state of product quality deterioration, and operational decisions lack effective dynamic basis.

[0003] In existing technologies, general retail management systems often adopt a uniform fixed shelf-life management approach, without using real-time environmental parameters such as temperature and humidity collected from stores to dynamically adjust the remaining shelf life. This results in pastries that have not yet reached the label printing time experiencing quality degradation due to high temperature and humidity. If they continue to be sold normally, it will lead to customer complaints, miss the best promotional window when the actual quality is still acceptable, and exacerbate unnecessary product losses. In cross-store transfer scenarios, existing systems only match inventory with demand gaps, without considering whether the actual remaining saleable time of the batch to be transferred is sufficient to support transportation and sales in the receiving store. This easily leads to ineffective transfers and secondary losses. Dynamic pricing strategies are also mostly based on discounts at fixed time points, failing to link with real-time shelf-life status and real-time inventory levels. Pricing is crude, and the various links are disconnected from each other.

[0004] Therefore, how to build a dedicated digital operation system for pastries that can fully perceive the store environment, dynamically calculate the actual remaining sales time of each batch of pastries based on real-time temperature and humidity data, and use this as the core to drive intelligent allocation and dynamic pricing across stores, so as to achieve accurate supply and demand matching and minimize losses, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the high losses and low efficiency issues in existing pastry operations caused by environmental sensitivity, static shelf-life management, and fragmented decision-making processes, this invention designs a digital operation system suitable for pastries. This system includes intelligent production terminals in a central factory, retail sales terminals in each store equipped with environmental sensors and electronic price tags, and a cloud-based operation platform. Within the cloud platform, a data aggregation module integrates sales, inventory, environmental, and external data; a demand forecasting module outputs predicted quantities for each item; a production scheduling module generates production orders based on these predictions; a shelf-life dynamic management module dynamically calculates the actual remaining saleable time for each batch of pastries based on preset standard shelf-life for each item and environmental parameters collected by the environmental sensors; a dynamic pricing module obtains this time and links it with inventory levels to generate real-time discount rates that are pushed to electronic price tags; and an intelligent allocation module generates allocation instructions based on inventory surplus and shortage. This achieves a synergistic effect of environmentally adaptive shelf-life management, precise supply and demand matching, dynamic pricing and clearing, and efficient and safe allocation, thereby reducing pastry losses and improving operational efficiency and profit margins.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a digital operation system suitable for pastries, the system comprising:

[0007] At least one intelligent production terminal, located in the central factory, is used to receive production work orders and execute pastry baking production;

[0008] Multiple store sales terminals are set up in each store. Each store sales terminal includes a POS cashier module for collecting pastry sales data, an environmental sensor for real-time monitoring of environmental parameters, and an electronic price tag for displaying pastry prices and discount information.

[0009] The cloud-based operation platform communicates and connects with the intelligent production terminals and various store sales terminals via the network.

[0010] The cloud-based operation platform includes: a data aggregation module, a demand forecasting module, a production scheduling module, an intelligent allocation module, a shelf-life dynamic management module, and a dynamic pricing module.

[0011] Furthermore, the data aggregation module is used to collect and aggregate the sales data, inventory data, environmental parameters, and external data of each store in real time;

[0012] The input terminal of the demand forecasting module is connected to the data aggregation module. Based on the sales data and external data collected by the data aggregation module, the demand forecasting module predicts the demand for each pastry item in each store within a preset time period.

[0013] Furthermore, the process by which the demand forecasting module predicts the demand for each pastry item is as follows:

[0014] The historical sales sequence of each pastry item in each store within a preset period is obtained from the data aggregation module. The historical sales sequence records the daily sales volume of each pastry item with a daily time granularity.

[0015] The external data is obtained from the data aggregation module, including weather type data, temperature data, holiday identification data, and promotional activity data.

[0016] The historical sales series were decomposed using STL time series decomposition to obtain trend component series, seasonal component series and residual component series;

[0017] The trend component sequence and the seasonal component sequence are used as the basic prediction features, and the weather feature vector after the weather type data is individually thermally encoded, the temperature data, the holiday identification data and the promotional activity data are used as auxiliary prediction features. The prediction feature vector is constructed by splicing them together.

[0018] The predicted feature vector is input into a pre-trained gradient boosting decision tree model, which outputs the preliminary demand prediction values ​​for each pastry item in each store on each day within a preset time period.

[0019] The residual component sequence is input into a pre-trained long short-term memory neural network model, which outputs the residual prediction values ​​of each pastry item in each store for each day within a preset time period.

[0020] The preliminary demand forecast and the residual forecast are superimposed daily to obtain the final demand for each pastry item in each store on each day within a preset time period.

[0021] Furthermore, the input end of the production scheduling module is connected to the demand forecasting module, and the output end of the production scheduling module is connected to the intelligent production terminal. The production scheduling module is used to generate and issue the production work order based on the demand output by the demand forecasting module, the capacity of the central factory, and the raw material inventory. Specifically, it includes:

[0022] The final demand for each pastry item at each store within a preset time period is summarized to obtain the total production demand for each pastry item at the central factory within the preset time period.

[0023] Obtain the current raw material inventory data of the central factory, which includes the current inventory level and availability status of each raw material;

[0024] Based on the recipe data of each pastry item, the total production demand is calculated back into the demand consumption of each raw material, and it is determined whether the current inventory of each raw material meets the demand consumption.

