A production scheduling method and system based on big data technology

Abstract

The present invention relates to the field of big data processing technology, and in particular to a production scheduling method and system based on big data technology, which includes an order data acquisition module, a customer classification module, a stable customer demand prediction module, a fluctuating customer demand prediction module, and a pre-production total acquisition module. The present invention can effectively distinguish stable customers from fluctuating customers by comprehensively analyzing a large amount of historical order data, including order frequency, order time difference, and order demand, and calculating the standard deviation of these data; for stable customers, an order demand prediction based on an LSTM model is used to improve the prediction accuracy of the order demand of these customers; for fluctuating customers, an exponential smoothing method adjusted by a particle swarm optimization algorithm is used for prediction, which can better adapt to the uncertainty and volatility of demand; this classification prediction method makes the production scheduling plan more accurate, and can quickly respond to market changes and fluctuations in customer demand.

Description

Technical Field

[0001] The present invention relates to the technical field of big data processing, and in particular to a production scheduling method and system based on big data technology. Background Art

[0002] In today's rapidly changing business environment, the manufacturing industry is facing growing market demand and fierce competitive pressure. In order to maintain competitiveness and meet customer expectations, manufacturing companies must optimize their production plans and scheduling strategies. Traditional scheduling methods rely on historical data and static forecasts, which often fail to accurately reflect the dynamic changes in the market and the volatility of customer demand, resulting in low production efficiency and inventory backlogs.

[0003] The existing production scheduling system still relies more on manual statistics and schedules production based on the number of orders. However, the data collected in the early stage, such as the number of orders, order demand, and order time, are rarely fully utilized to their full potential, especially in terms of demand forecasting to distinguish between stable customers and fluctuating customers.

[0004] Research has found that stable customers usually have predictable order patterns and demand volumes, while the demands of fluctuating, unstable customers are more uncertain and difficult to predict. For demand forecasting of fluctuating customers, although the traditional exponential smoothing method is simple and easy to use, it often lacks the ability to capture complex patterns. Therefore, if we can accurately distinguish between these two types of customers and formulate corresponding production scheduling strategies for each of them, we can greatly improve the flexibility and responsiveness of our company's production plan. Summary of the invention

[0005] Based on this, it is necessary to provide a production scheduling method and system based on big data technology to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention discloses a production scheduling method based on big data technology, comprising:

[0007] Obtain the order data within the current preset period, which includes the number of orders and order demand of all ordering customers; combine the order data obtained in the current preset period and the order data obtained in the most recent N-1 preset periods into an order data time series set in chronological order;

[0008] For any customer who has placed an order in the current preset period and the most recent N-1 preset periods, calculate the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand in the order data time series set; set threshold conditions. If the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand meet the threshold conditions, the customer is classified as a stable customer.

[0009] Filter out the order demands corresponding to all stable customers in the order data time series set, and obtain a stable customer data prediction set, which includes N stable customer order demands sorted by time; use the stable customer data prediction set as the input of the order demand prediction model, and output the stable customer prediction demand for the next preset period;

[0010] Filter out the order demand corresponding to the remaining customers except the stable customers in the order data time series set to obtain the fluctuating order data set, which contains the order demand of N remaining customers sorted by time; predict the fluctuating customer forecast demand in the next preset period based on the fluctuating order data set;

[0011] The predicted demand of stable customers and the predicted demand of fluctuating customers are combined into the overall predicted demand, and the current total inventory is obtained at the same time; the difference between the overall predicted demand and the current total inventory is calculated to obtain the initial pre-production quantity; a buffer value is set based on the overall predicted demand, the sum of the initial pre-production quantity and the buffer value is calculated, and the total pre-production quantity for the next preset period is obtained.

[0012] Preferably, the specific operations for stable customers are divided into:

[0013] For any customer, use the formula Calculate the order frequency P of the current customer, where n represents the number of orders placed by the current customer in N preset periods; arrange the n orders of the current customer in chronological order, calculate the time difference of all adjacent orders in turn, and obtain the time difference data set t i , i=1, 2, ..., n-1; calculate the average value of the time difference data set , using the formula Calculate the standard deviation T of the time difference of the current customer's order placement; combine the order demand of the current customer's n orders into an order demand data set a j , j=1, 2, ..., n; calculate the average value of the order demand data set , using the formula Calculate and obtain the standard deviation A of the current customer's order demand; set the frequency threshold, time difference threshold and demand threshold. If the current customer's order frequency P is greater than the frequency threshold, and the standard deviation T of the order time difference and the standard deviation A of the order demand are less than the time difference threshold and the demand threshold respectively, the current customer is classified as a stable customer.

