An energy dynamic prediction and multi-energy flow coupling optimization method and an energy management and control system

Abstract

The application provides an energy dynamic prediction and multi-energy flow coupling optimization method and an energy management and control system, and comprises the following steps: S1, obtaining energy production and consumption historical data, a work condition type list and future production and maintenance plans; S2, using an optimized long short-term memory neural network with a fused self-attention mechanism to output a prediction result; S3, using a Gantt chart work condition management module to realize automatic calling of a work condition sample curve or a work condition instance changing with a work condition rhythm; S4, superimposing real-time data, abnormal work condition information and a work condition sample curve, and combining historical work condition information to make corrections; S5, establishing a multi-energy coupling optimization scheduling model, and using the multi-energy coupling optimization scheduling model to schedule the final dynamic prediction result in S4. The application provides a practical and effective decision-making scheme for production scheduling personnel, and finally assists a steel enterprise in reducing energy dissipation, comprehensively improving system energy efficiency and economic benefits, and reducing production and operation costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of comprehensive energy scheduling of steel enterprises, in particular, and more particularly to an energy dynamic prediction and multi-energy flow coupling optimization method and an energy management and control system. BACKGROUND

[0002] In the steel production process, not only a large amount of energy and resources will be consumed, but also a large amount of secondary energy such as coal gas, steam and electricity will be produced. These three kinds of energy are interrelated and coupled with each other, and especially the production and consumption will change with the change of working condition and equipment running state, which increases the difficulty of energy system scheduling and management. Only relying on manual experience for utilization, it is difficult to utilize reasonably, resulting in resource waste and environmental pollution. Therefore, the steel enterprise not only needs to establish an optimization model to manage and optimize the allocation of coal gas, steam and electricity, but also needs to predict the imbalance of supply and demand of coal gas, steam and other energy media in advance, and to pre-control, which will greatly reduce the unnecessary dissipation of energy.

[0003] Under the current background of comprehensive digital transformation, data mining technology is increasingly widely used in various fields. As one of the key application fields, steel enterprise energy scheduling management has been significantly developed with the support of data mining technology. With the continuous maturity of deep learning models and the continuous upgrading of intelligent optimization algorithms, this technology gradually helps the actual production scheduling to make guidance based on more scientific and accurate data. Some scholars use scenario analysis method to establish real-time adjustment feedback model to cope with the adverse effects of the difference in response time of different energy flows on system optimization scheduling. Some scholars train LSSVM model based on historical data, and use the trained optimal prediction model to predict part of the historical data of coal gas production and consumption, so as to improve the dynamic prediction accuracy. Some scholars propose an online Wang-Mendel fuzzy method to extract knowledge from historical data and realize online learning ability.

[0004] However, the above studies all ignore the influence of uncertain factors in actual production, resulting in low prediction accuracy and unreasonable decision instructions. First, from the perspective of energy medium prediction, many composite prediction models only study the prediction problem under a single scenario, without considering the complex working conditions during equipment operation, and only analyze a certain link in the energy system, with less consideration of external factors (production plan, maintenance plan, process conditions, working condition information, etc.). Most are offline prediction modeling, which cannot be adjusted with the dynamic changes of production conditions. The corresponding prediction model does not consider the key information of relevant feature points when performing time series prediction, and is not corrected, resulting in low prediction accuracy and poor stability of the model. Secondly, from the perspective of optimization scheduling model, most models do not consider the superposition of multiple production conditions from the whole plant level, and do not consider that there are many uncertain factors in the production process, such as intermittent fluctuations in production rhythm, demand prediction deviation of various loads on the load side, and changes in production plan demand uncertainty, which will affect the normal operation of the energy system. There are many problems with the current intelligent energy system used by steel enterprises. Visualization cannot fully meet the demand for information presentation, operability is not strong, prediction is not accurate, and scheduling instructions are frequent, which leads to missing the best adjustment time. SUMMARY

[0005] Therefore, the purpose of the present application is to provide an energy dynamic prediction and multi-energy flow coupling optimization method and an energy management and control system, to establish a fusion of data-driven and deep learning, to realize a visual, multifunctional and convenient-to-operate energy dynamic prediction and optimization scheduling intelligent energy management and control system, to realize accurate prediction of future trends of energy instantaneous production and consumption of each energy equipment, accurate prediction of future trends of total instantaneous production and consumption of various coal gas and steam in the whole steel enterprise, and emergency response prediction of energy production and consumption trends under sudden abnormal working conditions.

[0006] The technical means adopted by the present application are as follows:

[0007] An energy dynamic prediction and multi-energy flow coupling optimization method, comprising the following steps:

[0008] S1, obtaining energy production and consumption historical data, working condition type list, and future scheduling period production and maintenance plan;

[0009] S2, preprocessing the historical data to obtain preprocessed historical data; using a template matching method to extract periodic features in the preprocessed historical data; obtaining strongly correlated feature points based on the periodic features; constructing a long short-term memory neural network short prediction model fused with self-attention mechanism and training the model; using a grey wolf optimization algorithm to optimize the trained long short-term memory neural network short prediction model fused with self-attention mechanism; and using the optimized long short-term memory neural network fused with self-attention mechanism to output prediction results;

[0010] S3, based on the prediction result of S2, a working condition sample library is established, a Gantt chart working condition management module is established, a device working condition type is defined in the Gantt chart working condition management module according to the energy equipment and the working condition type list, a working condition sample curve is defined based on the device working condition type, and the Gantt chart working condition management module is used to realize automatic calling of the working condition sample curve or the working condition instance with the working condition rhythm change;

[0011] S4, real-time data of energy production and consumption of different equipment in the SCADA system is collected, abnormal working condition information is collected, the real-time data, the abnormal working condition information and the working condition sample curve are superimposed to obtain the future trend curve of the energy production and consumption prediction curve of the equipment and the total production and consumption of coal gas and steam under multiple working conditions; the historical data sequence with the same working condition condition as the prediction day is found out as the correction information of the prediction curve, the production and consumption prediction curve and the future trend curve are corrected according to the correction information, the energy production and consumption prediction curve of different equipment, the coal gas tank position prediction curve and the prediction curve of the total production and consumption of coal gas and steam are obtained, and finally the dynamic prediction is realized;

[0012] S5, the scheduling model variables and the model constraint conditions are determined, the parameter value method of the cost term and the penalty term in the scheduling function is determined, the efficiency characteristic equation of the equipment is obtained by using a data-driven method combined with the historical production and consumption data of the equipment, the objective function is comprehensively optimized, a multi-energy coupling optimization scheduling model is established, the energy allocation of each equipment and the total energy change trend in the future are obtained based on the energy dynamic prediction curve in S4, that is, the production and consumption of each equipment in the future, and the multi-energy coupling optimization scheduling model is called to give an energy optimization allocation scheme based on the production and consumption of each equipment in the future.

[0013] Further, S1 specifically includes:

[0014] The enterprise comprehensive data integration platform server is used to obtain the enterprise energy topology network structure, the energy production and consumption historical data, the types and quantities of energy media generated and consumed under different working conditions are obtained through the enterprise comprehensive data integration platform server, and the production plan and the maintenance plan in the future scheduling period are obtained through the enterprise comprehensive data integration platform server.

[0015] The enterprise energy topology network structure and the energy production and consumption historical data include the boiler coal gas consumption and the steam generation, the turbine unit steam inlet and steam extraction and power generation, the coal gas diffuser tower diffuser, and the coal gas tank real-time tank position.