[0025] If the current inventory level is insufficient to meet the demand consumption, the total production demand is reduced proportionally according to the shortage ratio of the raw materials to obtain the adjusted total production demand.

[0026] Obtain equipment capacity data from the central factory, including equipment identification, equipment type, unit time capacity, and available time window for each production equipment.

[0027] With the dual objectives of minimizing production changeovers and maximizing equipment utilization, and with the delivery time constraints of each pastry item as a hard constraint, a genetic algorithm is used to schedule and optimize the total production demand after adjustment, generating a production sequence for each pastry item executed by each production equipment within each time window.

[0028] The production sequence of the pastry items is packaged into the production work order. The production work order includes the work order number, pastry item identifier, production quantity, target production equipment identifier, planned start time and planned end time, and the production work order is sent to the intelligent production terminal.

[0029] Furthermore, the process of the genetic algorithm performing scheduling optimization calculations is as follows:

[0030] The execution order of each pastry production task on each production equipment is encoded as a chromosome individual. Each chromosome consists of multiple gene segments, and each gene segment corresponds to the sequence of pastry production tasks arranged in the execution order on a production equipment.

[0031] Construct a fitness function, wherein the fitness function is: ,in, For the fitness function value, This refers to the total number of production switchovers across all production equipment. A production switchover refers to a situation where two adjacent pastry production tasks on the same production equipment have different pastry item identifiers. This refers to the total number of production equipment. For the first The actual production time occupied by each production equipment For the first Total available time window duration for each production unit. and These are the production changeover penalty weight and the equipment utilization weight, respectively. ;

[0032] With maximizing the fitness function value as the optimization direction, the population is subjected to selection, crossover, and mutation operations for iterative evolution. The iteration terminates when the preset maximum number of iterations is reached, and the production sequence of the pastry item corresponding to the chromosome with the largest fitness value is output.

[0033] Furthermore, the input end of the shelf life dynamic management module is connected to the data aggregation module, and the shelf life dynamic management module obtains the production completion timestamp of each batch of pastries and the preset standard shelf life of each pastry item;

[0034] The data aggregation module acquires environmental parameter sequences collected by the environmental sensors of the corresponding stores at a preset sampling period. These environmental parameter sequences include an ambient temperature sequence and an ambient humidity sequence. Based on the ambient temperature sequence and the ambient humidity sequence, a quality degradation rate factor for each batch of pastries is calculated. ,in, for The quality degradation rate factor at any given time. for The ambient temperature value collected by the environmental sensor at that time. for The ambient humidity value collected by the environmental sensor at that time. The standard temperature value under the aforementioned standard storage environment conditions. The standard humidity value under the aforementioned standard storage environment conditions. For temperature sensitivity coefficient, The humidity sensitivity coefficient is preset according to the type of pastry item;

[0035] Calculate the cumulative quality degradation of each batch of pastries based on the quality degradation rate factor. ,in, This represents the cumulative quality degradation from the completion of production to the present moment. The production completion timestamp, For the current moment, The preset sampling period;

[0036] Based on the cumulative quality degradation and the standard shelf life, calculate the actual remaining marketable time for each batch of pastries: ,in, The actual remaining saleable time, This is the preset upper limit threshold for quality degradation. The standard shelf life is the duration stated.

[0037] Furthermore, the input end of the dynamic pricing module is connected to the shelf life dynamic management module and the data aggregation module, and the output end of the dynamic pricing module is connected to the electronic price tag. The dynamic pricing module is used to obtain the actual remaining sales time from the shelf life dynamic management module, and to obtain the current inventory quantity of each pastry item in each store and the real-time customer flow data of each store in the current time period from the data aggregation module.

[0038] Based on the actual remaining saleable time and the standard shelf life, the shelf life stage of each batch of pastries is determined, including the normal stage and the near-expiration warning stage.

[0039] When the actual remaining saleable time Greater than or equal to the preset pre-sale threshold At that time, the pastry item is in the normal stage, and the real-time discount rate remains the original price discount rate;

[0040] When the actual remaining saleable time Less than the preset pre-sale threshold When the pastry item is in the near-expiration warning stage, the real-time discount rate is calculated based on the actual remaining sales time, the current inventory quantity, and the real-time customer flow data.

[0041] The dynamic pricing module generates a real-time discount rate and pushes it to the electronic price tag of the corresponding store via the network. The electronic price tag receives and updates the original price and discounted price of the corresponding pastry item.

[0042] Furthermore, when a pastry item is in the near-expiration warning stage, a near-expiration urgency index is calculated based on the actual remaining sales time, the current inventory quantity, and the real-time customer traffic data. ,in, The aforementioned urgency index, The actual remaining saleable time, To preset the pre-sale threshold, This represents the current inventory quantity of pastry items at the corresponding store. This refers to the average daily sales volume of pastry items for a predetermined number of days at the corresponding stores. Assuming inventory pressure as the weighting factor, obtain the real-time passenger flow data for the current time period and calculate the passenger flow correction factor: ,in, The passenger flow correction factor is... This is the real-time passenger flow data for the current time period. This represents the historical average customer traffic for the corresponding store during the same period.