[0014] Preferably, the order demand forecasting model is established based on the LSTM model, including N demand forecasting LSTM units, 1 fully connected layer and 1 output layer, wherein the demand forecasting LSTM unit is used to extract the temporal features in the stable customer data forecast set; the fully connected layer is used to extract features from the outputs of all demand forecasting LSTM units; and the output layer is used to output the predicted demand of stable customers.

[0015] Preferably, for the training of the demand forecasting model, the specific operations are:

[0016] Obtain several order demand training samples, each of which stores N+1 order demand data sorted by time, and use the first N order demand data in the order demand training samples as the input of the order demand prediction model, and use the last order demand data in the order demand training samples as the output of the order demand prediction model; divide all order demand training samples into an order demand training set and an order demand test set, send the order demand training set to the order demand prediction model with initialized parameters for training, and then input the order demand test set to the order demand prediction model to obtain the test results; set the training conditions to determine whether the test results meet the training conditions, and if so, output the trained order demand prediction model; otherwise, continue to train the order demand prediction model through the order demand training set.

[0017] Preferably, the specific operation of predicting the predicted demand of fluctuating customers in the next preset period according to the fluctuating order data set is:

[0018] For the currently acquired volatility order data set b k , k=1,2,…,N; using the formula Calculate and obtain the volatility order data set b k The order demand forecast value D in the k+1th time period k+1 , where α is the smoothing coefficient and is adjusted by the swarm optimization algorithm; D k represents the order demand forecast value for the kth time period, and when k=1, D1=b1; when k=N, the order demand forecast value D is calculated. N+1 That is, the predicted demand of fluctuating customers for the next preset period.

[0019] Preferably, the swarm optimization algorithm for adjusting the smoothing coefficient α is a particle swarm optimization algorithm, and the value of the smoothing coefficient α is updated once every preset period.

[0020] Preferably, the specific operation of adjusting the smoothing coefficient α using the particle swarm optimization algorithm is:

[0021] A1: Generate a particle swarm randomly in the range of 0-1, and the position of each particle represents a candidate smoothing coefficient;

[0022] A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where E represents the degree of deviation between the order demand forecast value obtained by applying the current candidate smoothing coefficient and the actual order demand;

[0023] A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position;

[0024] A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position;

[0025] A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle.

[0026] A6: Repeat steps A4 to A5 until the maximum number of iterations is reached. The global optimal position obtained is the optimal smoothing coefficient.

[0027] Preferably, the production scheduling system based on big data technology includes:

[0028] An order data acquisition module is used to compose an order data time series set by chronologically arranging the order data acquired in the current preset period and the order data acquired in the most recent N-1 preset periods;

[0029] The customer classification module is used to calculate the order frequency, standard deviation of order time difference and standard deviation of order demand of all customers in the order data time series set respectively; if the order frequency, standard deviation of order time difference and standard deviation of order demand of the customer meet the threshold conditions, the customer is classified as a stable customer;

[0030] The stable customer demand forecasting module is used to use the stable customer data forecast set as the input of the order demand forecasting model and output the stable customer forecast demand for the next preset period;

[0031] The fluctuating customer demand prediction module is used to filter out the order demand corresponding to the remaining customers except the stable customers in the order data time series set, and obtain the fluctuating order data set; based on the fluctuating order data set, the predicted demand of the fluctuating customers in the next preset period is predicted;

[0032] The module for obtaining the total pre-production volume is used to combine the predicted demand of stable customers and the predicted demand of fluctuating customers into the overall predicted demand, and calculate the difference between the overall predicted demand and the current total inventory to obtain the initial pre-production volume; calculate the sum of the initial pre-production volume and the buffer value to obtain the total pre-production volume for the next preset period.

