MULTI-SCALE COORDINATED CONTROL METHOD OF INTEGRATED ENERGY SYSTEM FOR GREEN HYDROGEN METALLURGY

Abstract

The present application discloses a multi-scale coordinated control method of an integrated energy system for green hydrogen metallurgy, and belongs to the field of energy system control technology. A detailed analysis is performed on various flexible and adjustable resources in the integrated energy system for green hydrogen metallurgy, and the regulation characteristics of these resources at different timescales are summarized. The cross-link regulation characteristics of these resources in multi-energy and mass interactions are then explored, with particular attention paid to the time-delay phenomenon in the transmission and conversion process of multi-energy and mass. An equivalent modeling method for heterogeneous links is proposed to provide support for the multi-link coordinated control of the system. Finally, the influence of different production tasks on the multi-energy and mass flow allocation is studied, and the dynamic association of different production processes on the multi-energy and mass flow allocation is analyzed.

Claims

1. A multi-scale coordinated control method of an integrated energy system for green hydrogen metallurgy, comprising the following steps: S1: analyzing multi-timescale regulation characteristics of flexible and adjustable resources in the integrated energy system for green hydrogen metallurgy to form a regulation response timescale matrix; S2: analyzing cross-link regulation characteristics in a multi-energy and mass interaction according to the regulation response timescale matrix, and further performing equivalent modeling based on heterogeneous links to obtain a regulation external-characteristic equivalent model; S3: analyzing influence of different production tasks in the integrated energy system for green hydrogen metallurgy on allocation of multi-energy and mass flows based on dynamic association analysis, and establishing an extended resource-task network model representing an association between the production tasks and the multi-energy and mass flows; S4: predicting random variables involved in a production process at a future moment based on a task-resource allocation result obtained by the extended resource-task network model, inputting the random variables into a multi-model prediction mechanism based on a local model combination and a multi-controller combination, and outputting an energy-mass flow control strategy at a current working condition to achieve the coordinated control of the integrated energy system for green hydrogen metallurgy; wherein in the step S2, the method of performing the equivalent modeling based on the heterogeneous links to obtain the regulation external-characteristic equivalent model specifically comprises: S21: collecting historical operation data of various devices in the integrated energy system for green hydrogen metallurgy; wherein the historical operation data is regulation characteristic data of the various devices under different working conditions; S22: selecting a corresponding deep neural network according to different regulation characteristic data of the various devices; S23: training the corresponding deep neural network according to the historical operation data and dynamically updating to obtain regulation external-characteristic models corresponding to different regulation characteristics of the various devices; the step S3 comprises the following substeps: S31: performing dynamic association analysis on production task types and the multi-energy and mass flow allocation in the integrated energy system for green hydrogen metallurgy, and constructing a dynamic association model for production tasks; S32: performing dynamic association analysis on production task processes and the multi-energy and mass flow allocation in the integrated energy system for green hydrogen metallurgy, and constructing an inter-process energy-mass flow coordinated control model; and S33: constructing an extended resource-task network model according to the constructed dynamic association model for production tasks and the inter-process energy-mass flow coordinated control model and by introducing dynamic characteristics of energy-mass flows in time and space; the step S31 comprises the following substeps: S31-1: classifying the production tasks in the integrated energy system for green hydrogen metallurgy according to production demands, and determining demand characteristics of various production tasks on energy-mass flows at different time periods; S31-2: fitting historical data corresponding to the demand characteristics of various production tasks on the energy-mass flows at different time periods, and establishing dynamic demand curves of the various production tasks; S31-3: according to the dynamic demand curves of the various production tasks, dynamically generating a plurality of association models of different production task types to obtain a dynamic association model for production task; wherein the production task types comprise a continuous type production task and a discrete type production task; the step S32 comprises the following substeps: S32-1: analyzing energy-mass demand characteristics of each production process of the production tasks in the integrated energy system for green hydrogen metallurgy; S32-2: adjusting an execution sequence of the production task processes and energy-mass flow allocation in real time through a dynamic planning or optimization algorithm according to the energy-mass demand characteristics corresponding to different production processes, so that the coordinated scheduling among the production processes is optimal, and further obtaining the inter-process energy-mass flow coordinated control model; wherein the inter-process energy-mass flow coordinated control model takes time scheduling cost among the production processes, dependency relationships among the production processes and resource limitation as constraint conditions; in the step S33, the extended resource-task network model is constructed as a model for dynamically adjusting resource supply and task demand in the integrated energy system for green hydrogen metallurgy according to time series data; when the extended resource-task network model is configured to perform coordinated control on resources related to a plurality of production tasks, multi-association processing is performed on the resources, an optimization model with constraint conditions is established, and multi-energy and mass supply coordinated matching with production task demands are achieved; wherein the constraint conditions comprise a resource supply capacity, dynamic allocation of energy-mass flows and time series dependency of production tasks; when the extended resource-task network model is configured to allocate resources and tasks, a dynamic optimization algorithm is adopted to dynamically adjust an execution sequence of the production tasks and an allocation strategy of the energy-mass flows according to current energy-mass flow allocation in the integrated energy system for green hydrogen metallurgy; the extended resource-task network model is defined as: $E_{\mathrm{elec}}\left(t,x\right)=h\left(t,x\right);$ $E_{H_2}\left(t,x\right)=k\left(t,x\right);$ wherein t is time, x is spatial position, and functions h and k describe dynamic allocation of different energy-mass flows in time and space dimensions.