[0016] Further, S2 specifically includes the following steps:

[0017] S201, the historical data of the influence of different types of working conditions on the energy production and consumption of the equipment is preprocessed to obtain preprocessed historical data; the preprocessing method includes time series stabilization, data repair, serialized data, normalization and data segmentation;

[0018] S202, determine the periodic characteristics in the pre-processed historical data, construct a periodic matching template, match the periodic matching template with the pre-processed historical data, and extract specific periodic characteristics in the pre-processed historical data; measure the matching degree of the template and the data through the Pearson correlation coefficient, and obtain strong correlation feature points; extract the key information of the strong correlation feature points for time series prediction;

[0019] The calculation formula of the Pearson correlation coefficient is as follows:

[0020]

[0021] In the formula, r xy is the correlation coefficient, X is the independent variable, and Y is the dependent variable;

[0022] S203, construct a long short-term memory neural network short prediction model fused with a self-attention mechanism to fully extract the dependency between time series;

[0023] The self-attention mechanism algorithm process is as follows:

[0024]

[0025] In the formula, Tr, Tm, and Tk are three length h vectors in the sample, which are combined into a matrix in parallel, and the Softmax function is an important activation function in the field of deep learning. The Softmax function converts the original score output by the neural network into a probability distribution of each class, so that the model can classify different classes in the form of probability;

[0026] S204, use the grey wolf optimization algorithm to optimize the parameters of the long short-term memory neural network short prediction model fused with the self-attention mechanism, train the long short-term memory neural network short prediction model fused with the self-attention mechanism with the optimal prediction ability, and output the prediction results of the influence of different working conditions on the energy production and consumption of the equipment;

[0027] The specific formula of the grey wolf optimization algorithm is as follows:

[0028]

[0029] In the formula, represents the predicted value of production and consumption, and y i represents the true value of production and consumption.

[0030] Further, S3 specifically includes the following steps:

[0031] S301, define the working condition type according to the equipment type, and record each working condition type in the Gantt chart management module;

[0032] Define the working condition type for the main gas production and consumption unit, including blast furnace replacement, blast furnace blowdown, coke oven maintenance, converter heat, converter maintenance, and rolling maintenance, for selection when recording working conditions;

[0033] S302, define the production and consumption curve of various media for various working conditions of energy equipment, and use the production and consumption curve as the working condition sample curve of the change of energy medium type and quantity of the equipment with time;

[0034] S303, each type of working condition of each device is one-to-one corresponding to the working condition sample curve or working condition instance of the working condition, and the working condition sample curve or working condition instance is automatically called according to the working condition rhythm;

[0035] When an abnormal working condition occurs, a certain production and consumption medium sample curve, i.e. working condition instance data, is directly recorded according to the working condition, and the priority of the working condition instance data is higher than that of the sample data.

[0036] Further, S4 specifically includes the following steps:

[0037] S401, collect real-time data in the SCADA system and pre-process the real-time data;

[0038] S402, select energy equipment and its working condition type that need to be automatically tracked, determine the working condition signal collection site in the steel plant, establish a working condition rhythm prediction logic based on the production / maintenance plan and real-time working condition signal, and automatically track and correct the working condition rhythm;

[0039] S403, introduce the abnormal working condition in the production / maintenance plan in the future period of time by dragging the Gantt bar; the imbalance caused by the superposition of abnormal working conditions in a certain period of time is as follows:

[0040]

[0041] In the formula, i, m, and z represent the i-th time point, the m-th energy medium, and the z-th working condition, respectively, (t i ,t i+1 ) is a certain period of time in h-1 periods;

[0042] S404, use K-neighbor algorithm to find out the historical data production and consumption change quantity correction prediction curve with the same working condition condition as the prediction day by deeply mining the energy historical data in the system, correct the production and consumption prediction curve and the future trend curve according to the correction information, and finally realize dynamic prediction.

[0043] Further, S5 includes the following steps:

[0044] S501, clarify the enterprise energy dispatching rules and security supply requirements, including the safe pressure of coal gas / steam pipe network operation, the safe position of coal gas tank, the safe pressure of steam accumulator, the safe load of hot blast stove and heating furnace, the minimum load of generator set, the safety range of purchased electricity demand;

[0045] S502, compile the list of adjustable energy equipment, including coal gas / steam generator set and coal gas tank, set the adjustment priority, establish the performance database of adjustable energy equipment, and track the real-time operation state of adjustable equipment;

[0046] S503, according to the corresponding relationship between the energy conversion efficiency of each boiler in a certain period and the operation load of the boiler, the operation load of the energy conversion equipment and the energy conversion efficiency under the corresponding load are characterized as a quadratic function by the method of quadratic curve fitting, and according to the test data and online production data of each steam turbine, the consumption characteristic equation of the steam turbine is established by using the principle of multiple linear regression, the formulas are as follows:

[0047]

[0048] In the formula, represents the actual load of the energy conversion equipment when it is running stably, C1, C2, C3 represent the fitting parameters; respectively represent the steam inlet quantity, steam extraction quantity and condensate quantity of the steam turbine in the t time period;

[0049] S504, according to the list of adjustable equipment, combining with the energy dispatching rules and security supply requirements, determine the dispatching model variables and model constraint conditions, and establish a multi-energy coupling optimization dispatching model;

[0050] S505, based on the energy dynamic prediction curve in S4, the energy distribution of each equipment and the total energy change trend in the future period are known, and the multi-energy coupling optimization dispatching model is called to give the energy optimization distribution scheme.

[0051] Further, the multi-energy coupling optimization dispatching model includes an objective function and a constraint condition, the dispatching cycle of the multi-energy coupling optimization dispatching model is T, a certain time period is recorded as t (1≤t≤T), D is the number of different equipment, N represents the number of different energy media, then the objective function formula of the model is as follows:

[0052]

[0053] In the formula, C buy,j , E j,t respectively represent the purchase price of fuel j and the purchase quantity of fuel j in the t period; C os,t , P trespectively represent the electricity price and the purchased electricity quantity in peak, valley and flat periods; C adj,j , ΔF i,j,t respectively represent the adjustment penalty coefficient of fuel j and the adjustment quantity of equipment i to fuel j in period t; C re,k , D k,t respectively represent the diffusion penalty coefficient of coal gas k and the diffusion quantity of coal gas k in period t; C st,i , L i,t respectively represent the shutdown cost of equipment i and the shutdown flag of equipment i, 0 represents shutdown and 1 represents start-up; C de,k , H r,k,t , H g,k respectively represent the rack deviation penalty coefficient, the rack height of coal gas k at time t and the target rack height;

[0054] The constraint conditions include equipment range constraints, equipment node balance constraints, supply and demand balance constraints and heat value balance constraints;

[0055] The equipment range constraints include equipment energy production and consumption upper and lower limit constraints, coal gas tank rack upper and lower limit constraints and equipment climbing speed upper and lower limit constraints;

[0056] The equipment energy production and consumption upper and lower limit constraints are as follows:

[0057] G i,j,min ≤G i,j,t ≤G i,j,max

[0058] The coal gas tank rack upper and lower limit constraints are as follows:

[0059] H k,min ≤H k,t ≤H k,max

[0060] The equipment climbing speed upper and lower limit constraints are as follows:

[0061] M i,speed,min ≤M i,speed,t ≤M i,speed,max

[0062] In the formula, G i,j,t represents the energy production or consumption quantity of equipment i, H k,t represents the rack height of coal gas k, and M i,speed,t represents the climbing speed of equipment i;

[0063] The equipment node balance constraints are as follows:

[0064]

[0065] In the formula, n ef,i,jConversion coefficient of energy j in device i into energy p, m represents the number of energy types consumed by device i, E in,i,j represents the consumption of energy j in device i, h j represents the calorific value of energy j, F out,i,p represents the amount of energy p generated by device i, h p represents the calorific value of energy p;

[0066] Supply-demand balance constraint:

[0067]

[0068] In the formula, (ΔT i , ΔT i+1 ) is a period in ΔT, W j,Δt represents the amount of imbalance introduced by the working condition within ΔT, represents the amount of energy j generated in the period (ΔT i , ΔT i+1 ), represents the change in the storage energy j cabinet in the period (ΔT i , ΔT i+1 ), if energy j is not gas, the value is 0, respectively represent the emission amount and consumption amount of energy j in the period (ΔT i , ΔT i+1 );

[0069] Calorific value balance constraint:

[0070]

[0071] In the formula, X i , Y i respectively represent the minimum specific calorific value and the maximum specific calorific value of the total calorific value of the energy consumed by device i;

[0072] Under different production working conditions, various energy production / consumption devices, energy conversion devices, and gas cabinets will make corresponding adjustments, and the adjustment penalty coefficient of each device is equivalent to providing priority for the adjustment sequence of the device, that is, under the condition that other conditions are the same, the model will preferentially adjust the device with a small adjustment penalty coefficient, and the deviation penalty coefficient can ensure that the gas cabinet position is maintained around the target cabinet position.

[0073] The application also provides an energy dynamic prediction and multi-energy flow coupling optimization energy management and control system for realizing any one of the above energy dynamic prediction and multi-energy flow coupling optimization methods, comprising:

[0074] A data acquisition and cleaning module, an energy management and diagnosis module, a Gantt chart working condition management module, an energy online prediction and optimization scheduling module, and an energy scheduling decision pushing module.

[0075] The data collection and cleaning module is used to perform data collection and data cleaning steps:

[0076] The data collection step is as follows:

[0077] Using advanced sensor technology and Internet of Things technology, the SCADA system collects real-time energy production and consumption data of various devices, converts data of different sources and formats into a unified data type through the use of ETL tools, ensures data consistency and compatibility, and performs preliminary processing and filtering, reduces the transmission of invalid data, and improves data processing efficiency;

[0078] The data cleaning step is as follows:

[0079] T1, missing value processing: the system automatically detects missing values in the data set, and for completely random missing data and non-critical data, uses the method of deleting according to the unified time granularity of the model to process, thereby saving data processing time; when processing random missing data and non-random missing data (MNAR), use the random forest method to fill in the missing data, thereby fully restoring the true situation of the original data.

[0080] T2, abnormal data identification and correction: use KMeans clustering algorithm to cluster the data, count the number of data points in each cluster, set a threshold, if the cluster number is less than the threshold, consider that the data point is an outlier, and take the average value interpolation method to correct;

[0081] T3, data deduplication: through similarity matching technology, identify and remove duplicate data, ensure the uniqueness and accuracy of the data set;

[0082] T4, data standardization: the system standardizes the data, including unit conversion, dimension unification, etc., to ensure data consistency and comparability;

[0083] The energy management and diagnosis module is used to perform energy management and diagnosis functions;

[0084] The energy management is as follows: establish a data dictionary to record data types, units, and other information, provide data query and use functions; record data sources, collection time, collection method, and other metadata information to ensure data transparency and traceability, trace the data source and collection process through metadata information, when data is abnormal, trace to specific sensor equipment and collection time through metadata, find out the problem cause, solve the data abnormality problem, use the distributed architecture of MySQL to realize storage function, and use the master-slave replication mechanism and asynchronous replication mode provided by MySQL;

[0085] The diagnosis function is as follows: the types and consumption of energy used by the enterprise are statistically analyzed, the main energy consumption points and energy saving potential are found out, the efficiency of the main production equipment of the enterprise is evaluated, and the problems of low equipment efficiency and energy waste are found out;

[0086] The Gantt chart working condition management module has the following functions: working condition samples are recorded into the system through the Gantt chart working condition management module, the recorded working condition sample information is displayed through a movable Gantt chart, and the imbalance amount is introduced by superimposing the working condition samples to simulate the change amount of the equipment production and consumption under the actual production plan condition.

[0087] The energy online prediction and optimization scheduling module displays the coal gas / steam production and consumption dynamic prediction curve through a visual chart, the production and consumption prediction curve includes various types of coal gas and steam production and consumption curves of various energy equipment, various types of coal gas and steam production and consumption curves of the whole plant, coal gas / steam pipe network pressure prediction curve, blast furnace / coke oven / convertor gas tank position prediction curve; based on the energy dynamic prediction curve, the user calls the multi-energy coupling optimization scheduling model to give an energy optimization distribution scheme, at this time the visual chart displays the equipment energy production and consumption, the energy total amount change trend and the tank position change trend in the future period after the implementation scheme is displayed;

[0088] The energy scheduling decision pushing module, when the energy online prediction and optimization scheduling module generates a new scheduling decision, displays the scheduling decision and optimization suggestion in the form of charts and texts through the system interface, for the scheduling decision confirmed by the user, the system automatically performs the corresponding adjustment operation; save all pushed scheduling decisions and their execution situations to form a historical record, analyze the execution effect of the historical decision, and evaluate the accuracy and effectiveness of the optimization suggestion to improve the future scheduling model and algorithm.

[0089] The application also provides a storage medium, which comprises a stored program, wherein the program runs to execute any one of the above energy dynamic prediction and multi-energy flow coupling optimization methods.

[0090] The application also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and the processor runs to execute any one of the above energy dynamic prediction and multi-energy flow coupling optimization methods through the computer program.

[0091] Compared with the prior art, the application has the following advantages:

[0092] The application explores the user load characteristics, the equipment energy medium production and consumption characteristics and the influence of external factors on the change trend of the two based on big data analysis technology, adaptively extracts specific periodic characteristics based on data driving and using a template matching method, and trains a prediction model;

[0093] The application adopts a "double correction" mode to improve the prediction model precision.The grey wolf optimization algorithm is used to correct the prediction model, and real-time data in the SCADA system are collected, advanced technologies such as big data mining and deep learning are used to deeply mine the energy historical data in the system, and the K-neighbor algorithm is used to correct the prediction curve.

[0094] The stability and rationality of the energy optimization scheduling model are crucially dependent on the accuracy of the prediction, the application uses advanced technologies such as big data technology, data mining and deep learning to establish a prediction model, and performs secondary correction, greatly improving the accuracy of the prediction, by using the application, the future working condition rhythm and the production and consumption of coal gas and steam in a period of time can be accurately predicted in advance, buffer time is left for unexpected situations, thereby improving the stability and rationality of the optimization scheduling model;

[0095] The application establishes a working condition sample library based on the GWO optimized Self Attention-LSTM prediction model to construct equipment production and consumption prediction curves under different production conditions, and the working condition sample is not fixed, but can be corrected after identifying the current production condition according to the actual operation of the system.