[0033] The present invention has the following advantages:

[0034] 1. The present invention can effectively distinguish stable customers from fluctuating customers by comprehensively analyzing a large amount of historical order data, including order frequency, order time difference and order demand, and calculating the standard deviation of these data; the order demand forecasting based on the LSTM model is used for stable customers, thereby improving the forecasting accuracy of the order demand of these customers; for fluctuating customers, the exponential smoothing method adjusted by the particle swarm optimization algorithm is used for forecasting, which can better adapt to the uncertainty and volatility of demand; this classification forecasting method makes the production scheduling more accurate, and can quickly respond to market changes and fluctuations in customer demand.

[0035] 2. The present invention determines the initial pre-production quantity by calculating the difference between the predicted demand and the current inventory, and introduces a buffer value to calculate the total pre-production quantity for the next preset cycle; this method can help companies manage inventory more effectively, reduce the risk of inventory backlogs, and ensure that production plans can meet actual demand and avoid overproduction or shortages. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of the structure of a production scheduling system based on big data technology adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0038] Embodiment 1, a production scheduling method based on big data technology, comprises the following steps:

[0039] Obtain order data within the current preset period. The preset period can be one week, one month or longer, depending on the production and operation needs of the enterprise. The order data includes the number of orders and order demand of all ordering customers. The obtained order data can reflect the recent market and customer needs; the order data obtained in the current preset period and the order data obtained in the most recent N-1 preset periods are arranged in chronological order to form an order data time series set. By analyzing the data of multiple periods including the current period, we can better understand the long-term trend of customer demand and provide more comprehensive data support for prediction.

[0040] For any customer who has placed an order in the current preset cycle and the most recent N-1 preset cycles, calculate the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand in the order data time series set; no matter how the customer's order frequency is, as long as he has placed an order in these cycles, he becomes the object of analysis, ensuring that the analysis covers all active customers and provides a comprehensive data basis for further demand forecasting; order frequency is an indicator to measure the activeness of customer purchasing behavior, reflecting the frequency of transactions between customers and enterprises; the standard deviation of order time difference is used to evaluate the consistency and predictability of customer order time, and a smaller standard deviation means that the customer's order interval is relatively stable; the standard deviation of order demand is used to measure the stability of customer order size, and a smaller standard deviation indicates that customer demand is relatively consistent and easy to predict; set threshold conditions, if the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand meet the threshold conditions, it is considered that his order behavior is stable and predictable, and the customer is thus classified as a stable customer.

[0041] The order demands corresponding to all stable customers are screened out in the order data time series set to obtain the stable customer data prediction set, which contains N stable customer order demands sorted by time. By screening the order demands of stable customers, we can focus on customer groups with highly predictable demand patterns, providing an accurate data basis for subsequent forecasting analysis. The stable customer data prediction set is used as the input of the order demand forecasting model to output the stable customer forecast demand for the next preset period, thereby helping enterprises plan production resources in advance, such as material procurement, personnel arrangement and machine scheduling, to meet the predicted order demands.

[0042] In the order data time series set, the order demands corresponding to the remaining customers except the stable customers are screened out to obtain the fluctuating order data set, which contains the order demands of N remaining customers sorted by time; focus on customer groups whose demand patterns may be unstable or difficult to predict, and establish a special data set for the order demands of these customers in order to perform special forecasting analysis; based on the fluctuating order data set, the predicted demand of fluctuating customers in the next preset period is predicted, thereby providing enterprises with forecasts of fluctuating customer demand, helping enterprises to better respond to market changes and adjust production plans and resource allocations in a timely manner to adapt to demands with higher uncertainty.

[0043] The forecast demand of stable customers and the forecast demand of fluctuating customers are combined into the overall forecast demand. At the same time, the current total inventory is obtained, including the quantity of all finished products and work-in-progress, to ensure that the production plan takes into account the existing inventory level and avoid overproduction or inventory backlogs; the difference between the overall forecast demand and the current total inventory is calculated to obtain the initial pre-production quantity, and the minimum number of products that the company needs to produce to meet the forecast demand is determined to provide preliminary guidance for the production plan; a buffer value is set based on the overall forecast demand to provide additional production volume to cope with demand uncertainty and potential forecast errors, reduce out-of-stock risks, calculate the sum of the initial pre-production quantity and the buffer value, and obtain the total pre-production volume for the next preset cycle to ensure that the production plan not only meets the forecast demand but also takes into account uncertainty and potential additional demand, thereby improving the flexibility and adaptability of the production plan.