[0043] The real-time discount rate is calculated based on the urgency index and the passenger flow correction factor. ,in, For real-time discount rates, This is the preset lower limit of the discount rate. To maximize the function, ensure that the real-time discount rate is not lower than the lower limit of the discount rate.

[0044] Furthermore, the input end of the intelligent allocation module is connected to the demand forecasting module and the data aggregation module. The intelligent allocation module obtains the real-time inventory of each pastry item in each store from the data aggregation module. The real-time inventory includes the inventory quantity of each batch of pastries and the actual remaining sales time. It also obtains the final demand of each pastry item in each store within a preset time period from the demand forecasting module.

[0045] For each pastry item, calculate the inventory surplus or inventory deficit for each store:

[0046] When the real-time inventory of pastry item M in a store is greater than the final demand and the excess exceeds the preset transfer trigger threshold, the store is marked as a candidate store for transfer of pastry item M, and the inventory surplus is calculated.

[0047] When the real-time inventory of pastry item M in a store is less than the final demand, the store is marked as a candidate store for pastry item M, and the inventory gap is calculated.

[0048] Based on the inventory surplus and inventory deficit, a pastry transfer instruction is generated. The pastry transfer instruction includes the sending store identifier, receiving store identifier, pastry item identifier, transfer batch number, transfer quantity, and transfer time window. The pastry transfer instruction is then pushed to the corresponding store sales terminals of the sending and receiving stores.

[0049] Compared with existing technologies, this digital operation system for pastries has the following advantages:

[0050] I. This invention utilizes environmental sensors deployed at store sales terminals to collect real-time temperature sequences of the pastry storage environment. A dynamic shelf-life management module then dynamically obtains the actual remaining saleable time for each batch of pastries, objectively reflecting the accelerated quality degradation caused by environmental temperature fluctuations. This makes the triggering of near-expiration warnings more precise. Store managers can use this information to prioritize clearing out pastries nearing their quality limits, avoiding the sale of products that have deteriorated in quality but have not yet reached their label expiration date. This effectively reduces the risk of customer complaints, minimizes product losses due to expiration, and improves the overall economic efficiency of pastry operations.

[0051] Second, the dynamic pricing module of this invention obtains the actual remaining saleable time of each pastry item for the current batch from the shelf-life dynamic management module, and obtains the real-time inventory level of the store from the data aggregation module and the predicted demand output from the demand forecasting module. It calculates and generates a real-time discount rate, and pushes the real-time discount rate to the electronic price tag of the corresponding store for display in real time. This pricing strategy closely combines the degree of expiration with the market supply and demand relationship. When the pastries still have good quality, they attract consumers to buy them at attractive prices, forming a rapid clearance under the guidance of price signals. This effectively smooths demand, reduces price discount losses and final write-offs, and enables stores to maximize profits while ensuring customer experience.

[0052] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0054] Figure 1 This is a data flow diagram of a digital operation system for pastries according to an embodiment of the present invention;

[0055] Figure 2 This is a flowchart of the demand forecasting module in an embodiment of the present invention.

[0056] Figure 3 This is an overall flowchart of a digital operation system for pastries according to an embodiment of the present invention. Detailed Implementation

[0057] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0058] In this embodiment, as Figure 1As shown, a digital operation system for pastries includes at least one intelligent production terminal, multiple store sales terminals, and a cloud operation platform. The cloud operation platform includes: a data aggregation module, a demand forecasting module, a production scheduling module, an intelligent allocation module, a shelf-life dynamic management module, and a dynamic pricing module.

[0059] The intelligent production terminal is deployed in the production workshop of the central factory. This intelligent production terminal is an industrial tablet computer or embedded industrial control computer that integrates functions such as work order reception, barcode scanning, production reporting, and equipment status reporting. It establishes a two-way communication connection with the cloud operation platform through a wired or wireless network. After receiving the production work order issued by the cloud, the terminal guides the production line equipment to execute production according to the specified pastry items, production quantities, and process parameters through a visual interface. After each production batch is completed, the terminal sends production completion information back to the cloud operation platform by scanning the raw material batch code or entering the production batch number. This information includes the production work order number, actual production quantity, production start and end timestamps, and the corresponding batch number.

[0060] Each store's sales terminal is deployed in the sales and storage area of ​​each chain store. Each sales terminal includes a POS module, an environmental sensor, and an electronic price tag. The POS module is specifically a network-enabled cash register or POS tablet, which runs a sales management system. This system records the item identification, quantity, sales time, and transaction amount of each pastry transaction in real time by scanning or inputting the pastry's barcode or QR code, thereby generating sales data. This data is then uploaded to the cloud operation platform in real time. The environmental sensor is preferably a temperature and humidity recorder integrating temperature and humidity sensors. It is installed in the store's refrigerated display cases or open shelves used for displaying and short-term storage of pastries to ensure that the collected environmental parameters accurately reflect the actual microenvironment of the pastries. The sensor continuously collects environmental temperature and humidity sequences at a preset sampling period and uploads the data via the store's sales terminal's gateway device. The electronic price tag is preferably a wireless electronic shelf label using e-ink technology. It connects to the store's sales terminal's base station via Bluetooth to receive and refresh price information sent from the cloud operation platform.

