A coal preparation plant energy storage system and energy storage method based on digital twin

Abstract

The present application provides a coal preparation plant energy storage system and an energy storage method based on digital twin. The energy storage system includes a hybrid energy storage system, a battery management system, a high-voltage DC protection system, a pre-charge / discharge circuit, a high-voltage AC power distribution system, a centralized PCS cluster, and a grid power electronic interface. The battery energy storage array is connected to the power-consuming equipment in the coal preparation plant. A digital twin model of the power-consuming equipment is constructed, the production plan of the coal preparation plant is obtained, the power-consuming equipment required for the technological process under the current production plan is determined, and the working load of all power-consuming equipment under the current production plan is simulated through the digital twin model; according to the working load of the power-consuming equipment, the required power W at the peak electricity price under the current production plan is predicted, and the corresponding energy storage capacity is configured. Through full-process efficiency control, the production efficiency is improved, an electric energy storage system is established, and the best charge / discharge planning plan is given through digital twin technology and local peak-valley electricity prices, thereby reducing the unit power consumption cost.

Description

Technical Field

[0001] This application relates to the technical field of energy storage, and particularly relates to a coal preparation plant energy storage system and an energy storage method based on digital twin. Background Art

[0002] The thermal coal preparation plant removes impurities such as gangue and minerals with high sulfur content from the raw coal mined by the coal mine by means of diversified coal preparation technologies and processes. Subsequently, according to different quality requirements and market demands, the raw coal is sorted and processed to produce thermal coal products that meet the corresponding standards. The purpose of setting up the coal preparation plant is to improve the quality of coal so that coal can better serve as a power fuel and is widely used in many power fields such as power generation, heat supply, coal for steam locomotives, and fuel for industrial boilers. In terms of product quality control, the thermal coal preparation plant needs to ensure that the calorific value of the slack coal reaches the specified standard, and at the same time, the ash content of the clean coal also meets the corresponding requirements.

[0003] In terms of production cost, electricity consumption is a major production cost. How to reduce the electricity cost is one of the main directions of cost control in the coal preparation plant. With the development of energy storage technology, by building an energy storage system and using the electricity price difference between the valley period and the peak period to control the electricity consumption cost is a common means for energy-consuming factories. For different coal media in the coal preparation plant, its production process is relatively large, and the energy consumption of equipment in each production process link is also different, and the daily power consumption also varies greatly. The existing industrial energy storage system charges according to the total capacity of the battery array during the valley period every day, rather than according to the actual electricity consumption required the next day, and frequent charge and discharge has a greater impact on the life of the battery array. Summary of the Invention

[0004] The purpose of the embodiments of this application is to provide a coal preparation plant energy storage system and an energy storage method based on digital twin to solve the above problems existing in the prior art and reduce the energy consumption of the coal preparation plant.

[0005] In a first aspect, a coal preparation plant energy storage system based on digital twin includes a hybrid energy storage system, a battery management system, a high-voltage DC protection system, a pre-charge / discharge circuit, a high-voltage AC power distribution system, a centralized PCS cluster, and a grid power electronics interface. The hybrid energy storage system is connected to the power-consuming equipment in the coal preparation plant, and it includes a lithium battery array and a supercapacitor array. An equipment-level digital twin model is constructed based on the dynamic data of the power-consuming equipment and coupled to generate a system-level digital twin model. According to the production plan, the electricity price at different local times, and the charge / discharge cost of the hybrid energy storage system, the start / stop combination, working period of the power-consuming equipment, and the power of the motors of the power-consuming equipment are simulated through the system-level digital twin model to find the production route with the lowest total electricity price for a single working day. The constraint conditions include the processing volume and start / stop frequency of all power-consuming equipment for a single working day. According to the optimal production route, the start / stop time period of the power-consuming equipment and the power of the motors are determined. According to the working period of the power-consuming equipment and the power of the motors, the energy storage capacity of the hybrid energy storage system is determined. According to the working period of the power-consuming equipment, the peak load time period at the moment of start / stop of the power-consuming equipment is predicted. The battery management system turns on the supercapacitor array during the peak load time period to perform high-frequency power compensation for the power-consuming equipment, and the lithium battery array supplies power to the power-consuming equipment during the time period other than the peak load of the peak electricity.