[0044] Specific operations to identify stable customers:

[0045] For any customer, use the formula Calculate the order frequency P of the current customer, where n represents the number of orders placed by the current customer in N preset periods; arrange the n orders of the current customer in chronological order, calculate the time difference of all adjacent orders in turn, and obtain the time difference data set t i , i=1, 2, ..., n-1; calculate the average value of the time difference data set , using the formula Calculate the standard deviation T of the time difference of the current customer's order placement; combine the order demand of the current customer's n orders into an order demand data set a j , j=1, 2, ..., n; calculate the average value of the order demand data set , using the formula Calculate and obtain the standard deviation A of the current customer's order demand; set the frequency threshold, time difference threshold and demand threshold. If the current customer's order frequency P is greater than the frequency threshold, and the standard deviation T of the order time difference and the standard deviation A of the order demand are less than the time difference threshold and the demand threshold respectively, the current customer is classified as a stable customer.

[0046] The order demand prediction model is built based on the LSTM model. The long short-term memory network is a special type of recurrent neural network that can learn long-term dependencies. LSTM solves the gradient vanishing or gradient exploding problems that occur in traditional RNNs when processing long sequence data by introducing three gating mechanisms: input gate, forget gate, and output gate. The input gate controls the inflow of new information, the forget gate determines which information should be forgotten from the cell state, and the output gate determines which information of the cell state should be included in the next hidden state. This makes LSTM very suitable for processing and predicting sequence information in time series data, and is widely used in speech recognition, natural language processing, stock market prediction and other fields.

[0047] The order demand forecasting model includes N demand forecasting LSTM units, 1 fully connected layer and 1 output layer. The demand forecasting LSTM unit is used to extract temporal features such as trends and seasonality from the stable customer data forecast set. By learning the long-term dependencies in the data, the demand forecasting LSTM unit can capture the pattern of stable customer order demand; the fully connected layer is used to comprehensively process the outputs of all demand forecasting LSTM units, enhance the expression ability of the model through nonlinear transformation, and provide richer information for the final forecasting results; the output layer is used to output the predicted demand of stable customers.

[0048] For the training of demand forecasting model, the specific operations are as follows:

[0049] Obtain several order demand training samples, each of which stores N+1 order demand data sorted by time, and use the first N order demand data in the order demand training samples as the input of the order demand prediction model, and use the last order demand data in the order demand training samples as the output of the order demand prediction model; divide all order demand training samples into an order demand training set and an order demand test set, and the test set is used to verify the prediction ability of the model to ensure that the model not only performs well on the training data, but also generalizes to unseen data; send the order demand training set to the parameter-initialized order demand prediction model for training, and then input the order demand test set to the order demand prediction model to obtain the test results; set the training conditions to determine whether the test results meet the training conditions. If the training conditions are met, output the trained order demand prediction model; otherwise, continue to train the order demand prediction model through the order demand training set.

[0050] According to the fluctuating order data set, the predicted demand of the fluctuating customers in the next preset period is predicted. The method used is exponential smoothing. Exponential smoothing is a time series prediction method. It predicts future values ​​by weighted averaging historical data. The weight decays exponentially over time. Recent data obtains a greater weight, while the weight of earlier data gradually decreases. In the present invention, the advantages of exponential smoothing are its simplicity, adaptability and flexibility. It can quickly respond to changes in demand and optimize the prediction accuracy through dynamic adjustment of the smoothing coefficient α. In particular, with the assistance of the particle swarm optimization algorithm, the smoothing coefficient α can be adaptively updated to further improve the accuracy and response speed of the prediction of fluctuating customer demand. When processing a data set with volatility, this method can provide more accurate and real-time prediction results than the traditional static prediction model. The specific operations are as follows:

[0051] For the currently acquired volatility order data set b k , k=1,2,…,N; using the formula Calculate and obtain the volatility order data set b k The order demand forecast value D in the k+1th time period k+1 , where α is the smoothing coefficient and is adjusted by the swarm optimization algorithm; D k represents the order demand forecast value for the kth time period, and when k=1, D1=b1; when k=N, the order demand forecast value D is calculated. N+1 That is, the predicted demand of fluctuating customers for the next preset period.