[0061] The cloud-based operation platform is deployed on a cloud server. Its internal data aggregation module, demand forecasting module, production scheduling module, intelligent allocation module, shelf-life dynamic management module, and dynamic pricing module are implemented through a microservice architecture or modular programming. The modules interact with each other through application programming interfaces. The cloud-based operation platform is equipped with a high-performance database to store all historical sales data, real-time inventory data, environmental parameters, master data, and various forecasting and decision-making intermediate results. The master data includes store information, product information, and formula information.

[0062] At the start of a complete operational cycle, the data aggregation module serves as the data entry point, continuously performing its data collection and aggregation functions. Specifically, this data aggregation module simultaneously establishes data connections with the POS cash register modules of each store's sales terminals, environmental sensors, external weather data, and the inventory database in the enterprise resource planning system. The module polls or listens to each data source at preset time intervals, receiving and parsing incremental sales data from the POS cash register modules, environmental parameter sequences from environmental sensors, real-time inventory data from each store and the central factory in the enterprise resource planning system, and external data obtained through the application programming interface (API). This external data includes weather type data, temperature data, holiday identification data for the target city on the current day and for the next few days, as well as promotional activity data pre-entered by the marketing department. All collected raw data, after being stamped with a collection timestamp and data source identifier, undergoes cleaning, format conversion, and standardization processing, and is stored in the corresponding real-time database and historical data warehouse of the cloud-based operational platform.

[0063] The input terminal of the demand forecasting module is connected to the data aggregation module. In a preferred embodiment, the demand forecasting module employs a combined forecasting strategy to improve the accuracy of predicting the daily demand for each pastry item in each store within a preset future time period, such as... Figure 2 As shown, the specific prediction process is as follows:

[0064] The data aggregation module extracts the historical sales sequence of each pastry item in each store within the past preset period. This sequence is organized with the day as the smallest time granularity, recording the daily sales volume of each type of pastry for each day.

[0065] The data aggregation module synchronously acquires external data corresponding to the historical sales sequence dates, including the weather type, average temperature, whether it is a statutory holiday or store anniversary or other special date, and whether a promotional activity was carried out on that day;

[0066] The extracted historical sales series is subjected to STL time series decomposition. STL is a seasonal trend decomposition based on local weighted regression. This decomposition process decomposes the original sales series into three components: trend component, seasonal component, and residual component. The trend component reflects the slow growth or decline trend of pastry sales in the long term, the seasonal component captures the regular fluctuations with a weekly or monthly cycle, and the residual component represents the influence of random noise or irregular events remaining after removing the trend and seasonality.

[0067] The trend component sequence and seasonal component sequence obtained from the decomposition are directly used as the basic prediction features. For external data, the weather type data is uniquely encoded to form a weather feature vector. At the same time, the temperature data, holiday identification data and promotional activity data are normalized and then concatenated with the weather feature vector to form an auxiliary prediction feature vector.

[0068] The basic prediction features and the auxiliary prediction features corresponding to the same prediction date are concatenated end to end to construct a complete prediction feature vector;

[0069] The constructed prediction feature vector is input into two independent prediction models. The first model is a pre-trained gradient boosting decision tree model, which is used to capture the nonlinear relationship and complex interaction between features. Its output is the preliminary demand prediction value of each pastry item in each store for each day within a preset time period. The second model is a pre-trained long short-term memory neural network model. The input of this model is the residual component sequence. It uses its ability to process time series memory to learn and predict the short-term dependence and fluctuation pattern in the residual sequence. Its output is the residual prediction value of each pastry item in each store for each day within a preset time period.

[0070] The preliminary demand forecast and the residual forecast for the same forecast date are superimposed daily. That is, the output of the gradient boosting decision tree model is added to the output of the long short-term memory neural network model, and the sum is used as the final demand for that pastry item in the store on the forecast date. By modeling and combining the basic trend seasonal item and the residual item separately, the accuracy and robustness of the forecast are improved.

[0071] The production scheduling module's input is connected to the demand forecasting module, and its output is connected to the intelligent production terminal of the central factory. This module performs a scheduling calculation once a day or per shift, summarizing the final demand of each store for each pastry item within a preset time period. Combined with the central factory's existing safety stock strategy, it calculates the total production demand that the central factory needs to produce within the corresponding time period. In this embodiment, the preset time period is the next 24 hours. For a pastry item M, its total production demand equals the sum of the predicted demand from all stores, minus the inventory available for allocation in the central factory's current inventory, plus the safety stock increment set to cope with sudden demand. The current raw material inventory data of the central factory is obtained from the cloud operation platform's database. This data includes the current inventory, allocated quantity, and shelf-life status of each raw material, including normal, near-expiry, and expired states. Simultaneously, the recipe data for each pastry item stored in the product recipe management module is read. This data includes the various raw materials required to produce a unit weight or quantity of pastries and their standard usage.

[0072] Based on the formula data, the total production demand is converted into the demand for each raw material. Each raw material is then assessed to determine whether its current inventory, after deducting the allocated amount, meets the demand. If the assessment results indicate that the current inventory of one or more raw materials does not meet the demand after deducting the allocated amount, the shortage ratio of each scarce raw material is calculated. The shortage ratio is the quotient obtained by dividing the shortage amount by the demand. The largest shortage ratio is selected, and the total production demand is proportionally reduced according to this ratio to generate an adjusted total production demand that is feasible under the existing raw material constraints.