[0006] In an optional implementation manner, the method for finding the production route with the lowest total electricity price for a single working day includes:

[0007] The time periods of the low electricity price period and the high electricity price period are respectively divided into T and T' time periods, and the duration of each time period is , j represents the jth time period of the low electricity price period, and k represents the kth time period of the high electricity price period;

[0008] After numbering the power-consuming equipment, a set is formed, and m is the total number of equipment; let the power of the power-consuming equipment numbered i in the j time period be , and let the power of the power-consuming equipment numbered i in the k time period be , , Characterize whether the power-consuming equipment numbered i works in the j time period. When , it means that the power-consuming equipment i works in the j time period during the low electricity price period. When , it means that the power-consuming equipment i does not work; , Characterize whether the power-consuming equipment numbered i works in the k time period. When , it means that the power-consuming equipment i works in the k time period during the high electricity price period. When , it means that the power-consuming equipment i does not work;

[0009] Let the low electricity price be C and the power supply cost price of the energy storage system be d; the total daily electricity cost C-low during the low electricity price period:

[0010]

[0011] Total electricity cost C-high during peak hours:

[0012]

[0013] Establish the objective function C-total = C-low + C-high, where C-total is the total electricity cost, and solve it according to the genetic algorithm and as well as to minimize the total electricity cost; determine the on / off state of the power-consuming equipment at each time period through and values.

[0014] In an optional implementation, the calculation method of the energy storage capacity of the hybrid energy storage system includes:

[0015] According to the optimal production route, determine the start / stop time periods of the power-consuming equipment and the power of the motor, and determine the total peak-hour electricity consumption W of a working day based on the working time periods of the power-consuming equipment and the motor power.

[0016]

[0017] The energy storage capacity E of the hybrid energy storage system = W × a / (η × DOD), where DOD is the depth of discharge, η is the charge / discharge efficiency, and a is the redundancy factor.

[0018] In an optional implementation, the following constraint conditions can also be added to solve the objective function C-total = C-low + C-high. Set a threshold for the overall load factor state Lsystem value, L system The calculation process of the value includes:

[0019] Based on the real-time simulation operation ability of the quantity and quality process in the digital twin model and the screening and sinking data in the real-time coal quality database, obtain the real-time processing capacity of each device and system, get the optimal raw coal processing capacity of the current system, and calculate the production efficiency of the current coal preparation plant in combination with the real-time raw coal washing volume (real-time raw coal washing volume / optimal raw coal processing capacity). Analyze and calculate the load rate of each device, divide the load levels (low load, optimal load, high load, over load, and can be assigned values for calculation respectively, such as 1, 2, 3, 4). Further, according to the situation of the devices included in the main production systems (raw coal preparation system, heavy medium separation system, coarse coal slime recovery system, product transportation system, etc.) in each interval (low load, optimal load, high load, over load), determine the probability distribution model it follows based on the historical load state of the device, collect the fault data and load abnormal data of the device within a certain period of time, such as the number of faults, fault time, average load rate value in the abnormal load interval, abnormal load duration, etc., and calculate the reliability R of a single device, R = R0(1 - F); where R0 is the basic credibility, R0 = e -λt , λ = number of faults / unit operation time, t is the device life; F represents the reliability decay factor caused by abnormal load, F = min(△L×p×k, 1), k is an adjustment coefficient that needs to be set according to experience, △L = (average load rate - rated load rate) / rated load rate, p = abnormal load duration / mean time between failures; the obtained R value can automatically adjust the allocation weight wi of each device, allocate the weight wi for each device, and the allocation weight wi is positively correlated with the reliability R, that is, the higher the R, the greater the weight. Thus, calculate the overall load rate state L of the main production system system = ∑(wi×Li) and divide the load levels, where Li represents the load level of the power-consuming device numbered i, such as 1, 2, 3, 4.

[0020] When the overall load rate state L of the coal preparation plant system is on the low side and there are high-load or over-load production systems, analyze the load situation and reliability of the devices included in the system. If the reliability is within the credible range (such as above 70%) and the device load is high, the reasons and solutions for its high load (including high load and over load) can be analyzed in combination with the upstream and downstream relationships of the process and the trained online inference model.