2. The multi-scale coordinated control method of the integrated energy system for green hydrogen metallurgy according to claim 1, wherein in the step S1, the method of analyzing the multi-timescale regulation characteristics of the flexible and adjustable resources comprises: analyzing regulation response characteristics of energy supply, demand, energy conversion and storage links in the integrated energy system for green hydrogen metallurgy on different timescales respectively, wherein the regulation response characteristics comprise energy supply regulation characteristics, demand side regulation characteristics, energy conversion link regulation characteristics, and storage link regulation characteristics.

3. The multi-scale coordinated control method of the integrated energy system for green hydrogen metallurgy according to claim 1, wherein in the step S2, the method for analyzing the cross-link regulation characteristics in the multi-energy and mass interaction specifically comprises: analyzing influence of mutual conversion and interaction between different energy-mass flows on overall scheduling of the integrated energy system for green hydrogen metallurgy by constructing an energy-mass flow coupling model and analyzing a time-delay effect in a cross-loop energy-mass interaction process; wherein the time-delay effect comprises cross-link energy-mass transmission delay and conversion delay.

4. The multi-scale coordinated control method of the integrated energy system for green hydrogen metallurgy according to claim 1, wherein in the step S4, the multi-model prediction mechanism based on the local model combination refers to: constructing a local model for achieving energy-mass flow allocation prediction under different working conditions, and switching to an optimal local model according to the operating state of the integrated energy system for green hydrogen metallurgy to perform corresponding energy-mass flow allocation prediction based on a conditional model switching mechanism; the multi-model prediction mechanism based on the multi-controller combination refers to: on the basis of an output energy-mass flow allocation strategy predicted by a multi-model prediction mechanism of local model combination, designing corresponding independent controllers according to different energy-mass flow characteristics, and introducing a global objective function to optimize a control target of each independent controller by adopting a constraint optimization-based coordinated mechanism according to different working conditions and energy-mass flow control requirements, so as to obtain an overall optimal energy-mass flow control strategy for the system; wherein an output of each independent controller is associated.

5. The multi-scale coordinated control method of the integrated energy system for green hydrogen metallurgy according to claim 4, wherein the step S4 comprises the following substeps: S41: performing time series characteristic analysis on the random variables involved in the production process based on task-resource allocation result obtained by the extended resource-task network model; S42: constructing a discrete-time prediction model according to a time series characteristic analysis result of the random variables, and predicting the random variables involved in the production process at the future moment by using the discrete-time prediction model; S43: inputting the predicted random variables into a multi-model prediction mechanism based on a local model, predicting energy-mass flow allocation strategies in different operating states, and inputting the energy-mass-flow allocation strategies into the multi-model prediction mechanism based on the multi-controller combination; and S44: in the multi-model prediction mechanism based on the multi-controller combination, on the basis of ensuring scheduling consistency among independent controllers, selecting a corresponding independent controller for energy-mass flow control according to the involved energy-mass flow characteristics, and outputting the energy-mass flow control strategy at the current working condition to achieve the coordinated control of the integrated energy system for green hydrogen metallurgy.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0053] FIG. 1 is a flow chart of a multi-scale coordinated control method of an integrated energy system for green hydrogen metallurgy according to the present application.

[0054] FIG. 2 is a principle framework diagram of constructing a regulation external-characteristic equivalent model according to the present application.

[0055] FIG. 3 is a principle framework diagram of constructing an extended resource-task network model according to the present application.

[0056] FIG. 4 is a principle framework diagram of energy-mass flow control by a multi-model prediction mechanism according to the present application.

DESCRIPTION OF EMBODIMENTS

[0057] The following description of the specific embodiments of the present application is provided to facilitate the understanding of the present application by those skilled in the art, however, it should be understood that the present application is not limited to the scope of the specific embodiments, and for those of ordinary skill in the art, various changes that are made without departing from the spirit and scope of the present application as defined and determined by the appended claims are apparent, and all inventions and creations that are made by using the concept of the present application are within the protective scope.

[0058] An embodiment of the present application provides a multi-scale coordinated control method of an integrated energy system for green hydrogen metallurgy, as shown in FIG. 1, which includes the following steps: [0059] S1: analyzing multi-timescale regulation characteristics of flexible and adjustable resources in the integrated energy system for green hydrogen metallurgy to form a regulation response timescale matrix; [0060] S2: analyzing cross-link regulation characteristics in a multi-energy and mass interaction according to the regulation response timescale matrix, and further performing equivalent modeling based on heterogeneous links to obtain a regulation external-characteristic equivalent model; [0061] S3: analyzing influence of different production tasks in the integrated energy system for green hydrogen metallurgy on allocation of multi-energy and mass flows based on dynamic association analysis, and establishing an extended resource-task network model representing an association between the production tasks and the multi-energy and mass flows; and [0062] S4: predicting random variables involved in a production process at a future moment based on a task-resource allocation result obtained by the extended resource-task network model, inputting the random variables into a multi-model prediction mechanism based on a local model combination and a multi-controller combination, and outputting an energy-mass flow control strategy at a current working condition to achieve the coordinated control of the integrated energy system for green hydrogen metallurgy.

[0063] In the step S1 of the embodiment of the present application, to implement efficient coordinated operation of the system, firstly, systematic analysis and modeling of the regulation characteristics of different devices in the integrated energy system for green hydrogen metallurgy are required. Based on this, in this embodiment, the method of analyzing the multi-timescale regulation characteristics of the flexible and adjustable resources includes: [0064] analyzing regulation response characteristics of energy supply, demand, energy conversion and storage links in the integrated energy system for green hydrogen metallurgy on different timescales respectively, wherein the regulation response characteristics include energy supply regulation characteristics, demand side regulation characteristics, energy conversion link regulation characteristics, and storage link regulation characteristics.