[0052] The swarm optimization algorithm for adjusting the smoothing coefficient α is a particle swarm optimization algorithm, and the value of the smoothing coefficient α is updated once every preset period.

[0053] Specific operations of using particle swarm optimization algorithm to adjust the smoothing coefficient α:

[0054] A1: Generate a particle swarm randomly in the range of 0-1, and the position of each particle represents a candidate smoothing coefficient;

[0055] A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where E represents the degree of deviation between the order demand forecast value obtained by applying the current candidate smoothing coefficient and the actual order demand;

[0056] A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position;

[0057] A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position;

[0058] A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle.

[0059] A6: Repeat steps A4 to A5 until the maximum number of iterations is reached. The global optimal position obtained is the optimal smoothing coefficient.

[0060] Example 2, a production scheduling system based on big data technology, such as Figure 1 As shown, including:

[0061] An order data acquisition module is used to compose an order data time series set by chronologically arranging the order data acquired in the current preset period and the order data acquired in the most recent N-1 preset periods;

[0062] The customer classification module is used to calculate the order frequency, standard deviation of order time difference and standard deviation of order demand of all customers in the order data time series set respectively; if the order frequency, standard deviation of order time difference and standard deviation of order demand of the customer meet the threshold conditions, the customer is classified as a stable customer;

[0063] The stable customer demand forecasting module is used to use the stable customer data forecast set as the input of the order demand forecasting model and output the stable customer forecast demand for the next preset period;

[0064] The fluctuating customer demand prediction module is used to filter out the order demand corresponding to the remaining customers except the stable customers in the order data time series set, and obtain the fluctuating order data set; based on the fluctuating order data set, the predicted demand of the fluctuating customers in the next preset period is predicted;

[0065] The module for obtaining the total pre-production volume is used to combine the predicted demand of stable customers and the predicted demand of fluctuating customers into the overall predicted demand, and calculate the difference between the overall predicted demand and the current total inventory to obtain the initial pre-production volume; calculate the sum of the initial pre-production volume and the buffer value to obtain the total pre-production volume for the next preset period.