[0021] In an optional implementation manner, the power-consuming device includes a thickening system, and the load level judgment of the thickening system includes:

[0022] Obtain the target data of the concentration system, where the target data includes the current solid content of the thickener feed, the turbidity of the thickener overflow, the torque of the thickener rake, the height of the slime layer in the thickener, the underflow concentration of the thickener, the model specifications of the thickener, the current number of operating filter presses, and the discharge amount of the filter presses. According to historical data, clean the target data, analyze the feature importance of the cleaned data, and use supervised learning algorithms to give the system load level.

[0023] In an alternative embodiment, the system-level digital twin model further includes a prediction model based on a Transformer-TCN hybrid architecture. The prediction model predicts the electrical energy required during the peak electricity period of the next working day according to the input data. The input data includes the power of power-consuming equipment, raw coal screening and float-sink data, sorting process timing data, the working temperature, humidity, vibration signals, and health assessment indicators of power-consuming equipment, production shifts, production targets, and equipment start-stop plans. The Transformer-TCN hybrid architecture includes a Transformer layer and a TCN layer. The Transformer layer is used to capture the long-term dependencies of load data, and the TCN layer extracts local temporal features through causal convolution and dilated convolution. The non-stationary load signal of the input data is decomposed by variational mode decomposition to extract modal components in different frequency bands, reduce noise interference, and enhance feature interpretability. A temporal feature matrix is generated through Z-score normalization and sliding window techniques. The model is trained using historical data, and transfer learning is combined to improve the generalization ability in small-sample scenarios. Adversarial training (GAN) is introduced to generate extreme operating condition data (such as equipment failures and coal quality mutations) to enhance the robustness of the model. When the deviation between the actual load and the prediction exceeds the threshold, the LSTM error correction module is triggered to adjust the prediction model parameters online. For example, when the coal quality mutation during the production process of the previous day causes a sudden increase in the load of the separator, the digital twin immediately generates simulation data, updates the prediction results, and adjusts the stored electrical energy for the next working day.

[0024] In an alternative embodiment, the hybrid energy storage system stores electrical energy during low electricity price periods and injects the excess electrical energy into the power grid during peak periods. The power grid power electronic interface further includes a virtual synchronous generator, which is used to simulate the rotor of the synchronous generator in the power grid system. The operating state of the power grid is obtained in real time, and it is judged whether the change in the power grid frequency reaches the threshold. When the set threshold is reached, the virtual synchronous generator simulates the inertial response of the synchronous generator in the power grid system, dynamically adjusts the inertia of the virtual generator to delay the frequency change, calculates the power of the virtual generator according to the frequency deviation and the adjusted inertia, and adjusts the power distribution of the virtual synchronous generator corresponding to the battery array according to the health status of the battery array in the hybrid energy storage system.

[0025] In a second aspect, a coal preparation plant energy storage method is provided, which adopts the digital twin-based coal preparation plant energy storage system described in the first aspect.

[0026] Beneficial effects: The whole-process twin calculation of the steam coal preparation plant can be carried out based on the digital twin technology; the efficiency bottleneck points can be determined by judging the loads of each system and equipment; through the whole-process efficiency control, the production efficiency can be comprehensively judged and improved, a new energy power energy storage system is established, and the best charge-discharge planning plan is given through the digital twin technology and the local peak-valley electricity price, so as to reduce the unit power consumption cost. Brief description of the drawings

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0028] Figure 1 A schematic diagram of an application scenario provided by an embodiment of the present application;

[0029] Figure 2 A schematic flow chart of a coal preparation plant energy storage method based on digital twin provided by an embodiment of the present application;

[0030] Figure 3 A block diagram of a coal preparation plant energy storage system with digital twin provided by an embodiment of the present application;

[0031] Figure 4 A schematic diagram of a device provided by an embodiment of the present application. Detailed implementation manners