[0065] In this embodiment, the flexible and adjustable resources include various continuous, discontinuous, and interruptible regulations devices in the integrated energy system for green hydrogen metallurgy. For example, the flexible and adjustable resources include power grid access, renewable energy power generation, electrolyzers, hydrogen storage devices, shaft furnaces, electric arc furnaces, and the like.

[0066] In an example of this embodiment, the following example is provided for analyzing the energy supply regulation characteristics.

[0067] Grid access and renewable energy power generation (such as wind energy and photovoltaic) are energy supply resources with volatility on a time scale, so that the volatility needs to be fitted by a probability model, and the characteristic description of energy supply fluctuations is constructed through a time-series model from an hour level to a minute level.

[0068] The volatility of energy supply can be described using a stochastic process model. Assuming that the time series of wind and photovoltaic power generation are P.sub.wind(t) and P.sub.pv(t) respectively, a total energy supply power P can be expressed as:

[00001]$\begin{array}{cc}{P}_{\mathrm{supply}}\left(t\right)=P_{\mathrm{grid}}\left(t\right)+P_{\mathrm{wind}}\left(t\right)+P_{\mathrm{pv}}\left(t\right)& \left(1\right)\end{array}$ [0069] wherein P.sub.grid(t) is a power supplied by the grid.

[0070] The volatility of wind and photovoltaic power generation can be described by probability distributions (P.sub.wind) and (P.sub.pv), which are usually represented by normal distribution or other distribution functions:

[00002]$\begin{array}{cc}f\left(P_{\mathrm{wind}}\right)N\left({}_{\mathrm{wind}},{}_{\mathrm{wind}}^2\right)& \left(2\right)\end{array}$$\begin{array}{cc}f\left(P_{\mathrm{pv}}\right)N\left({}_{\mathrm{pv}},{}_{\mathrm{pv}}^2\right)& \left(3\right)\end{array}$

[0071] In an example of this embodiment, the following example is provided for analyzing the demand side regulation characteristics.

[0072] There are significant differences in the operating characteristics of demand side devices (such as electrolyzers, shaft furnaces, and electric arc furnaces). For example, the electrolyzers can operate continuously and are suitable for frequent regulations, while shaft furnaces and electric arc furnaces have intermittent working modes and large thermal inertia, so the regulation frequency is relatively low, and special attention needs to be paid to the changes in energy consumption during startup and shutdown. The regulation frequency, response time and energy efficiency indicators of these devices are modeled in detail to generate the operating curves of the devices at different timescales.

[0073] A relationship between dynamic response time electrolyzer and power P.sub.elec(t) of the electrolyzer may be described as a first-order dynamic model:

[00003]$\begin{array}{cc}\mathrm{electrolyzer}\frac{{\mathrm{dP}}_{\mathrm{elec}}\left(t\right)}{\mathrm{dt}}+P_{\mathrm{elec}}\left(t\right)=P_{\mathrm{set}}\left(t\right)& \left(4\right)\end{array}$ [0074] wherein P.sub.set(t) is set power. Similarly, the dynamic characteristics of power P.sub.furnace(t) and thermal inertia P.sub.set_furnace(t) of shaft furnaces and electric arc furnaces can be expressed as:

[00004]$\begin{array}{cc}\mathrm{electrolyzer}\frac{{\mathrm{dP}}_{\mathrm{furnace}}\left(t\right)}{\mathrm{dt}}+P_{\mathrm{furnace}}\left(t\right)=P_{\mathrm{set}\_\mathrm{furnace}}\left(t\right)& \left(5\right)\end{array}$

[0075] In an example of this embodiment, the energy conversion link includes an electrolytic hydrogen production process, and the nonlinear conversion efficiency of this link needs to be modeled according to different working conditions and load conditions. Specifically, an example of analyzing the regulation characteristics is provided.

[0076] The power-to-hydrogen efficiency of the electrolyzer varies significantly under different loads. The conversion characteristics can be described by fitting experimental data or mechanism models, and the startup time-lag effect of the electrolyzer should be explicitly considered in the model.

[0077] The power-to-hydrogen efficiency .sub.elec(P) of the electrolyzer can be described by a nonlinear function. Assuming that the relationship between the power-to-hydrogen efficiency and input power P of the electrolyzer is:

[00005]$\begin{array}{cc}{}_{\mathrm{elec}}\left(P\right)=\frac{\mathrm{Hydrogen}\mathrm{production}}{\mathrm{Electricity}\mathrm{input}}=f\left(P\right)& \left(6\right)\end{array}$

[0078] In actual operation, the efficiency may show nonlinear characteristics with power changes, for example:

[00006]$\begin{array}{cc}{}_{\mathrm{elec}}\left(P\right)={\mathrm{aP}}^2+\mathrm{bP}+c& \left(7\right)\end{array}$ [0079] wherein a, b and c are fitting coefficients.

[0080] In an example of this embodiment, for the analysis of the storage link regulation characteristics, taking a hydrogen storage device as an example, the following example is provided.

[0081] The response time and charging and discharging rate of the hydrogen storage device need to be optimized according to different requirements. The charging and discharging behavior of the hydrogen storage device has nonlinear characteristics. For example, the efficiency decays significantly when the device in a full-charging or emptying state, and needs to be described and modeled through function fitting or physical models.

[0082] The charging and discharging efficiency .sub.storage(x) of the hydrogen storage device is usually related to the current storage state x (the amount of stored hydrogen), which can be expressed as:

[00007]$\begin{array}{cc}{}_{\mathrm{storage}}\left(x\right)={}_0.Math.\left(1-.Math.\frac{x}{x_{\max }}\right)& \left(8\right)\end{array}$ [0083] wherein .sub.0 is an initial charge and discharge efficiency, is an efficiency attenuation coefficient, and x.sub.max is a maximum capacity of the storage device.