[0073] The production scheduling module acquires equipment capacity data from the central factory. This capacity data includes a unique equipment identifier for each production piece of equipment, its type, unit-time capacity, and available time window determined by the factory shift schedule. The equipment types include rotary kilns, blast furnaces, and dough sheeters. The unit-time capacity is the number of kilograms or pieces of pastries that can be produced per hour. In this embodiment, the optimization focuses on minimizing production changeovers and maximizing equipment utilization. With the delivery time of each pastry item as a hard constraint, a genetic algorithm is used for multi-objective optimization. The delivery time is the latest delivery time required by the stores. The specific implementation process of the genetic algorithm is as follows:

[0074] Encoding and Population Initialization: A process-based encoding method is adopted, treating the production task of each pastry item as an independent workpiece, each production equipment as a machine, and each chromosome individual as a scheduling scheme composed of multiple gene segments. The number of gene segments is equal to the total number of production equipment K. Each gene segment corresponds to all pastry item production tasks that need to be executed on a specific production equipment, and these tasks are arranged in a predetermined production order within the gene segment. During initialization, multiple chromosome individuals that meet the delivery time constraints are randomly generated to form the initial population.

[0075] Fitness Function Design: This embodiment designs a fitness function that comprehensively considers production changeover penalties and equipment utilization, and its expression is: In the formula, Indicates the individual's fitness value. This represents the total number of production changeovers across all production equipment. A production changeover is defined as two adjacent production tasks on the same production equipment having different pastry item identifiers. Used to penalize frequent switching; the fewer the number of switches, the better. The larger the value, Indicates all The sum of the actual production time occupied by each production equipment. Indicates all The ratio of the total available time window of each production unit to the total available time window reflects the overall utilization rate of all equipment. and These are the production changeover penalty weight and the equipment utilization weight, both of which are positive and satisfy the following conditions: In this embodiment, the following settings are configured based on the factory's management priorities. .

[0076] Evolutionary operation: using the fitness function value To maximize the optimization direction, selection, crossover, and mutation operations are performed on the current population. The selection operation uses roulette wheel or tournament selection to give individuals with high fitness a greater probability of being selected to reproduce offspring. The crossover operation uses partial matching crossover or sequential crossover to exchange parts of the gene segments of two parent individuals to generate new offspring individuals. The mutation operation randomly selects an individual and swaps two task positions within a certain gene segment of it, or moves one task to another position, in order to maintain the diversity of the population.

[0077] Termination and Output: When the population iterative evolution reaches the preset maximum number of iterations, the iteration terminates. In this embodiment, the maximum number of iterations is 500 generations. The fitness function value is selected from the final population. The largest chromosome is decoded, and the production sequence of pastry items executed by each production device corresponding to that chromosome within each time window is output.

[0078] The optimized production sequence of the pastry items is packaged into a standardized production work order. The data structure of the work order includes at least the work order number, pastry item identifier, planned production quantity, target production equipment identifier, planned start time, and planned end time. The production work order is distributed to the corresponding intelligent production terminal in the central factory through the network to guide actual production activities.

[0079] The input terminal of the shelf-life dynamic management module is connected to the data aggregation module. In this embodiment, the shelf-life dynamic management module is triggered to execute at a fixed time step to dynamically evaluate the remaining saleable time for each batch of pastries in a saleable state. The fixed time step is 5 minutes, and its specific calculation logic is as follows:

[0080] When a batch of pastries is completed and put into storage, the shelf-life dynamic management module obtains the production completion timestamp of that batch of pastries from the information transmitted back from the intelligent production terminal. And compare it with the standard shelf life preset in the master data for this pastry item. Relatedly, in this embodiment, the standard shelf life is the number of days of shelf life under standard temperature and humidity conditions.

[0081] At each subsequent trigger point, the shelf-life dynamic management module retrieves the environmental sensor data from the data aggregation module at the store where the batch of pastries was located, based on the data from the moment production was completed. Up to the current moment Between preset sampling periods The collected complete environmental parameter sequence, including the environmental temperature sequence, is included. and ambient humidity sequence .

[0082] Based on the ambient temperature and humidity sequences, the cumulative quality degradation of this batch of pastries from production completion to the current moment is calculated. This embodiment introduces a quality degradation rate factor. This is used to quantify the impact of real-time environmental deviations from standard storage environments on the rate of pastry quality degradation. This quality degradation rate factor... The calculation formula is: In the formula, It represents The dimensionless quality decay rate factor at time. for The ambient temperature value collected by the environmental sensor at that time. for The ambient humidity value collected by the environmental sensor at that time. The standard temperature value under preset standard storage environment conditions, in this embodiment, is... Set to 25℃. The standard humidity value under preset standard storage environment conditions, in this embodiment, is... Set to 60%, For temperature sensitivity coefficient, These are humidity sensitivity coefficients. These two coefficients are pre-calibrated experimentally and stored in the system based on the sensitivity of different pastry items to temperature and humidity.

[0083] After calculating the instantaneous decay rate factor at each sampling moment, the cumulative quality decay from the completion of production to the current moment is calculated using an integral summation method. ,in, This converts the quality degradation process under actual non-standard environments into an equivalent degradation time under standard environments. For the preset sampling period, the actual remaining marketable time for each batch of pastries is calculated based on the cumulative quality degradation and the standard shelf life. The calculation formula is: ,in, This refers to the actual remaining saleable time. The preset upper limit threshold for quality degradation is reached when the cumulative quality degradation reaches... When this occurs, the batch of pastries is considered completely unsuitable for sale. In this embodiment, Set to be equal to the standard shelf life. , The standard shelf life is the duration stated.