[0032] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present application. Unless otherwise defined, the technical terms or scientific terms used in the present application should have the ordinary meaning understood by those of ordinary skill in the art to which the present invention pertains. The "first", "second" and similar terms used in the present application do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or items appearing before the term cover the elements or items listed after the term and their equivalents, without excluding other elements or items. The terms such as "connected", "coupled" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0033] The following is an explanation of the digital twin model. The digital twin model is a virtual model that comprehensively, dynamically, and real-time maps physical entities (such as devices, systems, processes, or environments) through digital technologies. Its core lies in using technologies such as the Internet of Things (IoT), big data, artificial intelligence (AI), cloud computing, and simulation technology to synchronize the operating status, behavioral characteristics, and environmental conditions of the physical world to the virtual space in real time, so as to achieve real-time monitoring, simulation prediction, optimization decision-making, and closed-loop control of physical entities. The essence of the digital twin model is to build a virtual mirror that "coexists" with the physical entity through virtual-real interaction and data-driven. It can not only reproduce the current state of the physical entity but also predict future behaviors through historical data accumulation and algorithm deduction, and even reverse-intervene in the operation of the physical entity to form a complete closed loop from perception, analysis, decision-making to execution. Its technical foundation includes multi-source heterogeneous data collection, high-precision modeling and simulation, real-time data synchronization, machine learning, and optimization algorithms. The specific implementation is usually divided into three levels: geometric twin (describing the physical structure), behavior twin (simulating the dynamic process), and rule twin (embedding business logic). In terms of the technical architecture, the digital twin model consists of a physical entity layer, a data transmission layer, a virtual model layer, and a functional application layer. The physical entity layer uses devices such as sensors, actuators, cameras, and RFID tags to collect data such as temperature, pressure, vibration, position, and energy consumption in real time; the data transmission layer relies on technologies such as 5G, industrial Ethernet, and edge computing to transmit the collected data to the cloud or local server at high speed and low latency, while ensuring data security and integrity; the virtual model layer is the core of the digital twin, building a dynamic virtual entity through three-dimensional modeling (such as CAD), physical equation simulation (such as finite element analysis), and data-driven models (such as neural networks); the functional application layer provides services such as status monitoring, fault diagnosis, performance optimization, and predictive maintenance based on the virtual model.

[0034] The application of the digital twin model in coal preparation plants is mainly reflected in the realization of real-time monitoring, intelligent optimization, and predictive maintenance of the entire coal preparation process by constructing a virtual mirror system that is highly synchronized with actual physical equipment, process flows, and production environments. In coal preparation plants, the digital twin model first collects real-time equipment operation data (such as motor speed, bearing temperature, separation density, slurry concentration, throughput, energy consumption, etc.) through sensors (such as vibration sensors, temperature sensors, pressure sensors, current sensors) deployed on key equipment such as crushers, dense medium separators, centrifuge dewaterers, vibrating screens, and belt conveyors, and combines with coal quality parameters such as screening and floating-sinking data, ash content, and sulfur content of raw coal, and transmits them to the cloud or local digital twin platform through Internet of Things (IoT) technology. The platform uses three-dimensional modeling technology (such as BIM or CAD) to construct geometric models of coal preparation equipment, and combines mechanism models (such as density control equations for dense medium separation, hydrodynamic models for centrifuge dewaterers) with data-driven models (such as LSTM neural networks to predict equipment failures, random forest algorithms to optimize separation parameters) to dynamically simulate the operation status of equipment under different working conditions. For example, in the dense medium separation link, the digital twin model can analyze the matching relationship between the density distribution of raw coal and the medium density in real time, dynamically adjust the set value of the suspension density, predict the separation efficiency and clean coal yield rate based on historical data, verify the feasibility of parameter adjustment through virtual commissioning, and then feedback the optimization instructions to the PLC control system to achieve closed-loop control of the separation density, increasing the clean coal recovery rate by 3%-5%. In terms of equipment health management, the digital twin model predicts potential failures such as crusher hammer wear and centrifuge bearing aging by analyzing the time-frequency characteristics of vibration signals (such as extracting high-frequency fault components through wavelet packet decomposition) and current harmonic characteristics, and generates maintenance suggestions (such as replacing the screen mesh 2 weeks in advance), reducing the unplanned downtime by 20%-30%. In addition, aiming at the problem of high energy consumption in coal preparation plants, the digital twin model can simulate the equipment start-stop combination and load distribution strategy under different production shifts, and dynamically optimize the operation frequency of crushers and the power of pump equipment in combination with time-of-use electricity prices and the state of charge (SOC) of the energy storage system, reducing the electricity consumption per ton of coal by 8%-12%. In safety control, the digital twin model constructs a three-dimensional panoramic map of the coal preparation workshop through lidar and video fusion technology, and real-time tracks the positions of personnel and the operation status of equipment. When belt deviation, coal bunker level overlimit, or personnel approaching dangerous areas are detected, it automatically triggers an audible and visual alarm and links to the emergency stop system. At the same time, the digital twin supports cross-system collaboration. For example, when the raw coal washing volume suddenly increases, the model can simulate the load matching relationship between the dense medium separation system and the coarse coal slime recovery system, and dynamically adjust the inlet pressure of the hydrocyclone and the excitation current of the magnetic separator to avoid system overload.Through digital twin technology, the coal preparation plant realizes the transformation from "experience-driven" to "data-driven", with the comprehensive production efficiency increased by 15%-20% and the equipment utilization rate increased to over 90%. At the same time, it provides a virtual test environment for process innovation (such as intelligent dry separation replacing wet separation), significantly reducing the technical transformation cost and risk.