[0084] In this application, the regulation characteristics of various types of devices in the flexible and adjustable resources are analyzed in detail to form a complete regulation response timescale matrix, which provides a basis for cross-scale regulation of the system.

[0085] In the embodiment of the present application, in the integrated energy system for green hydrogen metallurgy, each link not only has its own regulation characteristics, but also needs to consider the influence of the mutual conversion and interaction between different energy-mass (electricity, hydrogen, or heat) on the overall scheduling. For example, in the process of hydrogen production by electrolysis, the conversion efficiency of electrical energy to hydrogen energy is not only limited by the electrolyzer itself, but also affected by fluctuations in the energy supply of the upstream power grid.

[0086] Based on this, in the step S2 of this embodiment, the method for analyzing the cross-link regulation characteristics in the multi-energy and mass interaction specifically includes: [0087] analyzing influence of mutual conversion and interaction between different energy-mass flows on the overall scheduling of the integrated energy system for green hydrogen metallurgy by constructing an energy-mass flow coupling model and analyzing a time-delay effect in a cross-loop energy-mass interaction process; wherein the time-delay effect includes cross-link energy-mass transmission delay and conversion delay.

[0088] In this embodiment, the interaction between different energy-mass in the system is accurately described through an energy-mass flow coupling model. In an example of this embodiment, taking an electrolyzer and a hydrogen storage device as an example, the following process of constructing an energy-mass flow coupling model is provided.

[0089] The energy-mass flow coupling model of the electrolyzer includes a functional relationship between electricity input and hydrogen output, and considers factors such as power-to-hydrogen efficiency and dynamic changes in temperature and pressure. Similarly, the energy-mass flow coupling between the hydrogen storage device and the shaft furnace is also modeled by through coordinated scheduling optimization.

[0090] In the energy-mass flow coupling of the electrolyzer, the process of converting electrical energy into hydrogen energy can be described by the following relationship:

[00008]$\begin{array}{cc}{Q}_{H_2}\left(t\right)={}_{\mathrm{elec}}\left(P_{\mathrm{elec}}\left(t\right)\right).Math.P_{\mathrm{elec}}\left(t\right)& \left(9\right)\end{array}$ [0091] wherein Q.sub.H.sub.1(t) is a hydrogen production quantity, P.sub.elec(t) is input power of the electrolyzer, and .sub.elec(P.sub.elec(t)) is the power-to-hydrogen efficiency.

[0092] Similarly, the energy flow of the hydrogen storage device can be expressed as:

[00009]$\begin{array}{cc}\frac{{\mathrm{dS}}_{H_2}\left(t\right)}{\mathrm{dt}}=Q_{H_2}\left(t\right)-Q_{\mathrm{consumed}}\left(t\right)& \left(10\right)\end{array}$ [0093] wherein S.sub.H.sub.2(t) is a hydrogen storage capacity, and Q.sub.consumed(t) is a hydrogen consumption quantity.

[0094] In this embodiment, the analysis of time-delay effects during cross-link energy-mass interactions requires quantification through modeling. For example, when hydrogen generated by the electrolyzer flows to the hydrogen storage device or the shaft furnace, there may be pipeline transmission delays, and transmission efficiency is affected by temperature and pressure. These effects can be described by a delay equation and an efficiency decay model to accurately reflect the loss and time-delay in energy transmission.

[0095] The time-delay effect in energy-mass flow transmission can be described using a delay differential equation. For example, if there is the time-delay .sub.delay in the transmission of hydrogen from the electrolyzer to the hydrogen storage device, the equation for the change in hydrogen storage capacity is:

[00010]$\begin{array}{cc}{S}_{H_2}\left(t\right)=S_{H_2}\left(t-{}_{\mathrm{delay}}\right)& \left(11\right)\end{array}$

[0096] The transmission efficiency .sub.trans(t) may be affected by temperature and pressure and can be expressed as:

[00011]$\begin{array}{cc}{}_{\mathrm{trans}}\left(t\right)={}_0.Math.e^{-\left(T\left(t\right)-T_0\right)}& \left(12\right)\end{array}$ [0097] wherein .sub.0 is an initial transmission efficiency, T(t) is a current temperature, is an efficiency attenuation coefficient, and T.sub.0 is a reference temperature.

[0098] In this application, based on the above process, a global dynamic model is constructed by analyzing the dynamic coupling and time-delay effects of energy-mass flows to support subsequent system coordinated scheduling.

[0099] In the step S2 of the embodiment of the present application, after the cross-link regulation characteristics in multi-energy and mass interaction are analyzed, since the nonlinearity and uncertainty of various energy devices in the actual system are difficult to be fully described by traditional physical models, it is necessary to introduce deep learning technology to assist in modeling. The key point of this step is that a high-precision regulation external-characteristic equivalent model is constructed through analysis of historical data.

[0100] Based on this, in the step S2, the method of performing the equivalent modeling based on the heterogeneous links to obtain the regulation external-characteristic equivalent model specifically includes: [0101] S21: Historical operation data of various devices in the integrated energy system for green hydrogen metallurgy are collected.

[0102] The historical operation data is regulation characteristic data of various devices under different working conditions. For example, the historical operation data such as the regulation characteristics of electrolyzers, hydrogen storage devices, and shaft furnaces under different working conditions (such as input power, hydrogen output, energy efficiency, or temperature changes).

[0103] S22: A corresponding deep neural network is selected according to different regulation characteristic data of various devices.