[0084] The input end of the dynamic pricing module is connected to the shelf-life dynamic management module and the data aggregation module, and its output end is connected to the electronic price tags of each store. This module is triggered to execute each time the shelf-life dynamic management module completes an update, or each time a sale occurs. First, the actual remaining saleable time of each batch of pastries in each store is obtained from the shelf-life dynamic management module. The system retrieves the current inventory quantity of the corresponding pastry item for the corresponding store from the data aggregation module. and the store's real-time customer traffic data during the current period. .

[0085] Based on the actual remaining available sales time Compared with the standard shelf life The proportional relationship is used to determine the shelf life stage of each batch of pastries. In this embodiment, the shelf life stage is divided into a normal stage and a near-expiry warning stage, and the judgment criterion is: comparing the actual remaining saleable time. With a preset pre-sale threshold In this embodiment, the following is set ,when When the pastry is determined to be in the normal stage, the module does not adjust the discount, and the electronic price tag retains the original price; when When a pastry item is determined to be in the near-expiry warning stage, the module initiates a dynamic discount rate calculation process, whereby the real-time discount rate... The calculation process is as follows:

[0086] Calculate the urgency index This indicator is used to quantify the promotional pressure on this batch of pastries due to time constraints and inventory backlog. The calculation formula is as follows: ,in, The term "urgency index" is dimensionless; a higher value indicates greater urgency. The actual remaining saleable time, The preset pre-sale threshold, Current inventory quantity Average daily sales volume of pastry items in the corresponding stores The ratio reflects the current inventory pressure. The inventory pressure weighting coefficient is a pre-set positive number used to adjust the contribution of inventory pressure to the urgency level. In this embodiment, Set it to 0.5.

[0087] Obtain the real-time passenger flow data for the current time period. And calculate the passenger flow correction factor. This factor is used to dynamically adjust the discount based on the current customer flow, and its calculation formula is as follows: ,in, For dimensionless passenger flow correction factors, This refers to real-time customer traffic data for the current time period, such as the number of customers entering the store within the current hour. This corresponds to the store's historical average customer traffic during the same time period. When customer traffic is high, that is... When sales opportunities increase, the system appropriately reduces the discount rate; conversely, when customer traffic is low... At that time, it is necessary to attract customers with larger discounts.

[0088] Calculate the final real-time discount rate ,in, This is a preset lower limit for the discount rate, used to ensure that the discount is not too low, resulting in losses or damage to the brand image. The function that calculates the maximum value ensures that the calculated discount rate is not lower than the minimum discount rate set by the system. In this embodiment, Set to 0.5, the dynamic pricing module will generate the real-time discount rate. Along with the corresponding pastry item identification and batch information, the data is pushed to the electronic price tag management system of the corresponding store via the network. The management system formats the original price and the discounted price and sends them to the designated electronic price tag. The electronic price tag receives and updates the display of the original price and the discounted price of the corresponding pastry item.

[0089] The input terminal of the intelligent allocation module is connected to the demand forecasting module and the data aggregation module, and is triggered at preset times each day. In this embodiment, the preset times are 6:00 AM and 2:00 PM. The intelligent allocation module obtains real-time inventory details for each pastry item in each store from the data aggregation module. This real-time inventory includes not only the quantity of inventory but also the actual remaining saleable time for each batch, calculated by the shelf-life dynamic management module. Meanwhile, the intelligent allocation module obtains the final demand for each pastry item in each store within a preset time period from the demand forecasting module. In this embodiment, the preset time period is the next 24 hours.

[0090] For each active pastry item The intelligent allocation module performs a single traversal calculation to identify the inventory status of each store. Specifically, it calculates the available inventory level for each store, which is equal to the percentage of available inventory in all batches of inventory at that store. The sum of inventory levels exceeding a preset minimum transportation and sales cycle threshold (in this embodiment, the minimum transportation and sales cycle threshold is 8 hours) is compared with the final demand.

[0091] When the store Pastry items If the available inventory level exceeds its final demand, and the excess exceeds a preset transfer trigger threshold, then the store will be... Marked as pastry item The candidate stores are selected and their inventory surplus is calculated, which is equal to the available inventory level minus the final demand.

[0092] When the store Pastry items When the available inventory level is less than its final demand, the store will... Marked as pastry item The candidate stores are selected and their inventory gap is calculated, which is equal to the final demand minus the available inventory level.

[0093] After marking all stores, the system performs allocation matching, prioritizing the closest stores in the pending allocation batch, provided that the needs of the receiving stores are met. The longest-running store will be transferred, and the specific matching logic is as follows:

[0094] The shortage quantity from the receiving stores will be allocated to one or more sending stores according to the first-in, first-out (FIFO) principle. When generating the transfer order, ensure that the allocated transfer batch number corresponds to the pastry. Even after deducting estimated transportation and shelf-sales times, sufficient sales windows are still available for the receiving stores. Upon successful matching, a standardized pastry transfer instruction is generated. This instruction's data structure includes the sending store identifier, receiving store identifier, pastry item identifier, transfer batch number, transfer quantity, and transfer time window. The transfer time window includes the suggested outbound time and estimated delivery time. The pastry transfer instruction is pushed to the corresponding sending and receiving stores' sales terminals to prompt staff to perform preparation, outbound, and receiving operations. Simultaneously, cloud-based inventory data is updated, enabling optimized cross-store inventory allocation.

