Method and device for evaluating load adjustability of iron and steel industry, and electronic equipment

Abstract

The invention discloses a load adjustability assessment method and device for the iron and steel industry and electronic equipment, and relates to the field of power systems. The load adjusting capability of the steel industry can be quantified, and support is provided for participation in power grid demand side response. The method comprises the following steps: acquiring a plurality of state parameters in a steel production scene to obtain a current state; the current state is input into a strategy network, a plurality of action strategies output by the strategy network in the current state are obtained, the action strategies comprise power adjusting tracks of all the load units in a preset time period, and the power adjusting tracks comprise adjusting instructions of power down-regulation or up-regulation; the current state and the action strategy in the current state are input into a value network, a value evaluation value output by the value network is obtained, and the value evaluation value is an expected value of an accumulated reward which can be obtained by the action strategy in the current state; and determining an optimal action strategy in the current state through the value evaluation value, and evaluating the adjustable capability of the load unit in the preset time period according to the optimal action strategy.

Description

Technical Field

[0001] This application relates to the field of power systems, and more particularly to a method, apparatus, and electronic equipment for assessing the load adjustability of the steel industry. Background Technology

[0002] To maintain the real-time balance of the power grid, the demand-side response mechanism aims to guide electricity users, especially large industrial users, to adjust their electricity consumption behavior according to the real-time status of the power grid, thereby providing ancillary services to the power grid.

[0003] As a typical high-energy-consuming industry, the steel industry accounts for about 15% of the total industrial electricity consumption in the society. The steel production process is complex, involving multiple high-energy-consuming processes such as sintering, ironmaking, steelmaking, and rolling. Among them, equipment such as electric arc furnaces (EAFs) and rolling mills are characterized by high power, frequent start-ups and shutdowns, and drastic load fluctuations. This makes steel companies both "load impact sources" for the power grid and also contain huge adjustment potential, making them ideal candidates for participating in demand-side response.

[0004] Most existing methods for assessing load adjustability rely on historical data statistical analysis, feature extraction, and clustering (such as K-means). These methods typically provide a fixed, static potential value, failing to reflect the impact of dynamic changes in production plans, equipment status, and market electricity prices on adjustability. Some assessment models, for ease of solution, simplify complex production process constraints into linear or fixed boundary conditions, neglecting the strong coupling relationships between processes and the nonlinear and time-varying characteristics of the production flow, leading to significant deviations between assessment results and actual conditions. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic equipment for assessing the load adjustability of the steel industry, which can accurately quantify the load adjustability of the steel industry and is conducive to improving the benefits of the steel industry's participation in the grid demand-side response.

[0006] In a first aspect, this application provides a method for assessing the load adjustability of the steel industry, comprising: Collect multiple state parameters from the steel production scenario to obtain the current state; The current state is input into the policy network to obtain multiple action policies output by the policy network in the current state. The action policies include the power adjustment trajectory of each load unit within a preset time period. The power adjustment trajectory includes adjustment instructions for power reduction or increase. The current state and the action strategy under the current state are input into the value network to obtain the value evaluation value output by the value network. The value evaluation value is the expected value of the cumulative reward that the action strategy under the current state can obtain. The optimal action strategy under the current state is determined by the value assessment value, and the adjustability of the load unit within the preset time period is evaluated according to the optimal action strategy.

[0007] The load adjustability assessment method provided in this embodiment treats adjustable load units in a steel production scenario as a whole, and outputs power regulation action strategies for this whole. This allows for better learning of the strong coupling relationships between processes and the nonlinear and time-varying characteristics of the production process, making the model's output action strategies more closely aligned with actual conditions and improving the accuracy of adjustability assessment. The current state of the load unit changes in real time with the steel production scenario, and the power regulation strategy for a certain period is output through the strategy network, making the power regulation assessment more consistent with actual dynamic changes and further improving the accuracy of the assessment. Furthermore, this solution uses the current state of the steel production scenario to predict and output the power regulation strategy for a preset time period, proactively and forward-lookingly optimizing the scheduling strategy to maximize benefits. This provides a decision-making basis for the steel industry to respond to grid regulation demands, which is conducive to maximizing economic benefits.