[0096] The application establishes a multi-energy coupling optimization scheduling model, considers uncertain factors, and introduces a cabinet position deviation penalty method to better adapt to the actual operation of the equipment under different production conditions of the enterprise, and uses an efficient algorithm to adapt to the needs of dynamic scheduling.

[0097] The application overcomes the defects of traditional research that only balances and schedules a single energy medium, and considers uncertain factors under various production conditions, equipment operation characteristics and energy balance constraints, and introduces priority and penalty functions to effectively avoid frequent adjustment of the equipment under normal production conditions, and the output optimization scheduling instruction can meet the specific requirements of pipeline network operation safety, production and supply safety and equipment use safety.

[0098] The Gantt chart working condition management control developed by the application can realize manual entry and adjustment of working conditions, and the system has the functions of automatic data acquisition, automatic acquisition of working condition related signals, autonomous operation of the scheduling model, and pushing of optimization scheduling instructions, so that the energy scheduling personnel can manage the energy system more conveniently.

[0099] The application develops an energy medium production and consumption prediction and intelligent scheduling system that integrates data driving, deep learning, intelligent scheduling and uses a dynamic prediction method, realizes intelligent control, monitoring and management of the energy of a steel enterprise. BRIEF DESCRIPTION OF DRAWINGS

[0100] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0101] Figure 1 The method flowchart of the present application.

[0102] Figure 2 The GWO optimization Self Attention-LSTM prediction model construction flowchart of the present application.

[0103] Figure 3 The historical prediction data and production rhythm working condition sample library establishment flowchart of the present application.

[0104] Figure 4 The coal gas network operation structure schematic diagram of the present application.

[0105] Figure 5 The energy dynamic prediction and multi-energy flow coupling optimization method flowchart of the present application.

[0106] Figure 6 The integrated energy management and control system structure schematic diagram of the present application. DETAILED DESCRIPTION

[0107] In order to make the person skilled in the art better understand the present application scheme, the following will combine the drawings in the embodiments of the present application, and the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0108] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0109] The purpose of the present application is to invent a set of integrated energy management and control systems that combine data-driven, deep learning, intelligent scheduling, and energy dynamic prediction methods to achieve efficient and intelligent energy management. From the perspective of the whole plant, the prediction model is optimized considering the superposition of various production conditions, uncertain factors in the production process, and the interdependence between different time points, to improve the prediction accuracy, model stability, and the rationality of the scheduling instructions. Based on big data analysis technology, the present application explores the user load characteristics, the energy medium production and consumption change law, and the influence of external factors on the change trend of the two. The Self Attention-LSTM neural network model (Self Attention-LSTM) with optimized fusion self-attention mechanism greatly improves the prediction accuracy. The present application establishes a multi-energy flow coupling optimization model. Since most steel plants use SCADA systems to accumulate a large amount of real-time data, the present application proposes a dynamic optimization scheduling scheme, and finally establishes an integrated data-driven, deep learning, visual, multifunctional, and easy-to-operate energy dynamic prediction and optimization scheduling intelligent energy management and control system, which realizes the accurate prediction of the future trend of energy instantaneous production and consumption of various energy equipment, the accurate prediction of the future trend of the total instantaneous production and consumption of various coal gas and steam in the whole plant of the steel enterprise, the emergency response prediction of the energy production and consumption trend under abnormal conditions, provides effective decision-making schemes for production scheduling personnel, and finally helps steel enterprises to reduce energy dissipation, improve system energy efficiency and economic benefits, and reduce production and operation costs.

[0110] As shown in Figure 1 The present embodiment discloses an energy dynamic prediction and multi-energy flow coupling optimization method and an energy management and control system, comprising the following steps:

[0111] S1: Analyze the energy system layout of the steel enterprise, obtain energy production and consumption historical data and working condition type list, determine the energy medium production and consumption of each device under various working conditions, and obtain the production and maintenance plans for the future scheduling period;

[0112] S101: Obtain the enterprise energy topology network structure and energy production and consumption historical data through the enterprise comprehensive data integration platform server, including boiler gas consumption and steam production, generator set steam intake and steam extraction, and power generation, coal gas dissipation tower dissipation, coal gas tank real-time tank position, and obtain the types and quantities of energy media generated and consumed under different working conditions and their change rules over time, and obtain the production and maintenance plans for the future scheduling period;

[0113] S2: Based on big data analysis technology, explore the user load characteristics, equipment energy medium production and consumption characteristics and the influence of external factors on the change trend of the two, adopt a new template matching method to extract specific periodic characteristics based on data-driven self-adaptation, train Self Attention-LSTM neural network and use GWO algorithm to optimize Self Attention-LSTM prediction model, so that the model is in the optimal state, and the prediction results of the influence of different types of working conditions on the production and consumption of equipment energy are output, such as Figure 2 The GWO optimized Self Attention-LSTM prediction model flow chart of the application is shown in the figure;

[0114] S201: The historical data of the influence of different types of working conditions on the production and consumption of equipment energy is preprocessed, and the data is processed by time series stabilization, data repair, serialized data, normalization, data segmentation and the like;

[0115] S202: Determine the periodic characteristics in the data, construct a periodic matching template, match the constructed template with the data, extract specific periodic characteristics in the data, and adaptively adjust the template according to the actual data changes. The matching degree of the template and the data is measured by Pearson correlation coefficient, and the key information of the strong correlation feature points is extracted when the time series prediction is performed, so that the prediction is more accurate. The strong correlation feature points are measured according to the Pearson correlation coefficient, and the value is from -1 to 1, 1 indicates positive correlation, -1 indicates negative correlation, and the strong correlation feature points need to be analyzed according to the specific data, such as the characteristic between 0.6 and 0.8. Then 0.7-0.8 is considered as a strong correlation feature point.

[0116] The calculation formula of Pearson correlation coefficient is as follows:

[0117]

[0118] In formula (1), r xy is the correlation coefficient, X is the independent variable, and Y is the dependent variable;

[0119] S203: Based on the above analysis, a long short-term memory neural network short prediction model is constructed, which fully extracts the dependency relationship between time series;

[0120] The self-attention mechanism algorithm process is as follows:

[0121]

[0122] In formula (2), Tr, Tm, Tk are three h-length vectors combined in parallel into a matrix, and the Softmax function is an important activation function in the field of deep learning, which can convert the original score output by the neural network into a probability distribution of each class, so that the model can classify different classes in the form of probability;

[0123] S204: The grey wolf optimization algorithm is used to optimize the parameters of the prediction model, and the model with the optimal prediction ability is trained to output the prediction results of the influence of different working conditions on the energy production and consumption of the equipment;

[0124] Grey Wolf Optimizer (GWO) is an intelligent optimization algorithm, which has the advantages of not relying on gradient information, strong global search ability, fast convergence speed, etc. Its optimization process is divided into social hierarchical layering, surrounding prey, hunting, and attacking prey.

[0125] The model uses Mean Absolute Percentage Error (MAPE) and Root Mean Squared Error (RMSE) to evaluate the prediction results, and the specific formulas are as follows:

[0126]

[0127] In formulas (3) and (4), represents the predicted value of production and consumption, and y i represents the true value of production and consumption.

