CN-121998367-A - Intelligent month production plan adjustment optimizing method and system oriented to market demand influence

CN121998367ACN 121998367 ACN121998367 ACN 121998367ACN-121998367-A

Abstract

The invention provides an intelligent optimization method and an intelligent optimization system for month production plan adjustment facing market demand influence, and relates to the technical field of intelligent manufacturing and production plan optimization. The method specifically comprises the steps of obtaining commodity futures data of a target commodity, constructing a two-dimensional observation vector sequence, constructing an HMM (hidden Markov model) demand prediction model, training, inputting the two-dimensional observation vector sequence into the trained HMM demand prediction model, carrying out state decoding on output of the model to obtain a demand state sequence and a state transition matrix, carrying out monthly production planning by utilizing a multi-scene plan generation model to obtain an effective alternative plan scheme set of the target commodity, and screening the effective alternative plan scheme set by utilizing a production utility evaluation model to obtain an optimal plan scheme. According to the invention, three links of market demand prediction, multi-scene plan generation and intelligent evaluation optimization are organically connected in series to form a complete and data-driven decision closed loop.

Inventors

WANG JINGBO
HUANG MIN
WANG QING
SONG YANG
WANG XINGWEI
WANG XIANQIANG

Assignees

东北大学
长春黄金设计院有限公司

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. The month production plan adjustment intelligent optimization method facing the influence of market demands is characterized by comprising the following steps of: Obtaining market observation data, enterprise historical operation data and production resource and constraint data of a target bulk commodity; constructing a two-dimensional observation vector sequence of the target bulk commodity based on the market observation data; building and training an HMM demand prediction model to obtain a trained HMM demand prediction model; inputting the two-dimensional observation vector sequence into a trained HMM demand prediction model, and performing state decoding on the output of the model to obtain a demand state sequence and a state transition matrix of a target bulk commodity ; Based on the enterprise historical operation data, production resources and constraint data, a demand state sequence and a state transition matrix Performing monthly production planning by using a multi-scene plan generation model to obtain an effective alternative plan scheme set of the target bulk commodity; And screening the effective alternative plan scheme set by using a production utility evaluation model based on the production resources and the constraint data to obtain an optimal plan scheme.
2. The method for intelligently optimizing monthly production plan adjustment for market demand influence according to claim 1, wherein the market observation data comprises a futures price history sequence and a transaction amount history sequence, the enterprise history operation data comprises commodity history demand, and the production resources and constraint data comprise equipment maximum capacity and process limiting conditions.
3. The intelligent optimization method for month production plan adjustment facing market demand influence according to claim 2, wherein the specific content of constructing the two-dimensional observation vector sequence of the target bulk commodity based on the market observation data is as follows: Calculating the price log rate of return of each period of the target commodity by using the obtained futures price historical sequence; And taking the price log-rate of return and the transaction quantity log-rate of return in the same period as a group of observation vectors, and further obtaining a two-dimensional observation vector sequence of the commodity.
4. The method for intelligently optimizing monthly production plan adjustment facing market demand influence according to claim 2, wherein the specific contents of the HMM demand prediction model which is built and trained to obtain the trained HMM demand prediction model are as follows: Defining HMM demand prediction model as five-tuple Wherein The hidden state set comprises three market demand states of high demand, medium demand and low demand; For a sequence of two-dimensional observation vectors, , Is the first Observation vector of period; as the initial state probability distribution vector, , Is in state of Is provided for the state probability distribution of (c), ; As a matrix of state transition probabilities, , Is in a slave state Transition to State Is a function of the probability of (1), ; In order to observe the probability density function, , Is in a state of Under-generated observation vector Probability density of (c); Acquisition length of Commodity observation sequence of (2) And is constructed to have a length of Is a sequence of observations of (a) Wherein Represent the first Price data for the period; Represent the first Data of the amount of transactions in the period; And Stage 1 and 1 Observation vector of period; from the observation sequence Determining the observation sequence by adopting an adaptive window mechanism based on the fluctuation rate quantiles Is a sequence of initial hidden states of (a); based on the observation sequence And observing in the sequence Corresponding initial hidden state sequence, initializing initial state probability distribution vector in HMM demand prediction model State transition probability matrix And observing a probability density function ; The method comprises the steps of performing iterative training on an HMM demand prediction model by adopting a Baum-Welch algorithm, and executing the following operations in each iteration process: based on the observation sequence The HMM demand prediction model under the current iteration round calculates forward probability and backward probability respectively; calculating a state probability by using the forward probability and the backward probability Probability of transition ; According to the state probability Probability of transition Updating state transition probability matrices And observing a probability density function ; Judging whether the log-likelihood function is converged under the current iteration round, ending the iterative training to obtain a trained HMM demand prediction model if the log-likelihood function is converged, and continuing the next iteration process until the log-likelihood function is converged if the log-likelihood function is not converged.
5. The intelligent optimization method for month production plan adjustment facing market demand influence according to claim 4, wherein the method is based on observation sequence Determining the observation sequence by adopting an adaptive window mechanism based on the fluctuation rate quantiles The specific contents of the initial hidden state sequence are: Calculating an observation sequence Rolling annual fluctuation rate sequences of (2); acquiring a median and a third quartile of the historical fluctuation rate; For the first Rolling annual rate of fluctuation in period If rolling annual fluctuation rate If the rolling annual fluctuation rate is greater than the third quartile, selecting the first window length Selecting a second window length if the median is greater than the median of the historical volatility and not greater than the third quartile Selecting a third window length if the median of the historical volatility is not greater than the median of the historical volatility; According to the first Rolling annual rate of fluctuation in period Calculate the first using the selected window length Rolling average rate of return for periods ; Will be the first Rolling average rate of return for periods Comparing with a preset demand state dividing threshold value to determine the first Hidden state of period Is any one of high demand, medium demand and low demand, thereby obtaining the observation sequence A corresponding sequence of hidden states.