[0008] Secondly, this application provides a load adjustability assessment device for the steel industry, comprising: The data detection module is used to collect multiple state parameters in the steel production scenario to obtain the current state; The action output module is used to input the current state into the strategy network and obtain multiple action strategies output by the strategy network in the current state. The action strategies include the power adjustment trajectory of each load unit within a preset time period, and the power adjustment trajectory includes adjustment instructions for power reduction or increase. The action evaluation module is used to input the current state and the action strategy in the current state into the value network to obtain the value evaluation value output by the value network. The value evaluation value is the expected value of the cumulative reward that the action strategy in the current state can obtain. The adjustability assessment module is used to determine the optimal action strategy under the current state through the value assessment value, and to assess the adjustability of the load unit within the preset time period according to the optimal action strategy.

[0009] Thirdly, this application provides an electronic device including a memory and one or more processors. The memory stores one or more computer programs, each including instructions that, when executed by the processor, cause the electronic device to perform a load adjustability assessment method for the steel industry as described in the first aspect.

[0010] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the load adjustability assessment method for the steel industry as described in the first aspect.

[0011] Fifthly, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the load adjustability assessment method for the steel industry as described in the first aspect.

[0012] Understandably, the beneficial effects that the above-mentioned load adjustability assessment device, electronic equipment, computer-readable storage medium, and computer program product for the steel industry can achieve can be referred to the beneficial effects in the first aspect, and will not be repeated here. Attached Figure Description

[0013] Figure 1 A flowchart illustrating the load adjustability assessment method for the steel industry provided in this application embodiment; Figure 2 A schematic diagram of the structure of the load adjustability assessment device for the steel industry provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] To facilitate a clear description of the technical solutions in the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish identical or similar items with substantially the same function and effect. For example, "first chip" and "second chip" are only used to distinguish different chips and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" do not necessarily imply that they are different. It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be construed as being better or more advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. In the embodiments of this application, "at least one" means one or more, and "more than one" means two or more.

[0015] It should be noted that "at the time of..." in the embodiments of this application can be either at the instant when a certain situation occurs, or for a period of time after the occurrence of a certain situation. The embodiments of this application do not make specific limitations on this.

[0016] The implementation of this embodiment will now be described in detail with reference to the accompanying drawings.

[0017] This embodiment provides a method for assessing the load adjustability of the steel industry. For example, this method can be applied to various electronic devices such as computers (PCs), tablets, virtual reality / augmented reality devices, wearable devices, industrial computers, and vehicle-mounted systems; it can also be applied to servers, cloud computing, server clusters, etc. This embodiment does not impose any special limitations on these applications.

[0018] Figure 1 A schematic flowchart of the load adjustability assessment method for the steel industry provided in this application embodiment is shown.

[0019] like Figure 1 As shown, the method for assessing the load adjustability of the steel industry may include the following steps: Step 101: Collect multiple state parameters in the steel production scenario to obtain the current state.

[0020] The process of collecting multiple state parameters in the steel production scenario to obtain the current state specifically includes: acquiring the current load state of each load unit, the production process state of the steel production scenario, as well as the power consumption state and time state; and constructing the current load state, the production process state of the steel production scenario, as well as the power consumption state and time state into a multi-dimensional vector to obtain the current state.

[0021] The load unit includes an electric arc furnace, a heating furnace, a rolling mill, a blower, and a water pump.

[0022] To enable the agent to learn, a Markov Decision Process (MDP) framework can be constructed, including a state space, action space, and reward function. The agent can extract multiple state parameters from the steel production scenario based on the state space to determine the current state.

[0023] (1) State Space: Defines all the information that the agent needs to observe when making decisions, i.e., state parameters. State S is a multi-dimensional vector that can include: Equipment-level load status: Current operating power, operating stage, operating temperature, and remaining operating time of each key electrical equipment (such as electric arc furnace, LF furnace, rolling mill, heating furnace, fan, water pump, etc.).