[0128] S3: Based on the above historical working condition prediction data, a working condition sample library is established, which includes working condition type, interval time, duration, working condition sample curve setting, etc. A Gantt chart working condition management module is developed. Combined with the process mechanism of coal gas and steam production and consumption, the working condition management module defines the working condition type of the equipment according to the list of energy equipment and its working condition type, defines the variety and quantity of energy medium consumed by the equipment under a certain working condition with time, and defines the working condition combination of the equipment in a period of time. The process of establishing a working condition sample library based on historical prediction data and production rhythm is shown in Figure 3 ;

[0129] S301: Define the working condition type according to the equipment type, such as normal operation, maintenance, wind-off, shutdown, etc. Each working condition type will be entered into the Gantt chart management module. The more working condition types, the more detailed the management of the running mode of the equipment.

[0130] Define the working condition types of main gas production and consumption units such as coke oven, blast furnace, converter and rolling mill, including blast furnace replacement, blast furnace blowdown, coke oven maintenance, converter heat, converter maintenance, rolling mill maintenance, etc. These working condition types can be selected when entering the working condition;

[0131] S302: Define the production and consumption curves of various media for various working condition types of energy equipment. Use the working condition prediction data production and consumption curves as the working condition sample curves of the energy medium types and quantities of the equipment changing with time;

[0132] S303: Correspond each type of working condition of each device to the working condition sample curve or working condition instance of the working condition, and realize the automatic calling of the working condition sample curve or working condition instance with the working condition rhythm;

[0133] For abnormal working conditions, a production and consumption medium sample curve, i.e. working condition instance data, can also be directly entered with the working condition. The priority of working condition instance data is higher than that of sample data;

[0134] S4: Collect real-time data in SCADA system, and obtain abnormal working condition information of the equipment through maintenance plan, etc. Combine the working condition sample curves entered in the working condition sample library to obtain the future trend of the energy production and consumption curves of the equipment and the total production and consumption of gas and steam under multiple working conditions. Through advanced technologies such as data mining and deep learning, the energy historical data in the system is deeply mined. The K- nearest neighbor algorithm is used to find the historical data sequence with the same working condition as the prediction day as the correction information of the prediction curve, and finally the dynamic prediction is realized;

[0135] S401: The real-time data collected by SCADA in the field may have data missing or redundancy, so the data needs to be preprocessed to ensure the integrity and reliability of the data;

[0136] S402: Select the energy equipment and its working condition type that needs to be automatically tracked, determine the working condition signal collection site in the steel plant, establish a working condition rhythm prediction logic based on production / maintenance plan and real-time working condition signal, and automatically track and correct the working condition rhythm;

[0137] S403: Introduce the abnormal working condition in the future period of time in the production / maintenance plan by dragging the Gantt bar. The imbalance caused by the superposition of abnormal working condition in a certain period of time is as follows:

[0138]

[0139] In formula (5), i, m, z represent the i th time point, the m th energy medium, and the z th working condition. (t i ,t i+1 ) is a certain period of time in h-1 periods;

[0140] S404: Deep mining of energy history data in the system, using K-neighbor algorithm to find historical data with the same working condition as the prediction day, correcting the prediction curve, and finally realizing accurate prediction of future trends of instantaneous production and consumption of coal gas / steam of each energy equipment, accurate prediction of future trends of total instantaneous production and consumption of various types of coal gas and steam in the whole steel enterprise, and accurate prediction of future trends of coal gas / steam pipe network pressure, coal gas tank level, and steam accumulator pressure, and emergency response prediction of coal gas / steam production and consumption trend under abnormal working conditions;

[0141] S5: According to the list of adjustable equipment, combining energy dispatching rules and safety supply requirements, determining the dispatching model variables and model constraint conditions, determining the parameter value method of the cost term and the penalty term in the dispatching function by comprehensively considering factors such as peak-valley electricity price, adjustable equipment adjustment cost, adjustable equipment start-stop cost, coal gas emission cost, and coal gas tank level overrun maintenance cost; combining with the historical production and consumption data of the equipment to obtain the efficiency characteristic equation of the equipment in a data-driven manner, and comprehensively optimizing the objective function to establish a multi-energy coupling optimization dispatching model;

[0142] S501: Clarify the energy dispatching rules and safety supply requirements of the enterprise, refer to the coal gas network operation structure of the present application shown in Figure 4 , which includes coal gas / steam pipe network operation safety pressure, coal gas tank safety level, steam accumulator safety pressure, hot blast furnace blow safety production load, heating furnace safety production load, generator set minimum load, and purchased electricity demand safety range;

[0143] S502: Compile a list of adjustable energy equipment, including coal gas / steam generator set, coal gas tank, etc., and set the adjustment priority, establish an adjustable energy equipment performance database, and track the running state of the adjustable equipment in real time;

[0144] S503: According to the corresponding relationship between the energy conversion efficiency of each boiler in a certain period and the boiler operation load, the operation load of the energy conversion equipment and the energy conversion efficiency under the corresponding load are characterized by a quadratic function through a quadratic curve fitting method, and according to the test data and online production data of each steam turbine, a consumption characteristic equation of the steam turbine is established by using the principle of multiple linear regression, and the formulas are as follows:

[0145]

[0146] In formula (6), represents the actual load of the energy conversion equipment when it is stably running, C1, C2, and C3 represent fitting parameters; in formula (7), respectively represent the steam intake, steam extraction, and condensate amount of the steam turbine in the t time period;

[0147] S504: According to the adjustable device list, combined with the energy dispatching rules and the security supply requirements, the dispatching model variables and the model constraint conditions are determined, the objective function is comprehensively optimized, and a multi-energy coupling optimization dispatching model is established; refer to Figure 5 The energy flow prediction and multi-energy flow coupling optimization method flow chart is shown in FIG. 1;

[0148] The multi-energy coupling optimization control model includes an objective function and a constraint condition. It is assumed that the dispatching period of the model is T, a certain time period is t (1≤t≤T), D is the number of different devices, and N represents the number of different energy media. The objective function formula of the model is as follows:

[0149]

[0150] In formula (8), C buy,j , E j,t respectively represent the purchase price of fuel j and the purchase amount of fuel j at time t; C os,t , P t respectively represent the electricity price at peak, valley and flat periods and the purchased electricity amount at peak, valley and flat periods; C adj,j , ΔF i,j,t respectively represent the adjustment penalty coefficient of fuel j and the adjustment amount of device i to fuel j within the time period t; C re,k , D k,t respectively represent the diffusion penalty coefficient of gas k and the diffusion amount of gas k within the time period t; C st,i , L i,t respectively represent the shutdown cost of device i and the shutdown flag of device i, 0 represents shutdown, and 1 represents activation; C de,k , H r,k,t , H g,k respectively represent the cabinet position deviation penalty coefficient, the cabinet position height of gas k at time t, and the target cabinet position height.

[0151] The constraint conditions include device range constraints, energy balance constraints, supply and demand balance constraints, and calorific value constraints.