[0035] The method for constructing the energy storage of a coal preparation plant based on digital twins provided by the embodiments of this application can be applied to an analysis and decision-making system as shown in Figure 1 The analysis and decision-making system includes a transmission layer, a perception layer, an algorithm layer, and an application layer. The production efficiency bottlenecks can be analyzed and solved by means of digital twin real-time simulation and simulation technology, the load status of systems and equipment, the load balancing of multi-system material distribution, and the load judgment of the thickening system, etc., so as to improve the raw coal processing volume per unit time of the coal preparation plant, increase the equipment utilization rate, and reduce the power consumption per ton of coal and the unit power consumption cost. The "prediction-adjustment-validation" closed-loop and precise control of the whole-plant production efficiency of the thermal coal preparation plant is realized.

[0036] Among them, the data layer obtains the underlying real-time data through various basic sensors such as quality, water volume, flow rate, concentration, gate opening, and equipment motor current, constructs a full-process data lake, and realizes the real-time processing of high-frequency data in combination with a time series database (InfluxDB). The perception layer includes distributed edge nodes and multi-device cooperation protocols. Distributed edge nodes deploy edge computing modules on key equipment (such as dense medium shallow trough separators), and run lightweight AI models locally to achieve low-latency real-time control (such as fine-tuning the addition amount of heavy medium); the multi-device cooperation protocol is based on the communication protocol of the industrial Internet of Things (IIoT) to realize the autonomous cooperation between devices.

[0037] Digital Twin Calculation Model: Based on the process calculation rules of the coal preparation industry, online and dynamic calculations of the process system flow of coal preparation production are carried out. According to real-time coal quality fluctuations (such as ash content and calorific value), the raw coal screening and sinking data are automatically adjusted. The Digital Twin Calculation Model will perform simulation operations based on the latest data. Further, according to the machine learning neural network algorithm, the sinking data of each separation system obtained by the Digital Twin Calculation Model is used to fit the washability curve, so as to predict the best washing parameters (such as the density of heavy medium and the feed pressure of the coarse slime separator), and the decision-making target is sent to the corresponding execution system (such as intelligent density control and intelligent coarse slime separation system). The execution system is equipped with distributed edge node technology to achieve low-latency real-time control and reduce manual intervention. The LSTM neural network and the Digital Twin quantity and quality process calculation model are used to predict the quality of the washed clean coal (such as clean coal ash content and pulverized coal calorific value) in real time, and the XGBoost algorithm is combined to analyze the key influencing factors (such as the particle size distribution and density distribution characteristics of raw coal), and the process parameters are adjusted in advance. When the predicted quality deviates from the target value, the system automatically triggers a compensation strategy. For example, when the clean coal ash content exceeds the standard, the density of the shallow tank separator is adjusted to achieve a closed loop of "prediction - adjustment - verification". When the predicted quality deviates from the target value, the system automatically triggers a compensation strategy. The application layer is used for load condition analysis of the system and power-consuming equipment, concentration system load judgment, and power energy storage distribution.