[0024] Production process status: inventory of materials (molten steel, billets) for each process, logistics transfer status, process requirements of the current production batch, and progress of the production plan.

[0025] Grid and market conditions: real-time electricity price (or time-of-use price), demand response instructions issued by the grid (such as peak-shaving power and duration), and available grid capacity limitations.

[0026] Time status: current time, train / bus information, etc.

[0027] (2) Action Space: Defines the operations that the agent can perform. Action A is also a vector, representing a control command to one or more adjustable load units. The action space can be designed to be discrete or continuous, for example: Electric arc furnace / refining furnace: {Maintain current power, reduce power by 10%, reduce power by 20%, suspend power supply, resume power supply}.

[0028] Heating furnace: {Increase target temperature by 5℃, decrease target temperature by 5℃, maintain current heat preservation power}.

[0029] Rolling mill: {Maintain current rolling speed, reduce rolling speed by 5%, suspend feeding}.

[0030] Auxiliary equipment (fans / pumps): {Switch to high power mode, switch to energy saving mode}.

[0031] Step 102: Input the current state into the strategy network to obtain multiple action strategies output by the strategy network in the current state. The action strategies include the power adjustment trajectory of each load unit within a preset time period. The power adjustment trajectory includes adjustment commands for power reduction or increase.

[0032] Among them, the policy network is a deep neural network that includes a long short-term memory network.

[0033] In this embodiment, the agent adopts an Actor-Critic architecture, where the actor network is the policy network. Its input is the current state S, and its output is the probability distribution of performing each action A in this state. The actor network is responsible for decision-making, and its network structure adopts a deep neural network containing a Long Short-Term Memory (LSTM) layer to effectively process time-series related state information (such as load curves and production cycle time).

[0034] The current state is input into the policy network to obtain multiple action policies output by the policy network in the current state. Specifically, this includes: determining the state space and action space of the steel production scenario, wherein the action space includes multiple actions, each action representing a power adjustment command for one or more load units; inputting the current state into the policy network, wherein the policy network outputs the probability distribution of executing each action in the action space in the current state.

[0035] Step 103: Input the current state and the action strategy under the current state into the value network to obtain the value evaluation value output by the value network. The value evaluation value is the expected value of the cumulative reward that the action strategy under the current state can obtain.

[0036] Based on the probability distribution of each action in the action space in the current state output by the policy network, action policies with higher probabilities can be extracted and their value evaluated. The value network calculates the cumulative reward obtainable by the action policy using a reward function, which is:

[0037] in, The benefits are obtained through power regulation; The penalty is due to the impact on production caused by power regulation. It is a penalty for equipment wear caused by power regulation; , , It is the weighting coefficient.

[0038] This reward function can maximize the economic benefits of enterprises participating in demand-side response while ensuring production safety and product quality, thereby achieving optimal economic and production efficiency.

[0039] This implementation also includes a penalty function to determine the penalty value of the current action strategy. The penalty function can be preset and specifically evaluated based on factors such as equipment risks caused by power adjustment, production delays due to power reduction, and equipment losses caused by shutting down / starting load units. The penalty function determines the penalty value of the action strategy; if the penalty value of the current action strategy is within an acceptable range, the action strategy is feasible. Simultaneously, action strategies with penalty values ​​exceeding the acceptable range are excluded, thereby controlling the risk of power adjustment.

[0040] Before inputting the current state into the policy network, the method further includes optimizing the policy network using a proximal policy optimization (PPO) algorithm.

[0041] In this implementation, both the policy network and the value network are deep reinforcement learning (DRL) models. DRL models can learn continuously online, fine-tuning their policies as new data is generated, enabling them to adapt to changes in production processes, the introduction of new equipment, and changes in electricity market rules, exhibiting strong robustness and scalability. The policy network and value network are optimized using the states from the previous time period and the optimal action policy before the current state is reached, and the current action is then output.

[0042] Step 104: Determine the optimal action strategy under the current state based on the value assessment value, and evaluate the adjustability of the load unit within the preset time period according to the optimal action strategy.