[0152] Device range constraints:

[0153] G i,j,min ≤G i,j,t ≤G i,j,max (9)

[0154] H k,min ≤H k,t ≤H k,max (10)

[0155] M i,speed,min ≤M i,speed,t ≤M i,speed,max (11)

[0156] The above formulas (9), (10), (11) are respectively the upper and lower limit constraints of the energy consumption of the device, the upper and lower limit constraints of the cabinet position of the coal gas cabinet, and the upper and lower limit constraints of the climbing speed of the device; wherein G i,j,t represents the amount of energy generated or consumed by the device i, H k,t represents the cabinet position height of the coal gas k, M i,speed,t represents the climbing speed of the device i;

[0157] Device node balance constraint:

[0158]

[0159] In formula (12), n ef,i,j represents the conversion coefficient of energy j in device i to energy p, m represents the number of energy types consumed by device i, E in,i,j represents the consumption amount of energy j of device i, h j represents the calorific value of energy j, F out,i,p represents the amount of energy p generated by device i, h p represents the calorific value of energy p;

[0160] Supply and demand balance constraint:

[0161]

[0162] In formula (13), (ΔT i , ΔT i+1 ) is a period in ΔT, W j,Δt represents the imbalance amount introduced by the working condition in ΔT, represents the amount of energy j generated in (ΔT i , ΔT i+1 ) period, represents the change amount of the cabinet position of the storage energy j in (ΔT i , ΔT i+1 ) period, if energy j is not coal gas, the value is 0, respectively represent the dissipation amount and consumption amount of energy j in (ΔT i , ΔT i+1 ) period;

[0163] Calorific value balance constraint:

[0164]

[0165] In formula (14), X i , Y i respectively represent the minimum specific heat value and the maximum specific heat value of the total heat value of the energy consumed by device i;

[0166] Under different production conditions, various energy production / consumption equipment, energy conversion equipment, and coal gas tanks will make corresponding adjustments, and the adjustment penalty coefficient of each equipment corresponds to the priority of the equipment adjustment sequence, that is, under the same conditions, the model will preferentially adjust the equipment with a smaller penalty coefficient, and the deviation penalty coefficient can ensure that the tank position of the coal gas tank is kept around the target tank position, which is one of the important indicators of the rationality of the optimized scheduling instruction output;

[0167] Finally, based on the energy dynamic prediction and multi-energy flow coupling optimization model, through the development and integration of model database, data acquisition software, model service software, client software, etc., an energy medium production and consumption prediction and intelligent scheduling system is formed, and finally the functions of automatic data acquisition, automatic acquisition of working condition related signals, Gantt chart working condition management, autonomous operation of scheduling model, and pushing of optimized scheduling instructions are realized.

[0168] The present application also provides an energy dynamic prediction and multi-energy flow coupling optimization energy management and control system, which integrates core modules from basic energy management to energy production and consumption prediction, multi-energy medium coupling optimization distribution, and energy scheduling decision pushing, and finally realizes functions such as automatic data acquisition, automatic acquisition of working condition related signals, Gantt chart working condition management, autonomous operation of scheduling model, and pushing of optimized scheduling instructions, and provides a visual interface to facilitate dispatchers to more intuitively manage and allocate energy;

[0169] Specific function contents include:

[0170] 1. Data acquisition and cleaning module:

[0171] Data acquisition steps:

[0172] Using advanced sensor technology and loT technology, the SCADA system acquires real-time energy production and consumption data of various equipment, converts data of different sources and formats into a unified data type through the use of ETL tools, ensures the consistency and compatibility of the data, and performs preliminary processing and filtering to reduce the transmission of invalid data and improve data processing efficiency.

[0173] Data cleaning steps:

[0174] (1) Missing value processing: the system automatically detects missing values in the data set, and for completely random missing data (MCAR) and non-critical data, adopts a deletion method according to the unified time granularity of the model to save data processing time; when processing random missing data (MAR) and non-random missing data (MNAR), the random forest method is used to fill in the missing data, so as to fully restore the true situation of the original data.

[0175] (2) Abnormal data identification and correction: Use KMeans clustering algorithm to cluster data, count the number of data points in each cluster, set a threshold, if the cluster number is less than the threshold, consider the data point as an outlier, and correct it by using the average value interpolation method.

[0176] (3) Data deduplication: Through similarity matching technology, it can efficiently identify and remove duplicate data, ensuring the uniqueness and accuracy of the data set.

[0177] (4) Data standardization: The system standardizes the data, including unit conversion, dimension unification, etc., to ensure the consistency and comparability of the data.

[0178] 2. Energy management and diagnosis module:

[0179] (1) Energy management: Establish a data dictionary to record the type, unit, and other information of the data, provide data query and use functions; record the source, collection time, collection method, and other metadata information of the data, ensure the transparency and traceability of the data, through the metadata information, trace the source and collection process of the data, when the data is abnormal, can trace to the specific sensor equipment and collection time, find out the problem reason, solve the data abnormal problem. Use MySQL's distributed architecture to realize the above storage functions, and use MySQL's built-in master-slave replication mechanism and asynchronous replication mode to meet the enterprise's demand for large-scale data management and high-performance database system.

[0180] (2) Diagnosis function: Statistical analysis of the types and consumption of energy used by the enterprise, find out the main energy consumption points and energy saving potential. Evaluate the efficiency of the main production equipment of the enterprise, find out the problems of low equipment efficiency and energy waste.

[0181] 3. Gantt chart working condition management module:

[0182] Through the Gantt chart working condition management module, the working condition samples are entered into the system, and the entered working condition sample information will be displayed through the movable Gantt chart. The unbalance quantity is introduced by superimposing the working condition samples to simulate the change quantity of the equipment production and consumption under the actual production plan conditions. The specific functions are as follows:

[0183] (1) Manually enter and adjust the working condition, users can intuitively change the start and end time of the plan by dragging the Gantt bar, and manually modify the working condition Gantt bar corresponding to the working condition sample;

[0184] (2) The working condition combination of the production equipment can be extracted from the production scheduling plan, or directly selected from the device working condition type definition function to combine the device working condition.

[0185] (3) The working condition can be automatically tracked in the background and refreshed regularly to reflect the latest changes in the working condition;

[0186] Energy online prediction and optimal scheduling module:

[0187] (1) The module displays the dynamic prediction curve of coal gas / steam production and consumption through a visual chart, and the prediction curve includes various types of coal gas and steam production and consumption curves of each energy equipment, various types of coal gas and steam production and consumption curves of the whole plant, coal gas / steam pipe network pressure prediction curve, blast furnace / coke oven / converter gas tank level prediction curve, etc.

[0188] (2) Based on the energy dynamic prediction curve, the user can call the multi-energy coupling optimal scheduling model to give an energy optimal allocation scheme, and the visual chart displays the equipment energy production and consumption, energy total quantity change trend and tank level change trend in the future period after the implementation scheme.

[0189] 4. Energy scheduling decision pushing module:

[0190] When the energy online prediction and optimal scheduling module generates a new scheduling decision, the scheduling decision and optimization suggestion are displayed in the form of charts and texts through the system interface. For the scheduling decision confirmed by the user, the system can automatically perform the corresponding adjustment operation, such as adjusting the equipment load, starting and stopping the machine operation, etc. Save all pushed scheduling decisions and their execution situations to form a historical record, and analyze the execution effect of the historical decision to evaluate the accuracy and effectiveness of the optimization suggestion, so as to improve the future scheduling model and algorithm.