6. The intelligent optimization method for month production plan adjustment facing market demand influence according to claim 5, wherein the multi-scenario plan generation model is: setting the planning period of the month production plan as Month of the planning period In the period of time, the time period, Based on the enterprise historical operation data, the demand state sequence and the state transition matrix Respectively calculate future th Month reference predicted values for periods under different demand conditions; defining three risk preference types, and respectively generating disturbance of each type of risk preference by using Gaussian noise; For the status of being demanded by And risk preference Arbitrary scene scheme combination of constitution For future th During the demand state Lower month benchmark predictions Repeatedly apply Minor disturbance Obtaining Future (future) Period scene scheme combination Lower demand forecast value, wherein ; Future will be achieved by using a predetermined mapping function Month-to-month scene plan combination Mapping the demand predicted value to a production plan to obtain the future Month-to-month scene plan combination The following alternative planning schemes; defining a filter function based on the production resources and constraint data, and utilizing the filter function for all future Month-to-month scene plan combination Constraint filtering is carried out on the following alternative plan scheme to obtain the future Month-to-month scene plan combination The following set of active alternative planning schemes.
7. The method for intelligently optimizing monthly production plan adjustment for market demand impact according to claim 6, wherein said method is based on said enterprise historical operating data, a demand state sequence and a state transition matrix Respectively calculate future th The method for predicting the month benchmark predicted value under different demand states comprises the following steps: Dividing the enterprise historical operating data into a set of high market demand state days based on the demand state sequence Status day collection for medium market demand Low market demand status day collection Three kinds of demand state subsets, and respectively calculating the average value of the historical daily demands of each kind of demand state subsets ; Defining a state transition probability matrix A kind of electronic device Step-by-step exponentiation ; For the first Status of demand of period By using the transfer power Calculate future th State probability distribution of periods; Average value of historical daily requirement according to the requirement state subsets of each type And future (future) And calculating the month reference predicted value under each type of demand state according to the state probability distribution of the period.
8. The method for intelligently optimizing monthly production plan adjustment for market demand influence according to claim 7, wherein the specific content of the effective alternative plan scheme set is obtained by screening the effective alternative plan scheme set by using a production utility evaluation model based on the production resources and constraint data, and the specific content is as follows: Based on the production resources and constraint data, for the future Month-to-month scene plan combination The following active alternative plan schema set Respectively calculating effective alternative plan scheme sets by using DEA efficiency evaluation system based on BCC model A plan benefit value for each of the available alternative plan scenarios, wherein Representing scene plan combinations The number of available alternative plans; Representing scene plan combinations Lower first A number of available alternative planning schemes; normalizing all calculated planned benefit values to obtain normalized planned benefit values; Constructing scene plan combinations Under AHP hierarchical structure, and calculate scene scheme combination by normalized plan benefit value The comprehensive score of each effective alternative plan scheme is lower; from the active alternative plan schema set In the method, the effective alternative plan scheme with the largest comprehensive score is selected as the future Month-to-month scene plan combination The following optimal planning scheme.
9. The intelligent optimization method for month production plan adjustment facing market demand influence according to claim 8, wherein the scene scheme construction combination Under AHP hierarchical structure, and calculate scene scheme combination by normalized plan benefit value The specific content of the comprehensive score of each effective alternative plan scheme is as follows: Constructing scene plan combinations The lower AHP hierarchical structure comprises a target layer, a criterion layer and a scheme layer; wherein the target layer is used for selecting scene scheme combination The optimal production plan is set; the criterion layer comprises three-dimensional indexes, namely DEA comprehensive efficiency Resource utilization Matching with demand ; The scheme layer is scene scheme combination A set of active alternative planning schemes; for scene scheme combination Lower (th) Individual effective alternative planning schemes Will effectively alternate planning schemes Normalized plan benefit value As a means of DEA comprehensive efficiency index of (C) ; Calculating an effective alternative plan scheme based on the production resources and constraint data Resource utilization index of (2) ; According to an effective alternative plan scheme Calculating an effective alternative plan solution based on demand forecast values of (1) Is a requirement matching degree index of (2) ; Construction scene scheme combination The lower criterion layer judgment matrix is used for calculating the global weight vector of the criterion layer, and meanwhile, consistency test is carried out on the criterion layer judgment matrix, if the test fails, the criterion layer judgment matrix is modified until the test is successful; according to an effective alternative plan scheme DEA comprehensive efficiency index of (C) Index of resource utilization Index of matching degree with requirement Calculating an effective alternative plan scheme by using a weighted summation mode by using a global weight vector corresponding to a checked criterion layer judgment matrix Is a composite score of (2).
10. A market demand impact oriented month production plan adjustment intelligent optimization system for implementing the market demand impact oriented month production plan adjustment intelligent optimization method of any one of claims 1-9, characterized in that the system comprises: the data acquisition module is used for acquiring market observation data, enterprise historical operation data and production resource and constraint data of the target bulk commodity; The observation sequence construction module is used for constructing a two-dimensional observation vector sequence of the target bulk commodity based on the market observation data; The HMM model training module is used for constructing an HMM demand prediction model and training to obtain a trained HMM demand prediction model; The demand state decoding module is used for inputting the two-dimensional observation vector sequence into a trained HMM demand prediction model, and carrying out state decoding on the output of the model to obtain a demand state sequence and a state transition matrix of a target commodity; The multi-scene plan generation module is used for carrying out monthly production planning by utilizing a multi-scene plan generation model based on the enterprise historical operation data, the production resource and constraint data, the demand state sequence and the state transition matrix to obtain an effective alternative plan scheme set of the target bulk commodity; And the production plan decision module is used for screening the effective alternative plan scheme set by utilizing a production utility evaluation model based on the production resources and the constraint data to obtain an optimal plan scheme.