[0043] The optimal action strategy can refer to one or more action strategies with a high expected cumulative reward, or one or more action strategies with a high expected cumulative reward and an acceptable penalty value. Evaluating the adjustability of the load unit according to the optimal action strategy specifically includes: if the adjustment demand corresponding to the current state is a power reduction, then iterating through each adjustment command in the optimal action strategy to determine the maximum feasible power reduction; if the adjustment demand corresponding to the current state is a power increase, then iterating through each adjustment command in the optimal action strategy to determine the maximum feasible power increase.

[0044] For example, the current state and the current adjustment demand are input into the strategy network, and the strategy network outputs the action strategy of the adjustment command of power increase or power decrease. Then, the adjustable capacity in the steel production scenario is determined by the maximum power increase or maximum power decrease of each load in the optimal action strategy.

[0045] The above process in this embodiment can be executed online in real time. The optimal action strategy is determined based on the current state of the steel production scenario. The optimal action strategy is maximized by satisfying the constraints of the steel production scenario and ensuring production safety and product quality through reward and penalty functions, thereby achieving optimal economic and production benefits for enterprises participating in demand-side response.

[0046] Before using the agent to determine the action policy, the agent is trained. This embodiment uses the advanced Proximal Policy Optimization (PPO) algorithm as the core training algorithm. The agent adopts an Actor-Critic architecture.

[0047] Offline training: (1) Actor Network (Policy Network): The input is the current state S, and the output is the probability distribution of each action A to be performed in this state. The Actor Network is responsible for decision-making. Its network structure adopts a deep neural network containing a Long Short-Term Memory (LSTM) layer to effectively process time-series related state information (such as load curves and production cycle).

[0048] (2) Critic Network (Value Network): The input is the current state S and action A, and the output is the value assessment V(S, A) after making action A, that is, the expected cumulative reward in the future that can be obtained by starting from the current state and following the current strategy. The Critic Network is responsible for evaluating the quality of decisions.

[0049] (3) Actor and Critic Update: The Clipped Surrogate Objective and Advantage Function of the PPO algorithm are used to continuously update the parameters of the Actor and Critic networks until the current optimal action strategy is output.

[0050] By leveraging a wealth of historical production data from steel enterprises (such as production logs from the MES system and equipment operation data from the SCADA system) and historical electricity price data, an offline MDP environment model is constructed. In this simulation environment, the agent undergoes millions, or even more, trial-and-error explorations of "simulated production days." In each exploration, the agent interacts with the environment, collecting a large number of samples (state, action, reward, next state). These samples are then used to continuously update the parameters of the Actor and Critic networks using the PPO algorithm until the agent's policy converges, enabling it to make near-optimal decisions under various conditions.

[0051] Online assessment and deployment: The trained agent is then deployed into the energy management system (EMS) or production scheduling system of the steel company. The evaluation process is as follows: a) The system collects the current status S (including production, equipment and power grid information) in real time.

[0052] b) Input the state S into the trained Actor network, and the network outputs the current optimal action policy.

[0053] c) Virtual simulation and boundary detection: In order to evaluate the "adjustability", the agent performs multiple virtual simulations based on the current state S, simulating the execution of a series of extreme actions of "up" and "down" (e.g., simulating shutting down all interruptible loads), and inputs the simulated state into the Critic network for value evaluation.

[0054] d) Output Adjustability Assessment Results: If the simulation results (such as estimated production delays and equipment risks) are within an acceptable penalty range, the system considers the adjustment amount feasible. By simulating different adjustment depths, the system can quickly analyze the optimal combination of adjustment strategies for the current moment.

[0055] In this embodiment, the DRL model can generate minute-level adjustability assessment results in real time based on dynamic factors such as production plans, equipment operating conditions, and market electricity prices, far exceeding the accuracy and timeliness of traditional static assessment methods. Based on the environmental model of MDP and the powerful fitting capabilities of deep neural networks, it can learn and handle complex, nonlinear process coupling constraints in steel production. Compared with other methods that regulate individual production processes, the assessment results are closer to actual production feasibility.