[0191] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solution deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An energy dynamic prediction and multi-energy flow coupling optimization method, characterized in that, Comprise the following steps: S1, obtain energy production and consumption history data, working condition type list, production and maintenance plan of future scheduling period; S2, pre-process the historical data to obtain pre-processed historical data; use a template matching method to extract periodic features from the pre-processed historical data; Based on the periodic feature, obtain the strongly correlated feature points; construct a long short-term memory neural network short prediction model fused with self-attention mechanism and train it; use grey wolf optimization algorithm to optimize the trained long short-term memory neural network short prediction model fused with self-attention mechanism; use the optimized long short-term memory neural network fused with self-attention mechanism to output the prediction results; S3, based on the prediction results of S2, establish a working condition sample library and a Gantt chart working condition management module; define the device working condition type in the Gantt chart working condition management module according to the energy equipment and its working condition type list, define the working condition sample curve based on the device working condition type, and use the Gantt chart working condition management module to realize the automatic calling of the working condition sample curve or the working condition instance with the working condition rhythm change; S4, collect real-time data of energy production and consumption of different devices in the SCADA system, collect abnormal working condition information, superimpose the real-time data, abnormal working condition information and working condition sample curve to obtain the future trend curve of the total production and consumption of coal gas and steam under multi-working condition conditions; find out the historical data sequence with the same working condition as the prediction day as the correction information of the prediction curve, correct the production and consumption prediction curve and the future trend curve according to the correction information, obtain the different device energy production and consumption prediction curve, coal gas tank position prediction curve and coal gas and steam total production and consumption prediction curve, and finally realize dynamic prediction; S5, determine the scheduling model variables and model constraint conditions, determine the parameter value method of the cost term and the penalty term in the scheduling function, obtain the efficiency characteristic equation of the device in a data-driven manner combined with the historical production and consumption data of the device, optimize the objective function comprehensively, establish a multi-energy coupling optimization scheduling model, and based on the energy dynamic prediction curve in S4, know the energy distribution of each device and the total energy change trend in a future period of time, i.e. the production and consumption of each device in the future, call the multi-energy coupling optimization scheduling model to give an energy optimization distribution scheme based on the production and consumption of each device in the future.

2. The method of claim 1, wherein, S1 specifically comprises: Obtain the enterprise energy topology network structure, energy production and consumption history data through the enterprise comprehensive data integration platform server, obtain the types and quantities of energy media generated and consumed under different working conditions and their change rules over time through the enterprise comprehensive data integration platform server, and obtain the production plan and maintenance plan in the future scheduling period through the enterprise comprehensive data integration platform server; The enterprise energy topology network structure and energy production and consumption history data include boiler coal gas consumption and steam generation, generator set steam inlet and steam extraction and power generation, coal gas diffuser tower diffuser quantity, and coal gas tank real-time tank position. 3.The energy dynamic prediction and multi-energy flow coupling optimization method according to claim 1, characterized in that, S2 specifically comprises the following steps: S201, data preprocessing is performed on historical data of influences of different types of working conditions on energy production and consumption of equipment, to obtain preprocessed historical data; the preprocessing method includes time series stationarization, data repair, serialized data, normalization and data segmentation; S202, periodic characteristics in the preprocessed historical data are determined, a periodic matching template is constructed, the periodic matching template is matched with the preprocessed historical data, and specific periodic characteristics in the preprocessed historical data are extracted; The matching degree of the template and the data is measured by a Pearson correlation coefficient, to obtain strongly correlated feature points; key information of the strongly correlated feature points is extracted for time series prediction; The calculation formula of the Pearson correlation coefficient is as follows: In the formula, r xy is a correlation coefficient, X is an independent variable, and Y is a dependent variable. S203, a long short-term memory neural network short prediction model fused with a self-attention mechanism is constructed, to fully extract the dependency between time series; The self-attention mechanism algorithm process is as follows: In the formula, Tr, Tm and Tk are three h-length vectors in parallel combination of w sequences in the sample, the Softmax function is an important activation function in the field of deep learning, and the Softmax function converts original scores output by the neural network into probability distribution of each category, so that the model can classify different categories in the form of probability; S204, a grey wolf optimization algorithm is used to optimize parameters of the long short-term memory neural network short prediction model fused with the self-attention mechanism, to train the long short-term memory neural network short prediction model fused with the self-attention mechanism with optimal prediction ability, and output a prediction result of influences of different working conditions on energy production and consumption of equipment; The specific formula of the grey wolf optimization algorithm is as follows: In the formula, y represents the predicted value of the production and consumption, y i y represents the true value of the production and consumption. 4.The method of claim 1, wherein, S3 specifically includes the following steps: S301, defining working condition types according to equipment types, and recording each working condition type in a Gantt chart management module; Defining working condition types for main coal gas production and consumption units, including blast furnace replacement, blast furnace blowdown, coke oven maintenance, converter heat, converter maintenance and rolling steel maintenance, for selection when recording working conditions; S302, defining production and consumption curves of various media for various working condition types of energy equipment, and using the production and consumption curves as working condition sample curves of changes of types and quantities of media of the equipment with time; S303, one-to-one correspondence between continuous periods of various working conditions of each equipment and working condition sample curves or working condition instances of the working conditions, to realize automatic calling of the working condition sample curves or the working condition instances with changes of working condition rhythm; When an abnormal working condition occurs, a certain production and consumption medium sample curve, i.e., working condition instance data, is directly recorded with the working condition, and the working condition instance data has higher priority than sample data.

5. The method of claim 1, wherein, S4 specifically includes the following steps: S401, collecting real-time data in a SCADA system and preprocessing the real-time data; S402, selecting energy equipment and working condition types thereof that need to be subjected to automatic tracking of working conditions, determining working condition signal collection sites in a steel plant, establishing a working condition rhythm pre-judgment logic based on production / maintenance plans and real-time working condition signals, and automatically tracking and correcting working condition rhythm; S403, introducing abnormal working conditions in a future time period in the production / maintenance plan through dragging a Gantt bar; and an unbalance formula caused by superposition of abnormal working conditions in a certain time period is as follows: In the formula, i, m, z represent the ith time point, the mth energy medium, and the zth working condition, respectively, (t i ,t i+1 ) is a certain time period in h-1 time periods. S404, Deep mining of energy historical data in the system uses K-neighbor algorithm to find historical data with the same working condition as the prediction day to correct the prediction curve, and corrects the production and consumption prediction curve and future trend curve according to the correction information, and finally realizes dynamic prediction. 6.The method of claim 1, wherein, S5 includes the following steps: S501, clarify the enterprise energy dispatching rules and safety supply requirements, including the safe pressure of the coal gas / steam pipe network, the safe position of the coal gas tank, the safe pressure of the steam accumulator, the safe load of the hot blast furnace blast for production, the safe load of the heating furnace for production, the minimum load of the generator set, and the safety range of the demand for purchased electricity; S502, compile a list of adjustable energy equipment, including coal gas / steam generator sets and coal gas tanks, and set adjustment priorities, establish a performance database of adjustable energy equipment, and track the real-time operation state of adjustable equipment; S503, according to the corresponding relationship between the energy conversion efficiency of each boiler in a certain period and the boiler operating load, the operating load of the energy conversion equipment and the energy conversion efficiency under the corresponding load are characterized as a quadratic function by a quadratic curve fitting method, and according to the test data and online production data of each steam turbine, a consumption characteristic equation of the steam turbine is established by using the principle of multiple linear regression, and the formulas are as follows: wherein represents the actual load of the energy conversion plant when it is operating stably, C1, C2, C3 represent fitting parameters; respectively represent the steam admission, extraction, and condensing amount of the steam turbine in the t time period; S504, according to the list of adjustable equipment, combined with the energy dispatching rules and safety supply requirements, determine the dispatching model variables and model constraints, and establish a multi-energy coupling optimization dispatching model. S505, based on the energy dynamic prediction curve in S4, the energy distribution of each device in a certain period of time and the total energy change trend are known, and the multi-energy coupling optimization dispatching model is called to give an energy optimization distribution scheme.