[0095] In this embodiment, as Figure 3 As shown, the overall collaborative workflow of this digital operation system for pastries is as follows:

[0096] The data aggregation module continuously collects sales data, inventory data, environmental parameters, and external data from each store;

[0097] The demand forecasting module outputs the final demand for each pastry item in each store based on the collected data;

[0098] The production scheduling module generates the optimal production work order based on the final demand, combined with the raw material inventory and equipment capacity of the central factory, and sends it to the intelligent production terminal to execute baking production.

[0099] After the pastries are delivered to each store, the shelf life dynamic management module calculates the actual remaining saleable time of each batch of pastries in real time based on the environmental parameter sequence collected by the store's environmental sensors.

[0100] When the actual remaining sales time is lower than the set threshold, the dynamic pricing module calculates the real-time discount rate by combining the current inventory quantity and real-time customer traffic data, and pushes it to the electronic price tag for dynamic price reduction promotion;

[0101] The intelligent allocation module identifies candidate stores for allocation and candidate stores for allocation based on the real-time inventory and final demand of each store, and generates a pastry allocation instruction containing safe batch information to achieve cross-store inventory optimization.

[0102] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A digital operation system suitable for pastries, characterized in that, The system includes: At least one intelligent production terminal, located in the central factory, is used to receive production work orders and execute pastry baking production; Multiple store sales terminals are set up in each store. Each store sales terminal includes a POS cashier module for collecting pastry sales data, an environmental sensor for real-time monitoring of environmental parameters, and an electronic price tag for displaying pastry prices and discount information. The cloud-based operation platform communicates and connects with the intelligent production terminals and various store sales terminals via the network. The cloud-based operation platform includes: a data aggregation module, a demand forecasting module, a production scheduling module, an intelligent allocation module, a shelf-life dynamic management module, and a dynamic pricing module.

2. The digital operation system for pastries according to claim 1, characterized in that, The data aggregation module is used to collect and aggregate the sales data, inventory data, environmental parameters, and external data of each store in real time. The input terminal of the demand forecasting module is connected to the data aggregation module. Based on the sales data and external data collected by the data aggregation module, the demand forecasting module predicts the demand for each pastry item in each store within a preset time period.

3. The digital operation system for pastries according to claim 2, characterized in that, The process by which the demand forecasting module predicts the demand for each type of pastry is as follows: The historical sales sequence of each pastry item in each store within a preset period is obtained from the data aggregation module. The historical sales sequence records the daily sales volume of each pastry item with a daily time granularity. The external data is obtained from the data aggregation module, including weather type data, temperature data, holiday identification data, and promotional activity data. The historical sales series were decomposed using STL time series decomposition to obtain trend component series, seasonal component series and residual component series; The trend component sequence and the seasonal component sequence are used as the basic prediction features, and the weather feature vector after the weather type data is individually thermally encoded, the temperature data, the holiday identification data and the promotional activity data are used as auxiliary prediction features. The prediction feature vector is constructed by splicing them together. The predicted feature vector is input into a pre-trained gradient boosting decision tree model, which outputs the preliminary demand prediction values ​​for each pastry item in each store on each day within a preset time period. The residual component sequence is input into a pre-trained long short-term memory neural network model, which outputs the residual prediction values ​​of each pastry item in each store for each day within a preset time period. The preliminary demand forecast and the residual forecast are superimposed daily to obtain the final demand for each pastry item in each store on each day within a preset time period.

4. The digital operation system for pastries according to claim 1, characterized in that, The input of the production scheduling module is connected to the demand forecasting module, and the output of the production scheduling module is connected to the intelligent production terminal. The production scheduling module is used to generate and issue the production work order based on the demand output by the demand forecasting module, the capacity of the central factory, and the raw material inventory. Specifically, it includes: The final demand for each pastry item at each store within a preset time period is summarized to obtain the total production demand for each pastry item at the central factory within the preset time period. Obtain the current raw material inventory data of the central factory, which includes the current inventory level and availability status of each raw material; Based on the recipe data of each pastry item, the total production demand is calculated back into the demand consumption of each raw material, and it is determined whether the current inventory of each raw material meets the demand consumption. If the current inventory level is insufficient to meet the demand consumption, the total production demand is reduced proportionally according to the shortage ratio of the raw materials to obtain the adjusted total production demand. Obtain equipment capacity data from the central factory, including equipment identification, equipment type, unit time capacity, and available time window for each production equipment. With the dual objectives of minimizing production changeovers and maximizing equipment utilization, and with the delivery time constraints of each pastry item as a hard constraint, a genetic algorithm is used to schedule and optimize the total production demand after adjustment, generating a production sequence for each pastry item executed by each production equipment within each time window. The production sequence of the pastry items is packaged into the production work order. The production work order includes the work order number, pastry item identifier, production quantity, target production equipment identifier, planned start time and planned end time, and the production work order is sent to the intelligent production terminal.