[0056] Taking a steel company with two 100-ton electric arc furnaces (EAF), two LF refining furnaces and one hot continuous rolling production line as an example, the specific implementation process of this embodiment is explained.

[0057] Step 1: Data Preparation and Environment Setup (1) Data source: Obtain historical data for at least one year from the enterprise's Manufacturing Execution System (MES) and Supervisory Control and Data Acquisition System (SCADA). The data time resolution is 1 minute.

[0058] MES data includes: production plan, steel grade, casting number, start / end time of each process (steelmaking, refining, continuous casting, rolling), molten steel temperature, material tracking, etc.

[0059] SCADA data includes: real-time power, voltage, current, and equipment status (running, standby, fault) of major equipment (EAF, LF, rolling mill main motor, heating furnace, fan, water pump, etc.).

[0060] External data: historical time-of-use electricity price data, historical demand response event data.

[0061] (2) Definition of state space: The dimension of the state vector S is approximately 50, specifically including:

[0062] Where P represents power, T represents temperature, S represents stage state (such as the melting period and oxidation period of EAF), L represents inventory (slab inventory), and t represents time.

[0063] (3) Action Space Definition: Action vector A is for the main adjustable equipment (load unit). For example, the action for EAF1 is { , , , The total action space is the Cartesian product of all device actions, selected through the probability distribution output by the Actor network.

[0064] (4) Reward function setting:

[0065] The first item is the response benefit. In response to time-of-use electricity pricing, This represents the actual response power. The second item is a production penalty. The first term is a nonlinear penalty function for the excessive temperature drop of molten steel; the third term is a penalty for equipment loss. It is a penalty function for the number of times the device starts and stops.

[0066] Step 2: Model Training (1) Network structure: Both the Actor and Critic networks employ a structure consisting of two LSTM layers and three fully connected (FC) layers. The LSTM layers are used to capture the temporal dependencies of the states, while the FC layers are used for feature extraction and decision output.

[0067] (2) Training process: a) Initialize the Actor and Critic network parameters.

[0068] b) In the constructed simulation environment, train for 2 million episodes. Each episode simulates a complete 24-hour production process.

[0069] c) At each time t, the agent observes the state. Actor network outputs actions After the environment performs an action, it returns to the next state. and rewards .

[0070] d) Sample Store in the experience replay pool.

[0071] e) Use the Clipped Surrogate Objective and Advantage Function of the PPO algorithm to update the weights of the Actor and Critic networks.

[0072] Step 3: Online evaluation and experimental verification (1) Deployment: The trained model is packaged into an API service and deployed in the enterprise energy management center.

[0073] (2) Online assessment process: The system calls the assessment service once per minute.

[0074] a) Get the current real-time status .

[0075] b) Downgrade the competency assessment: The system generates a series of virtual downward adjustment actions, such as ={EAF1 power 60%, EAF2 power 40%}. For each virtual action The system starts from the current state The simulation begins by recreating the production evolution over the next 30 minutes after the action is executed, and the cumulative expected reward V of this trajectory is evaluated using a Critic network. , If the value of V is not lower than the preset "production safety" threshold, then the downward adjustment is considered the optimal action strategy, and the power corresponding to this downward adjustment action is... It is feasible; iterate through all possible down-adjustment combinations to find the maximum feasible down-adjustment power. .

[0076] c) Upgrade capacity assessment: The process is similar, but virtual upgrade actions are performed (such as turning on the heating furnace in advance to preheat).

[0077] d) Output: The system outputs a regulation capacity assessment report.

[0078] To verify the effectiveness, the DRL method in this embodiment was compared with two benchmark methods through simulation experiments. The simulation results are shown in Table 1: Method 1: Historical Average (HA): Based on historical data, calculate the average adjustable amount under similar production periods and load levels.

[0079] Method 2: Mixed Integer Linear Programming (MILP): Establish a simplified production scheduling optimization model and solve it with the goal of maximizing response benefits.