Description

Intelligent month production plan adjustment optimizing method and system oriented to market demand influence Technical Field The invention relates to the technical field of intelligent manufacturing and production plan optimization, in particular to a month production plan adjustment intelligent optimization method and system facing the influence of market demands. Background In the field of mass commodity production management, enterprises generally formulate a monthly production plan according to historical data and static models. Currently, existing methods for optimizing month production schedule adjustment rely mainly on periodic (e.g., monthly or quarterly) schedule reformulation. When market demand changes, common practices include: and (3) manually adjusting parameters such as yield, variety proportion and the like in the original month plan by virtue of experience according to recent sales data and market feedback by a planner. Part of the system adopts a rolling plan mode, for example, the plan of the following months is updated monthly according to the latest actual sales and inventory data. The core of the method is to predict the point of the demand based on a time sequence prediction model (such as ARIMA and exponential smoothing method) and generate a single optimal planning scheme according to the point. In combination with traditional operational planning methods, some optimization methods introduce linear programming or integer programming models to solve for optimal throughput given the demand predictions and objective functions (e.g., lowest cost, greatest profit). In addition, in the planning evaluation link, tools such as data envelope analysis (Data Envelopment Analysis, DEA) and analytic hierarchy Process (ANALYTIC HIERARCHY Process, AHP) are used to evaluate the efficiency of the production scheme or make multi-index decisions, respectively. The existing methods can play a role when the market demand is relatively stable or changes slowly, and also provide a basic framework for the production plan management of enterprises. However, despite the application of the existing methods in the planning of the production of large quantities of commodity months, the following technological limitations of gradual exposure still exist when faced with dynamic changes in market demand: 1. the existing methods lack the ability to dynamically identify and predict market demand conditions. The method depends on static historical data or single scene prediction, and can not quantitatively capture the transition rule of the market state, so that the plan adjustment is often delayed from the actual demand change, and the plan and the actual demand are easy to be disjointed. 2. Due to the lack of modeling of the multi-state evolution of demand, existing methods have difficulty systematically generating alternative plans covering different market scenarios and risk preferences. Enterprises lack diversified and flexibly selectable plan sets when facing uncertainty, and production target optimization under risk control is difficult to realize. 3. In a planning evaluation link, the existing evaluation means cannot deeply integrate efficiency quantification and multi-objective decision, the evaluation process often depends on manual experience or single index, and an objective evaluation system for comprehensively considering the demand matching degree, the resource utilization rate and the production stability is lacking, so that an evaluation result is one-sided and high in subjectivity. In summary, the existing method has the problem of insufficient overall adaptability in the establishment of the large-scale commodity month production plan, is difficult to realize rapid and closed-loop month production plan dynamic optimization and adjustment when the external market frequently fluctuates, and limits the agility of enterprises to deal with market changes and the utilization efficiency of production resources. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a month production plan adjustment intelligent optimization method and system facing the influence of market demands by constructing a systematic technical scheme of a fused hidden Markov model (Hidden Markov Model, HMM) demand prediction model, a multi-scene plan generation model and a production plan utility evaluation model, and aims to realize dynamic optimization of a production plan. On one hand, the invention provides an intelligent optimization method for month production plan adjustment facing to market demand influence, which comprises the following steps: Obtaining market observation data, enterprise historical operation data and production resource and constraint data of a target bulk commodity; constructing a two-dimensional observation vector sequence of the target bulk commodity based on the market observation data; building and training an HMM demand prediction model to obtain a trained HMM deman