[0104] For example, a deep neural network includes a feedforward neural network (FFNN) or a long short-term memory network (LSTM), with the appropriate network structure selected based on the time response characteristics of the device. Furthermore, the input of the network is a device control variable (such as input power), the output of the network is a regulation result (such as hydrogen production or energy efficiency), and complex regulation behaviors can be accurately fitted through training.

[0105] S23: The corresponding deep neural network is trained according to the historical operation data and dynamically updated to obtain regulation external-characteristic models corresponding to different regulation characteristics of various devices.

[0106] In the step S21 of this embodiment, the collected historical operation data is also standardized, including eliminating dimensional differences and removing abnormal data to improve modeling precision.

[0107] In an example of this embodiment, taking the construction of a regulation external-characteristic model corresponding to an electrolyzer as an example, the following example is provided.

[0108] The input of the network is control variable x=[P.sub.elec(t), T(t), p(t)] (input power, temperature, pressure) of the device, and the output is the hydrogen production quantity Q.sub.H.

[0109] The prediction model of the neural network can be expressed as:

[00012]$\begin{array}{cc}{Q}_{H_2}\left(t\right)=f_{\mathrm{NN}}\left(x\left(t\right).Math.\right)& \left(13\right)\end{array}$ [0110] wherein F.sub.NN is a neural network function, and is the network parameter. The network parameter is optimized by minimizing a loss function L(:

[00013]$\begin{array}{cc}L\left(\right)=\underset{i=1}{\overset{n}{.Math.}}{\left(Q_{H_2}^{\left(i\right)}-f_{\mathrm{NN}}\left(x^{\left(i\right)}.Math.\right)\right)}^2& \left(14\right)\end{array}$

[0111] The online learning update process can update the network weights based on the new observation data x.sub.new and Q.sub.H.sub.2,new through an incremental learning method:

[00014]$\begin{array}{cc}{}_{\mathrm{new}}={}_{\mathrm{old}}-{}_{}{L}_{\mathrm{new}}\left(\right)& \left(15\right)\end{array}$

[0112] is a learning rate, and L.sub.new is a loss function based on new data.

[0113] In this application, the control precision and the response capability of the green hydrogen metallurgy system can be further improved by introducing the deep learning model.

[0114] As shown in FIG. 2, a principle framework diagram from analyzing multi-timescale regulation characteristics to constructing the regulation external-characteristic equivalent model is provided to support the multi-link coordinated control of the integrated energy system for green hydrogen metallurgy.

[0115] As shown in FIG. 3, the step S3 of the embodiment of the present application includes the following substeps: [0116] S31: performing dynamic association analysis on production task types and the multi-energy and mass flow allocation in the integrated energy system for green hydrogen metallurgy, and constructing a dynamic association model for production tasks; [0117] S32: performing dynamic association analysis on production task processes and the multi-energy and mass flow allocation in the integrated energy system for green hydrogen metallurgy, and constructing an inter-process energy-mass flow coordinated control model; and [0118] S33: constructing an extended resource-task network model according to the constructed dynamic association model for production tasks and the inter-process energy-mass flow coordinated control model and by introducing dynamic characteristics of energy-mass flows in time and space.

[0119] In the step S31 of this embodiment, according to the integrated energy system for green hydrogen metallurgy, each production task (such as hydrogen preparation, or storage) is accompanied by the dynamic change of the multi-energy and mass flows. Since the temporal and spatial requirements of each task are different, it is necessary to establish a dynamic association model between the production task and the multi-energy mass and flows to accurately describe this change.

[0120] Based on this, as shown in FIG. 3, the step S31 in this embodiment includes the following substeps:

[0121] S31-1: classifying the production tasks in the integrated energy system for green hydrogen metallurgy according to production demands, and determining demand characteristics of various production tasks on energy-mass flows at different time periods;

[0122] for example, in the integrated energy system for green hydrogen metallurgy, tasks such as hydrogen production, electricity storage, hydrogen storage, and shaft furnace steelmaking have different energy-mass flow demand types and intensities; by classifying the tasks, the energy-mass flow demand characteristics of various tasks across different time periods can be clarified.

[0123] S31-2: fitting historical data corresponding to the demand characteristics of various production tasks on the energy-mass flows at different time periods, and establishing dynamic demand curves of the various production tasks;

[0124] S31-3: according to the dynamic demand curves of the various production tasks, dynamically generating a plurality of association models of different production task types to obtain a dynamic association model for production task; wherein the production task types include a continuous type production task and a discrete type production task.

[0125] In an example of this embodiment, in the step S31-1, assuming that the demand of a task T.sub.i for an energy-mass flow (such as electricity, or hydrogen) can be represented by time t and task progress p.sub.i(t), the energy-mass flow demand function can be defined as:

[00015]$\begin{array}{cc}{E}_i\left(t\right)=f\left(p_i\left(t\right),t\right)& \left(16\right)\end{array}$ [0126] wherein E.sub.i(t) represents the energy-mass flow demand of the task T.sub.i at t. A function (p.sub.i(t),t) is fitted based on historical data to describe the dynamic changes of task demands over time.

[0127] In an example of this embodiment, in the steps S31-2 and S31-3, the allocation of energy-mass flows can be regarded as a function of time and task progress; by fitting historical data, a dynamic demand curve for the task is established, and a plurality of association models are generated for different task types to ensure that the allocation of energy-mass flows meets real-time production requirements. For example, for the entire system, the sum of the energy-mass flow demands in the production task set {T.sub.1, T.sub.2, . . . , T.sub.n} is:

[00016]$\begin{array}{cc}{E}_{\mathrm{total}}\left(t\right)=\underset{i=10}{\overset{n}{.Math.}}{E}_i\left(t\right)& \left(17\right)\end{array}$

[0128] The above describes the total demands of all tasks for an energy-mass flow at any time t.