[0080] Table 1: Comparison of Evaluation Results and Simulation Verification by Different Methods

[0081] Simulation results show that the DRL method provided in this embodiment is far more accurate than traditional methods in assessing adjustability, with a prediction error of only 5.2%. Furthermore, due to its optimization of decision-making, it can bring the highest economic benefits in simulated daily demand response scheduling while minimizing the impact on production. Its rapid online computing capabilities also fully meet the real-time application needs of steel enterprises.

[0082] Furthermore, this embodiment also provides a load adjustability assessment device for the steel industry, which can be used to perform the aforementioned load adjustability assessment method for the steel industry. Figure 2As shown, the load adjustability assessment device 200 for the steel industry specifically includes: a data detection module 201, used to collect multiple state parameters in the steel production scenario to obtain the current state; an action output module 202, used to input the current state into a strategy network to obtain multiple action strategies output by the strategy network in the current state, wherein the action strategies include the power adjustment trajectory of each load unit within a preset time period, and the power adjustment trajectory includes adjustment instructions for power reduction or increase; an action assessment module 203, used to input the current state and the action strategies in the current state into a value network to obtain the value assessment value output by the value network, wherein the value assessment value is the expected value of the cumulative reward that the action strategy in the current state can obtain; and an adjustability assessment module 204, used to determine the optimal action strategy in the current state through the value assessment value, and to assess the adjustability of the load unit within the preset time period according to the optimal action strategy.

[0083] In one implementation, the adjustable capability assessment module 204 is specifically used to determine the maximum feasible power reduction by traversing each adjustment instruction in the optimal action strategy if the adjustment requirement corresponding to the current state is power reduction; and to determine the maximum feasible power increase by traversing each adjustment instruction in the optimal action strategy if the adjustment requirement corresponding to the current state is power increase.

[0084] In one implementation, the data detection module 201 is specifically used to obtain the current load status of each load unit, the production process status of the steel production scenario, and the power consumption status and time status; and to construct a multi-dimensional vector from the current load status, the production process status of the steel production scenario, and the power consumption status and time status to obtain the current status.

[0085] The value network calculates the cumulative reward obtainable by the action policy through a reward function, which is:

[0086] in, The benefits are obtained through power regulation; The penalty is due to the impact on production caused by power regulation. It is a penalty for equipment wear caused by power regulation; , , It is the weighting coefficient.

[0087] In one embodiment, the load adjustability assessment device 200 for the steel industry further includes an optimization module for optimizing the policy network using a near-end policy optimization algorithm.

[0088] In one implementation, the action output module 202 is specifically used to determine the state space and action space of the steel production scenario, wherein the action space includes multiple actions, each action representing a power adjustment command for one or more load units; input the current state into the strategy network, wherein the strategy network outputs the probability distribution of executing each action in the action space under the current state.

[0089] For example, the policy network is a deep neural network including a long short-term memory network. The load unit includes an electric arc furnace, a heating furnace, a rolling mill, a blower, and a water pump.

[0090] The specific details of each module or unit in the aforementioned load adjustability assessment device for the steel industry have been described in detail in the corresponding load adjustability assessment method for the steel industry, so they will not be repeated here.

[0091] This application also provides an electronic device. Figure 3 A schematic diagram of the structure of an electronic device suitable for implementing embodiments of the present disclosure is shown. Figure 3 The electronic device 600 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments disclosed herein.

[0092] like Figure 3 As shown, the electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes based on a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0093] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0094] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs the functions defined in the embodiments of this application.

[0095] For example, when the computer program is executed by the central processing unit (CPU) 601, it can perform the following: collect multiple state parameters in the steel production scenario to obtain the current state; input the current state into a strategy network to obtain multiple action strategies output by the strategy network in the current state, wherein the action strategies include the power adjustment trajectory of each load unit within a preset time period, and the power adjustment trajectory includes adjustment instructions for power reduction or increase; input the current state and the action strategies in the current state into a value network to obtain a value assessment value output by the value network, wherein the value assessment value is the expected value of the cumulative reward that the action strategies in the current state can obtain; determine the optimal action strategy in the current state through the value assessment value, and evaluate the adjustability of the load units within the preset time period according to the optimal action strategy.