5. A digital operation system for pastries according to claim 4, characterized in that, The process of the genetic algorithm performing scheduling optimization calculations is as follows: The execution order of each pastry production task on each production equipment is encoded as a chromosome individual. Each chromosome consists of multiple gene segments, and each gene segment corresponds to the sequence of pastry production tasks arranged in the execution order on a production equipment. Construct a fitness function, wherein the fitness function is: ,in, For the fitness function value, This represents the total number of production switchovers across all production equipment. This refers to the total number of production equipment. For the first The actual production time occupied by each production equipment For the first Total available time window duration for each production unit. and These are the production changeover penalty weight and the equipment utilization weight, respectively. ; With maximizing the fitness function value as the optimization direction, the population is subjected to selection, crossover, and mutation operations for iterative evolution. The iteration terminates when the preset maximum number of iterations is reached, and the production sequence of the pastry item corresponding to the chromosome with the largest fitness value is output.

6. A digital operation system for pastries according to claim 1, characterized in that, The input terminal of the shelf life dynamic management module is connected to the data aggregation module. The shelf life dynamic management module obtains the production completion timestamp of each batch of pastries and the preset standard shelf life of each pastry item. The data aggregation module acquires environmental parameter sequences collected by the environmental sensors of the corresponding stores at a preset sampling period. These environmental parameter sequences include an ambient temperature sequence and an ambient humidity sequence. Based on the ambient temperature sequence and the ambient humidity sequence, a quality degradation rate factor for each batch of pastries is calculated. ,in, for The quality degradation rate factor at any given time. for The ambient temperature value collected by the environmental sensor at that time. for The ambient humidity value collected by the environmental sensor at that time. The standard temperature value under the aforementioned standard storage environment conditions. The standard humidity value under the aforementioned standard storage environment conditions. For temperature sensitivity coefficient, Humidity sensitivity coefficient; Calculate the cumulative quality degradation of each batch of pastries based on the quality degradation rate factor. ,in, This represents the cumulative quality degradation from the completion of production to the present moment. The production completion timestamp, For the current moment, The preset sampling period; Based on the cumulative quality degradation and the standard shelf life, calculate the actual remaining marketable time for each batch of pastries: ,in, The actual remaining saleable time, This is the preset upper limit threshold for quality degradation. The standard shelf life is the duration stated.

7. A digital operation system for pastries according to claim 1, characterized in that, The input end of the dynamic pricing module is connected to the shelf life dynamic management module and the data aggregation module, and the output end of the dynamic pricing module is connected to the electronic price tag. The dynamic pricing module is used to obtain the actual remaining sales time from the shelf life dynamic management module, and to obtain the current inventory quantity of each pastry item in each store and the real-time customer flow data of each store in the current time period from the data aggregation module. Based on the actual remaining saleable time and the standard shelf life, the shelf life stage of each batch of pastries is determined, including the normal stage and the near-expiration warning stage. When the actual remaining saleable time Greater than or equal to the preset pre-sale threshold At that time, the pastry item is in the normal stage, and the real-time discount rate remains the original price discount rate; When the actual remaining saleable time Less than the preset pre-sale threshold When the pastry item is in the near-expiration warning stage, the real-time discount rate is calculated based on the actual remaining sales time, the current inventory quantity, and the real-time customer flow data. The dynamic pricing module generates a real-time discount rate and pushes it to the electronic price tag of the corresponding store via the network. The electronic price tag receives and updates the original price and discounted price of the corresponding pastry item.

8. A digital operation system for pastries according to claim 7, characterized in that, When a pastry item is in the near-expiry warning stage, a near-expiry urgency index is calculated based on the actual remaining sales time, the current inventory quantity, and the real-time customer traffic data. ,in, The aforementioned urgency index, The actual remaining saleable time, To preset the pre-sale threshold, This represents the current inventory quantity of pastry items at the corresponding store. This refers to the average daily sales volume of pastry items for a predetermined number of days at the corresponding stores. As the inventory pressure weighting coefficient, obtain the real-time passenger flow data for the current period and calculate the passenger flow correction factor: ,in, The passenger flow correction factor is... This is the real-time passenger flow data for the current time period. This represents the historical average customer traffic for the corresponding store during the same period. The real-time discount rate is calculated based on the urgency index and the passenger flow correction factor. ,in, For real-time discount rates, This is the preset lower limit of the discount rate. This is the function for finding the maximum value.

9. A digital operation system for pastries according to claim 1, characterized in that, The input end of the intelligent allocation module is connected to the demand forecasting module and the data aggregation module. The intelligent allocation module obtains the real-time inventory of each pastry item in each store from the data aggregation module and obtains the final demand of each pastry item in each store within a preset time period from the demand forecasting module. For each pastry item, calculate the inventory surplus or inventory deficit for each store: When the real-time inventory of pastry item M in a store is greater than the final demand and the excess exceeds the preset transfer trigger threshold, the store is marked as a candidate store for transfer of pastry item M, and the inventory surplus is calculated. When the real-time inventory of pastry item M in a store is less than the final demand, the store is marked as a candidate store for pastry item M, and the inventory gap is calculated. Based on the inventory surplus and inventory deficit, a pastry transfer instruction is generated. The pastry transfer instruction includes the sending store identifier, receiving store identifier, pastry item identifier, transfer batch number, transfer quantity, and transfer time window. The pastry transfer instruction is then pushed to the corresponding store sales terminals of the sending and receiving stores.

Images