[0129] In the step S32 of this embodiment, the energy-mass flow interaction between the production processes is complex, and the execution order and the time series of the production processes greatly influence the dynamic allocation of the energy-mass flows. To achieve reasonable control of the energy-mass flows, the execution condition of each process and the energy-mass flow demands need to be modeled.

[0130] Based on this, the step S32 includes the following substeps: [0131] S32-1: analyzing energy-mass demand characteristics of each production process of the production tasks in the integrated energy system for green hydrogen metallurgy; [0132] for example, the hydrogen and electricity requirements of a shaft furnace during startup and operation are nonlinear, consuming a large amount of energy during startup and requiring less energy during steady operation. Similarly, the energy requirements of an electrolyzer vary at different phases.

[0133] S32-2: adjusting an execution sequence of the production task processes and energy-mass flow allocation in real time through a dynamic planning or optimization algorithm according to the energy-mass demand characteristics corresponding to different production processes, so that the coordinated scheduling among the production processes is optimal, and further obtaining the inter-process energy-mass flow coordinated control model; wherein [0134] the inter-process energy-mass flow coordinated control model takes time scheduling cost among the production processes, dependency relationships among the production processes and resource limitation as constraint conditions; and for example, the energy-mass demands of a process may affect the subsequent startup time and efficiency.

[0135] In an example of this embodiment, in the step S32-1, assuming that the energy-mass flow demand of a process may vary with the startup and operation phases of the process, the demand of the process may be expressed by the function

[00017]$\begin{array}{cc}{E}_j\left(t\right)=g_j\left(t\right)& \left(18\right)\end{array}$

[0136] wherein E.sub.j (t) is the energy-mass flow demand of the process J.sub.j at t, and g.sub.j(t) can be defined according to the operating characteristics of the process, such as the nonlinear energy-mass demand in the startup phase and the lower demand in the stable phase.

[0137] In an example of this embodiment, in the step S32-2, assuming that J.sub.j and J.sub.j+1 are two adjacent processes, the energy-mass flow demands of the process J.sub.j affect the execution time t.sub.j+1 of the process J.sub.j+1:

[00018]$\begin{array}{cc}{t}_{j+1}=t_j+{}_j& \left(19\right)\end{array}$ [0138] wherein .sub.j represents the execution time of the process J.sub.j, and the energy-mass demands of the process J.sub.j+1 meet the resource constraint:

[00019]$\begin{array}{cc}{E}_{j+1}\left(t\right)R_j\left(t\right)& \left(20\right)\end{array}$ [0139] wherein R.sub.j(t) is the remaining energy-mass flow resources after the process J.sub.j is completed.

[0140] In the step S33 of this embodiment, the traditional resource-task network model usually assumes that resources are static and independent; however, in the multi-energy and mass flow system, the interaction between resources and tasks is dynamic and interdependent. To adapt to this complexity, it is necessary to expand the traditional resource-task network model. In this embodiment, the dynamic characteristics of energy-mass flows in time and space are introduced to construct an expanded resource-task network model.

[0141] Based on this, the extended resource-task network model in this embodiment is constructed as a model for dynamically adjusting resource supply and task demand in the integrated energy system for green hydrogen metallurgy according to time series data.

[0142] In this embodiment, the temporal and spatial allocation of different energy-mass flows (such as electricity, hydrogen, and thermal energy) is crucial to the overall operation of the system. In a specific example of this embodiment, in the process of hydrogen production by electrolysis, the supply of electricity is closely related to the demand for hydrogen production. It is necessary to define a supply and demand node in the network model and introduce a time-varying function to describe the relationship between the supply and the demand. Assuming that the temporal and spatial allocation of multi-energy and mass flows (e.g., electricity E.sub.lec and hydrogen E.sub.H.sub.2) are represented by functions h(t,x) and k(t,x) respectively, the extended resource-task network model can be defined as:

[00020]$\begin{array}{cc}{E}_{\mathrm{elec}}\left(t,x\right)=h\left(t,x\right)& \left(21\right)\end{array}$$\begin{array}{cc}{E}_{H_2}\left(t,x\right)=k\left(t,x\right)& \left(22\right)\end{array}$ [0143] wherein t is time, x is spatial position, and the functions h and k describe the dynamic allocation of different energy-mass flows in the time and space dimensions.

[0144] In this embodiment, when the extended resource-task network model is configured to perform coordinated control on resources related to a plurality of production tasks, multi-association processing is performed on the resources, an optimization model with constraint conditions is established, and multi-energy and mass supply coordinated matching with production task demands are achieved; wherein the constraint conditions include a resource supply capacity, dynamic allocation of energy-mass flows and time series dependency of production tasks.

[0145] In a specific example of this embodiment, for a task T.sub.i involving a plurality of resources, the objective of this task is to minimize the consumption of energy-mass flows to meet the task requirements. The optimization can be expressed as:

[00021]$\begin{array}{cc}\min \underset{r=1}{\overset{m}{.Math.}}{E}_{r,i}\left(t\right)& \left(23\right)\end{array}$ [0146] wherein E.sub.r,i(t) represents the consumption of the resource r in the task T.sub.i, and the constraints are:

[00022]$\begin{array}{cc}{E}_{r,i}\left(t\right)D_{r,i}\left(t\right)& \left(24\right)\end{array}$ [0147] wherein D.sub.r,i(t) represents the minimum demand of the task T.sub.i for the resource .