[0096] It should be noted that the computer-readable medium disclosed herein may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0097] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0098] The units described in the embodiments of this disclosure can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the unit itself.

[0099] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which include instructions that, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.

[0100] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for assessing the load adjustability of the steel industry, characterized in that, include: Collect multiple state parameters from the steel production scenario to obtain the current state; The current state is input into the policy network to obtain multiple action policies output by the policy network in the current state. The action policies include the power adjustment trajectory of each load unit within a preset time period. The power adjustment trajectory includes adjustment instructions for power reduction or increase. The current state and the action strategy under the current state are input into the value network to obtain the value evaluation value output by the value network. The value evaluation value is the expected value of the cumulative reward that the action strategy under the current state can obtain. The optimal action strategy under the current state is determined by the value assessment value, and the adjustability of the load unit within the preset time period is evaluated according to the optimal action strategy.

2. The method for assessing the load adjustability of the steel industry according to claim 1, characterized in that, The step of evaluating the adjustability of the load unit according to the optimal action strategy includes: If the adjustment requirement corresponding to the current state is power reduction, then iterate through each adjustment instruction in the optimal action strategy to determine the maximum feasible power reduction. If the adjustment requirement corresponding to the current state is an increase in power, then iterate through each adjustment instruction in the optimal action strategy to determine the maximum feasible power increase.

3. The method for assessing the load adjustability of the steel industry according to claim 1, characterized in that, The current state is obtained by collecting multiple state parameters from the steel production scenario, including: Obtain the current load status of each load unit, the production process status of the steel production scenario, as well as the power consumption status and time status; The current load status, the production process status of the steel production scenario, the electricity consumption status, and the time status are combined to form a multi-dimensional vector to obtain the current status.

4. The method for assessing the load adjustability of the steel industry according to claim 1, characterized in that, The value network calculates the cumulative reward obtainable by the action policy through a reward function, which is: in, The benefits are obtained through power regulation; The penalty is due to the production disruption caused by power regulation. It is a penalty for equipment wear caused by power regulation; , , It is the weighting coefficient.

5. The method for assessing the load adjustability of the steel industry according to claim 1, characterized in that, Before inputting the current state into the policy network, the method further includes: The policy network is optimized using a near-end policy optimization algorithm.

6. The method for assessing the load adjustability of the steel industry according to claim 5, characterized in that, The step of inputting the current state into the policy network to obtain multiple action policies in the current state output by the policy network includes: The state space and action space of the steel production scenario are determined. The action space includes multiple actions, each of which represents a power adjustment command for one or more load units. The current state is input into the policy network, which outputs the probability distribution of each action in the action space to be executed in the current state.

7. The method for assessing the load adjustability of the steel industry according to claim 6, characterized in that, The policy network is a deep neural network that includes a long short-term memory network.

8. The method for assessing the load adjustability of the steel industry according to claim 3, characterized in that, The load unit includes an electric arc furnace, a heating furnace, a rolling mill, a fan, and a water pump.

9. A device for assessing the load adjustability of the steel industry, characterized in that, include: The data detection module is used to collect multiple state parameters in the steel production scenario to obtain the current state; The action output module is used to input the current state into the strategy network and obtain multiple action strategies output by the strategy network in the current state. The action strategies include the power adjustment trajectory of each load unit within a preset time period, and the power adjustment trajectory includes adjustment instructions for power reduction or increase. The action evaluation module is used to input the current state and the action strategy in the current state into the value network to obtain the value evaluation value output by the value network. The value evaluation value is the expected value of the cumulative reward that the action strategy in the current state can obtain. The adjustability assessment module is used to determine the optimal action strategy under the current state through the value assessment value, and to assess the adjustability of the load unit within the preset time period according to the optimal action strategy.

10. An electronic device, characterized in that, The device includes a processor and a memory, the memory storing one or more computer programs, the one or more computer programs including instructions that, when executed by the electronic device, cause the electronic device to perform the load adjustability assessment method for the steel industry as described in any one of claims 1 to 8.

Images