[0066] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A production scheduling method based on big data technology, characterized in that: The steps include: Obtain the order data within the current preset period, which includes the number of orders and order demand of all ordering customers; combine the order data obtained in the current preset period and the order data obtained in the most recent N-1 preset periods into an order data time series set in chronological order; For any customer who has placed an order in the current preset period and the most recent N-1 preset periods, calculate the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand in the order data time series set; set threshold conditions, and if the customer's order frequency, standard deviation of order time difference, and standard deviation of order demand meet the threshold conditions, the customer is classified as a stable customer; the specific steps for classifying stable customers in this step are as follows: For any customer, use the formula Calculate the order frequency P of the current customer, where n represents the number of orders placed by the current customer in N preset periods; arrange the n orders of the current customer in chronological order, calculate the time difference of all adjacent orders in turn, and obtain the time difference data set t i , i=1, 2, ..., n-1; calculate the average value of the time difference data set , using the formula Calculate the standard deviation T of the current customer's order time difference; combine the order demand of the current customer's n orders into an order demand data set a j , j = 1, 2, …, n; Calculate the average value of the order demand data set , using the formula Calculate and obtain the standard deviation A of the current customer's order demand; set the frequency threshold, time difference threshold and demand threshold. If the current customer's order frequency P is greater than the frequency threshold, and the standard deviation T of the order time difference and the standard deviation A of the order demand are less than the time difference threshold and the demand threshold respectively, then the current customer is classified as a stable customer; Filter out the order demands corresponding to all stable customers in the order data time series set, and obtain a stable customer data prediction set, which includes N stable customer order demands sorted by time; The stable customer data prediction set is used as the input of the order demand prediction model, and the stable customer prediction demand for the next preset period is output; in this step, the order demand prediction model is established based on the LSTM model, including N demand prediction LSTM units, 1 fully connected layer and 1 output layer, wherein the demand prediction LSTM unit is used to extract the temporal features in the stable customer data prediction set; the fully connected layer is used to extract features from the outputs of all demand prediction LSTM units; the output layer is used to output the stable customer prediction demand; in this step, the specific operation steps for the training of the demand prediction model are as follows: obtain several order demand training samples, each of which stores N+1 order demand data sorted by time, and use the first N order demand data in the order demand training sample as the input of the order demand prediction model, and use the last order demand data in the order demand training sample as the output of the order demand prediction model; divide all order demand training samples into an order demand training set and an order demand test set, send the order demand training set to the parameter-initialized order demand prediction model for training, and then input the order demand test set into the order demand prediction model to obtain the test results; Set training conditions and determine whether the test results meet the training conditions. If so, output the trained order demand forecasting model; otherwise, continue to train the order demand forecasting model using the order demand training set. Filter out the order demand corresponding to the remaining customers except the stable customers in the order data time series set to obtain the fluctuating order data set. The fluctuating order data set contains the order demand of N remaining customers sorted by time. According to the fluctuation order data set, the predicted demand of the fluctuation customers in the next preset period is predicted; in this step, the specific operation steps for predicting the predicted demand of the fluctuation customers in the next preset period according to the fluctuation order data set are as follows: for the currently obtained fluctuation order data set b k , k=1,2,…,N; using the formula Calculate and obtain the volatility order data set b k The order demand forecast value D in the k+1th time period k+1 , where α is the smoothing coefficient and is adjusted by the swarm optimization algorithm; D k represents the order demand forecast value for the kth time period, and when k=1, D1=b1; when k=N, the order demand forecast value D is calculated. N+1 That is, the forecast demand of fluctuating customers for the next preset period; Combine the forecast demand of stable customers and the forecast demand of fluctuating customers into the overall forecast demand, and obtain the current total inventory at the same time; Calculate the difference between the overall forecast demand and the current total inventory to obtain the initial pre-production quantity; set a buffer value based on the overall forecast demand, calculate the sum of the initial pre-production quantity and the buffer value, and obtain the total pre-production quantity for the next preset period.

2. The production scheduling method based on big data technology according to claim 1 is characterized in that: The swarm optimization algorithm for adjusting the smoothing coefficient α is a particle swarm optimization algorithm, and the value of the smoothing coefficient α is updated once every preset period.

3. The production scheduling method based on big data technology according to claim 2 is characterized in that: Specific operations of using particle swarm optimization algorithm to adjust the smoothing coefficient α: A1: Generate a particle swarm randomly in the range of 0-1, and the position of each particle represents a candidate smoothing coefficient; A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where E represents the degree of deviation between the order demand forecast value obtained by applying the current candidate smoothing coefficient and the actual order demand; A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position; A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position; A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle. A6: Repeat steps A4 to A5 until the maximum number of iterations is reached. The global optimal position obtained is the optimal smoothing coefficient.

4. A production scheduling system based on big data technology, characterized in that: The system is applied to a production scheduling method based on big data technology as claimed in any one of claims 1 to 3, comprising: An order data acquisition module is used to compose an order data time series set by chronologically arranging the order data acquired in the current preset period and the order data acquired in the most recent N-1 preset periods; The customer classification module is used to calculate the order frequency, standard deviation of order time difference and standard deviation of order demand of all customers in the order data time series set respectively; if the order frequency, standard deviation of order time difference and standard deviation of order demand of the customer meet the threshold conditions, the customer is classified as a stable customer; The stable customer demand forecasting module is used to use the stable customer data forecast set as the input of the order demand forecasting model and output the stable customer forecast demand for the next preset period; The fluctuating customer demand prediction module is used to filter out the order demand corresponding to the remaining customers except the stable customers in the order data time series set, and obtain the fluctuating order data set; based on the fluctuating order data set, the predicted demand of the fluctuating customers in the next preset period is predicted; The module for obtaining the total pre-production volume is used to combine the predicted demand of stable customers and the predicted demand of fluctuating customers into the overall predicted demand, and calculate the difference between the overall predicted demand and the current total inventory to obtain the initial pre-production volume; calculate the sum of the initial pre-production volume and the buffer value to obtain the total pre-production volume for the next preset period.

Images