[0148] In this embodiment, in the extended resource-task network model, each task node not only represents the execution progress of the task, but also reflects the consumption of the energy-mass flows. Based on this, when the resource-task network model is configured to allocate resources and tasks, a dynamic optimization algorithm is adopted to dynamically adjust the execution sequence of the production tasks and the allocation strategy of the energy-mass flows according to the current energy-mass flow allocation in the integrated energy system for green hydrogen metallurgy, so that the optimal efficiency and stability of the system are ensured.

[0149] In this embodiment, the dynamic optimization algorithm in the above process may be a heuristic algorithm or a large-scale linear programming method. For example, the heuristic algorithm includes a genetic algorithm and a particle swarm optimization algorithm.

[0150] In the step S4 of the embodiment of this application, in a complex multi-energy and mass flow system, due to the volatility of energy-mass flows and the uncertainty of system demands, traditional control strategies are difficult to achieve efficient production control. Therefore, it is necessary to use the multi-model predictive control (MPC) strategy to achieve real-time control of system operation. In this embodiment, a plurality of random variable prediction models are designed, and production control is performed in a multi-model combination mode to effectively deal with system complexity and uncertainty.

[0151] Based on this, as shown in FIG. 4, the step S4 in this embodiment includes the following substeps: [0152] S41: performing time series characteristic analysis on the random variables involved in the production process based on task-resource allocation result obtained by the extended resource-task network model; [0153] for example, the random variables involved in the production process include wind power, photovoltaic power generation, user-side load demand, and the like, the prediction of these random variables is crucial for production regulation, and it is necessary to establish a discrete-time prediction model to estimate the future changes of these variables in advance; [0154] S42: constructing a discrete-time prediction model according to a time series characteristic analysis result of the random variables, and predicting the random variables involved in the production process at the future moment by using the discrete-time prediction model; [0155] S43: inputting the predicted random variables into a multi-model prediction mechanism based on a local model, predicting energy-mass flow allocation strategies in different operating states, and inputting the energy-mass-flow allocation strategies into the multi-model prediction mechanism based on the multi-controller combination; and [0156] S44: in the multi-model prediction mechanism based on the multi-controller combination, on the basis of ensuring scheduling consistency among independent controllers, selecting a corresponding independent controller for energy-mass flow control according to the involved energy-mass flow characteristics, and outputting the energy-mass flow control strategy at the current working condition to achieve the coordinated control of the integrated energy system for green hydrogen metallurgy.

[0157] In an example of this embodiment, in the step S41, the method for performing time series characteristic analysis on the random variables involved in the production process includes: [0158] performing time series analysis on each random variable to capture the volatility characteristics in the historical data of the variable. Classic time series models such as an autoregressive integrated moving average (ARIMA) model and a generalized autoregressive conditional heteroskedasticity (GARCH) model can be used to fit the volatility patterns of various random variables on different timescales.

[0159] Each random variable x.sub.t is subjected to time series analysis to capture the historical volatility characteristics of this variable. The autoregressive integrated moving average (ARIMA) model can be used to model the random variable:

[00023]$\begin{array}{cc}{x}_t={}_1{x}_{t-1}+{}_2{x}_{t-2}+.Math.+{}_p{x}_{t-p}+{}_t& \left(25\right)\end{array}$ [0160] wherein .sub.t is the noise term.

[0161] In an example of this embodiment, in the step S42, a discrete-time prediction model is constructed based on the results of the time series characteristic analysis. This model can discretize future time series fluctuations into several time steps (such as minutes, hours, and days), thereby providing accurate prediction values for production control.

[0162] In this embodiment, to further improve the prediction precision, an adaptive weighted prediction model can be introduced to perform weighted fusion on the results of a plurality of prediction methods such as neural networks and decision trees to generate more accurate prediction outputs.

[0163] The discrete-time prediction model is constructed based on the time series characteristics to predict the value of random variables at future moments. The prediction results of a plurality of models (such as neural networks and decision trees) are combined through weighted prediction methods:

[00024]$\begin{array}{cc}{\hat{x}}_{t+1}={}_1{\hat{x}}_{\mathrm{NN}}\left(t+1\right)+{}_2{\hat{x}}_{\mathrm{DT}}\left(t+1\right)& \left(26\right)\end{array}$

[0164] In this embodiment, the discrete-time model constructed in the step S42 not only outputs a single predicted value but also provides a probability distribution of random variables. For example, in the prediction of photovoltaic power generation, it is necessary not only to predict the expected power generation, but also to provide the probability distribution of power generation within a certain range. The prediction of the probability density function can better deal with the uncertainty in the system, thereby providing richer information for subsequent control strategies.

[0165] In this embodiment, the uncertainty in the system is handled by predicting the probability distribution of random variables. For example, the distribution information of random variables is obtained by predicting the probability density function (x)

[00025]$\begin{array}{cc}P\left(axb\right)=\underset{a}{\overset{b}{}}f\left(x\right)\mathrm{dx}& \left(27\right)\end{array}$

[0166] In the step S4 of this embodiment, due to the complexity of the system, a single model is difficult to cover all operating conditions of the system. In this embodiment, a multi-model predictive control (MPC) mechanism is introduced to flexibly respond to different production scenarios and energy-mass flow allocation requirements through the combination of a plurality of local models. Therefore, in the step S43 of this embodiment, the energy-mass flow allocation strategy in different operating states is predicted by the multi-model prediction mechanism based on the local models.

[0167] In this embodiment, the multi-model prediction mechanism based on the local model combination refers to: [0168] constructing a local model for achieving energy-mass flow allocation prediction under different working conditions, and switching to an optimal local model according to the operating state of the integrated energy system for green hydrogen metallurgy to perform corresponding energy-mass flow allocation prediction based on a conditional model switching mechanism.