7. The method of claim 6, wherein, The multi-energy coupling optimization dispatching model includes an objective function and a constraint condition, and the dispatching cycle of the multi-energy coupling optimization dispatching model is T, a certain time period is t (1≤t≤T), D is the number of different devices, and N represents the number of different energy media. The objective function formula of the model is as follows: In the formula, C buy,j , E j,t respectively represent the ex-factory price of fuel j and the ex-factory quantity of fuel j in period t; C os,t , P t respectively represent the electricity price in peak, valley and flat periods and the purchased electricity quantity in peak, valley and flat periods; C adj,j , ΔF i,j,t respectively represent the adjustment penalty coefficient of fuel j and the adjustment quantity of device i to fuel j in period t; C re,k , D k,t respectively represent the diffusion penalty coefficient of coal gas k and the diffusion quantity of coal gas k in period t; C st,i , L i,t respectively represent the shutdown cost of device i and the shutdown flag of device i, 0 represents shutdown and 1 represents start; C de,k , H r,k,t , H g,k respectively represent the rack position deviation penalty coefficient, the rack position height of coal gas k at time t and the target rack position height; The constraint conditions include device range constraints, device node balance constraints, supply and demand balance constraints, and calorific value balance constraints; The device range constraints include device energy production and consumption upper and lower limit constraints, coal gas tank position upper and lower limit constraints, and device climbing speed upper and lower limit constraints; The device energy production and consumption upper and lower limit constraints are as follows: G i,j,min ≤G i,j,t ≤G i,j,max The coal gas tank position upper and lower limit constraints are as follows: H k,min ≤H k,t ≤H k,max The device climbing speed upper and lower limit constraints are as follows: M i,speed,min ≤M i,speed,t ≤M i,speed,max where G i,j,t represents the amount of energy produced or consumed by device i, H k,t represents the height of the coal gas k, M i,speed,t represents the climbing speed of device i; The device node balance constraint is: where n ef,i,j represents the conversion factor of energy j in device i to energy p, m represents the number of energy types consumed by device i, E in,i,j represents the amount of energy j consumed by device i, h j represents the heat value of energy j, F out,i,p represents the amount of energy p produced by device i, h p represents the heat value of energy p; The supply and demand balance constraint is: where (ΔT i , ΔT i+1 ) is a period in ΔT, W j,Δt represents the amount of imbalance introduced by the working condition within ΔT, represents the amount of energy j produced in the period (ΔT i , ΔT i+1 ), represents the amount of change in the energy j storage in the period (ΔT i , ΔT i+1 ), which is 0 if energy j is not gas, respectively represent the amount of energy j dissipated and consumed in the period (ΔT i , ΔT i+1 ). The calorific value balance constraint is: wherein X i , Y i respectively represent the minimum and maximum specific heat values of the energy mix total heat value consumed by the device i; Under different production conditions, various energy production / consumption devices, energy conversion devices, and coal gas tanks will make corresponding adjustments, and each device adjustment penalty coefficient is equivalent to providing priority for device adjustment order. Under the condition that other conditions are the same, the model will preferentially adjust the device with a small penalty coefficient, and the deviation penalty coefficient can ensure that the coal gas tank position is maintained around the target position.

8. An energy management system for energy dynamic prediction and multi-energy flow coupling optimization, used to implement the energy dynamic prediction and multi-energy flow coupling optimization method of any one of claims 1-7, characterized in that: It includes a data collection and cleaning module, an energy management and diagnosis module, a Gantt chart working condition management module, an energy online prediction and optimization dispatching module, and an energy dispatching decision pushing module; The data collection and cleaning module is used to perform data collection and data cleaning steps: The data collection step is as follows: By using advanced sensor technology and Internet of Things technology, the SCADA system collects real-time energy production and consumption data of various devices, converts data of different sources and formats into a unified data type through the use of ETL tools, ensures data consistency and compatibility, and performs preliminary processing and filtering to reduce the transmission of invalid data and improve data processing efficiency; The data cleaning step is as follows: T1, missing value processing: the system automatically detects missing values in the data set, and for completely random missing data and non-critical data, it uses the method of deleting according to the unified time granularity of the model to process, thereby saving data processing time; When dealing with random missing data and non-random missing data (MNAR), the random forest method is used to fill in the missing data, thereby fully restoring the true situation of the original data; T2, abnormal data identification and correction: use KMeans clustering algorithm to cluster the data, count the number of data points in each cluster, set a threshold, if the cluster number is less than the threshold, consider that the data point is an outlier, and use the average value interpolation method for correction; T3, data deduplication: through similarity matching technology, identify and remove duplicate data to ensure the uniqueness and accuracy of the data set; T4, data standardization: the system standardizes the data, including unit conversion, dimension unification, etc., to ensure data consistency and comparability; The energy management and diagnosis module is used to perform energy management and diagnosis functions; The energy management is as follows: establish a data dictionary to record data types, units, and other information, provide data query and use functions; record data source, collection time, collection method, and other metadata information to ensure data transparency and traceability, trace the data source and collection process through metadata information, when data is abnormal, trace to specific sensor equipment and collection time through metadata to find out the problem cause and solve the data abnormality problem, use the distributed architecture of MySQL to realize storage function, and use the master-slave replication mechanism and asynchronous replication mode provided by MySQL; The diagnosis function is as follows: statistical analysis of the types and consumption of energy used by the enterprise, finding out the main energy consumption points and energy saving potential, efficiency evaluation of the main production equipment of the enterprise, finding out the problems of low equipment efficiency and energy waste; The Gantt chart working condition management module functions as follows: through the Gantt chart working condition management module, the working condition samples are recorded into the system, the recorded working condition sample information is displayed through a movable Gantt chart, and the imbalance quantity is introduced by superimposing the working condition samples to simulate the change quantity of device production and consumption under actual production plan conditions; The energy online prediction and optimal scheduling module displays the dynamic prediction curve of coal gas / steam production and consumption through a visual chart, the production and consumption prediction curve including various types of coal gas and steam production and consumption curves of each energy equipment, various types of coal gas and steam production and consumption curves of the whole plant, a coal gas / steam pipe network pressure prediction curve, and a blast furnace / coke oven / convertor gas tank level prediction curve; based on the energy dynamic prediction curve, the user calls the multi-energy coupling optimal scheduling model to give an energy optimal distribution scheme, at this time the visual chart displays the equipment energy production and consumption, the energy total quantity change trend, and the tank level change trend in the future period after the implementation scheme is displayed; The energy scheduling decision pushing module, when the energy online prediction and optimal scheduling module generates a new scheduling decision, displays the scheduling decision and the optimal suggestion in the form of charts and texts through a system interface, for the scheduling decision confirmed by the user, the system automatically performs the corresponding adjustment operation; saves all the pushed scheduling decisions and the execution situations, forms a historical record, analyzes the execution effect of the historical decision, evaluates the accuracy and effectiveness of the optimal suggestion, so as to improve the future scheduling model and algorithm.

9. A storage medium, characterized by The storage medium includes a stored program, wherein the program runs to execute the energy dynamic prediction and multi-energy flow coupling optimization method in any one of claims 1 to 7.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the energy dynamic prediction and multi-energy flow coupling optimization method in any one of claims 1 to 7 by running the computer program.

Images