[0169] Specifically, in this embodiment, the local models are constructed for different operation phases of the system. For example, under low-load and high-load conditions, the energy-mass flow allocation rules and response characteristics of the system are different. For these different working conditions, local models based on linear systems, nonlinear models or machine learning models (such as a long short-term memory network (LSTM) or a support vector machine (SVM)) can be established respectively.

[0170] The local models are constructed for different operating conditions of the system. For example, under different load conditions, the energy-mass flow allocation rules are different, and different local models can be established. Assuming that the system state equation is:

[00026]$\begin{array}{cc}{x}_{t+1}=f_i\left(x_t,u_t\right)& \left(28\right)\end{array}$ [0171] wherein .sub.t is the i-th local model.

[0172] In this embodiment, a conditional model switching mechanism is introduced into the multi-model prediction mechanism based on local models, which can select an appropriate local model under different system operating conditions. For example, when the system load fluctuates greatly, a local model suitable for a high fluctuation condition may be switched to; and when the system is operating smoothly, a simple linear model may be selected. The model switching mechanism can be implemented by setting thresholds, state observers, or dynamic discrimination methods based on real-time data.

[0173] In an example of this embodiment, a model switching threshold is set, and an appropriate local model is selected according to the system state:

[00027]$\begin{array}{cc}f\left(x_t,u_t\right)=\left\{\begin{array}{c}{f}_1\left(x_t,u_t\right),.Math.x_t-{\hat{x}}_t.Math.\\ f_2\left(x_t,u_t\right),.Math.x_t-{\hat{x}}_t.Math.>\end{array}\right\}& \left(29\right)\end{array}$

[0174] In this embodiment, for the multi-model prediction mechanism based on local models, in addition to using the conditional model switching mechanism to switch a single local model, the control precision can also be improved through the multi-model combination optimization. A model fusion strategy is adopted to perform weighted averaging or fusion of the prediction results of a plurality of local models to obtain a more accurate system state prediction. The Bayesian model averaging method can be used to dynamically adjust the weights according to the historical prediction precision of each model to obtain the optimal prediction results.

[0175] In an example of this embodiment, the outputs of a plurality of models are weighted averaged or fused through a model fusion strategy.

[00028]$\begin{array}{cc}\hat{y}=\underset{i=1}{\overset{N}{.Math.}}{}_i{\hat{y}}_i& \left(30\right)\end{array}$ [0176] wherein .sub.i is the output of the 1-th model, and is the weight.

[0177] In the step S4 of the embodiment of this application, on the basis of the multi-model prediction mechanism based on the local model, a multi-controller combination strategy is also designed in this embodiment to cope with different working conditions and energy-mass flow control demands. Local optimization and collaborative work of different controllers can ensure the stable operation of the system. Based on this, in the step S44 of this embodiment, energy-mass flow control for different working conditions is performed through a multi-model prediction mechanism based on a plurality of controllers, thereby achieving coordinated control of the integrated energy system for green hydrogen metallurgy.

[0178] In this embodiment, the multi-model prediction mechanism based on the multi-controller combination refers to: [0179] on the basis of an output energy-mass flow allocation strategy predicted by a multi-model prediction mechanism of local model combination, designing corresponding independent controllers according to different energy-mass flow characteristics, and introducing a global objective function to optimize a control target of each independent controller by adopting a constraint optimization-based coordinated mechanism according to different working conditions and energy-mass flow control requirements, so as to obtain an overall optimal energy-mass flow control strategy for the system; wherein an output of each independent controller is associated.

[0180] Specifically, in this embodiment, independent controllers are designed for different energy-mass flow characteristics. For example, for hydrogen production by electrolysis, an energy controller based on model prediction can be designed to adjust the operating power of the electrolyzer in real time. For hydrogen storage systems, a hydrogen storage controller based on inventory prediction can be designed to ensure the appropriateness of hydrogen reserves.

[0181] In this embodiment, the outputs of the independent controllers are interrelated and need to be coordinated and controlled through a coordinated mechanism. For example, the controller output of hydrogen production by electrolysis may affect the operating state of the hydrogen storage controller. Therefore, a multi-controller coordinated algorithm is required to ensure the scheduling consistency among the controllers. A coordinated mechanism based on constrained optimization can be adopted to introduce a global objective function and optimize the control objectives of each controller, thereby achieving optimal control of the entire system.

[0182] In this embodiment, on the basis of the above multi-model prediction mechanism based on the plurality of controllers, a multi-controller combination based on distributed combination optimization can also be constructed to deal with the information lag and computational burden problems that the centralized controller may face when the system scale is large and the energy-mass flow links are widely allocated. In this case, a distributed optimization algorithm can be introduced, the system is divided into a plurality of subsystems, each subsystem is independently controlled by a corresponding controller, and the subsystems are coordinated through information exchange. Distributed optimization algorithms such as the alternating direction method of multipliers (ADMM) can be used to coordinate decisions between local controllers, thereby improving the global control efficiency.

[0183] The principle and implementation of the present application are described herein by using specific examples. The descriptions about embodiments of the present application are merely provided, to help understand the method and core ideas of the present application. In addition, those of ordinary skill in the art can make variations and modifications to the present application in terms of the specific implementations and application scopes according to the ideas of the present application. Therefore, the content of specification shall not be construed as a limit to the present application.

[0184] It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to help readers understand the principles of the present application, and it should be understood that the protection scope of the present application is not limited to such specific descriptions and embodiments. Those of ordinary skill in the art may make various other specific modifications and combinations based on the technical inspirations disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the protection scope of the present application.