[0038] As Figure 3 shown, the coal preparation plant energy storage system based on Digital Twin in the embodiment of the present application includes a hybrid energy storage system, a battery management system, a high-voltage DC protection system, a pre-charge / discharge circuit, a high-voltage AC power distribution system, a centralized PCS cluster, and a grid power electronics interface. The battery energy storage array is connected to the power-consuming equipment in the coal preparation plant, and the total capacity of its hybrid energy storage system is determined by the above method. The hybrid energy storage system stores electrical energy and provides a DC power supply. The battery management system (BMS) monitors the battery state (SOC / SOH), performs equalization management, thermal management, and fault protection. The high-voltage DC protection system provides overvoltage, overcurrent, and short-circuit protection to ensure the safety of the DC side; the pre-charge / discharge circuit limits the inrush current during system startup and releases the residual energy during shutdown; the centralized PCS cluster provides DC-AC bidirectional conversion, grid-connected / off-grid mode switching, and power regulation; the high-voltage AC power distribution system distributes AC electrical energy and integrates protection devices (circuit breakers, fuses). The grid power electronics interface needs to meet grid connection standards (voltage, frequency, harmonics) to achieve bidirectional power flow.

[0039] As Figure 2 shown, the coal preparation plant energy storage method based on Digital Twin includes the following steps.

[0040] Deploy sensors to obtain the dynamic data of power-consuming equipment in the coal preparation plant. The dynamic data includes electrical parameters, operating parameters, and environmental parameters under various coal quality processes. The electrical parameters are used to collect current and voltage waveforms through an intelligent electricity meter at intervals of 1 second to 1 minute. The vibration / temperature parameters are used to monitor the mechanical state of the equipment through an industrial-grade vibration sensor at a sampling rate of 10 kHz. An edge gateway (such as Huawei Atlas500) can be deployed at the equipment end to filter, denoise, and compress the original data. For example, perform FFT analysis on the current signal to extract fundamental and harmonic components. The high-frequency vibration / current waveforms can be transmitted to the cloud or local server in real time through OPCUA and MQTT communication protocols. The coal quality data can be wirelessly transmitted within the workshop through ModbusTCP and LoRaWAN communication protocols.

[0041] The deployment of sensors is shown in the following table.

[0042]

[0043] Obtain the production plan of the coal preparation plant, and the obtained parameters are as follows:

[0044] Coal quality parameters: ash content, moisture content, sulfur content, particle size distribution.

[0045] Process parameters: crushing particle size, separation density, dehydration time.

[0046] Production plan: batch quantity, time window, priority.

[0047] Determine the power-consuming equipment required for the coal quality and process flow under the current production plan. The coal quality-process-equipment mapping rules are shown in the following table.

[0048]

[0049] Construct an equipment-level digital twin model based on the dynamic data of power-consuming equipment to simulate and measure the working state of power-consuming equipment; the equipment-level architecture includes a geometric model and a physical model; the geometric model includes the 3D appearance and structural topology of power-consuming equipment; the physical model includes the mathematical equations of motor power and heat conduction equations based on physical laws.

[0050] For geometric model modeling, software such as SolidWorks / Blender can be used to establish a 3D appearance model of the equipment (such as crushers, centrifugal pumps), including the topological structures of key components such as bearings and gearboxes. It is also possible to calibrate the geometric dimensions through laser scanning reverse engineering, with an error <0.1 mm. Synchronize the equipment maintenance records (such as bearing replacement) to the 3D model to update the topological relationship of components.

[0051] The motor power model is the core module of the digital twin physical model. It is necessary to combine the characteristics of the motor body and real-time operation data to achieve accurate energy consumption modeling. The electrical parameters collected by the motor power model include voltage U (V) and current I (A), which are collected in real time by an intelligent electricity meter. The mechanical parameters include rotational speed n (rpm), which is analyzed by an encoder or FFT of a vibration sensor, and the friction coefficient k friction (N·m·s / rad²) is calibrated through no-load experiments, and the motor efficiency η is obtained through manufacturer's nameplate data or IEEE112 standard tests. According to the collected parameters, electrical power calculation, mechanical power correction, harmonic loss compensation, and temperature drift compensation are carried out to generate the power model of the motor.

[0052] Couple the device-level digital twin models to generate a system-level digital twin model for collaborative simulation between devices. Taking the crushing-separation system of a coal preparation plant as an example, in the energy flow, couple the motor power of the crusher with the load torque of the centrifugal pump, and in the material flow, couple the particle size distribution of the crushed coal with the processing efficiency of the centrifugal pump. In the signal flow, couple the fault signal of the vibrating screen with the feeding rate of the feeder. The dynamic power balance equation for energy flow coupling:

[0053] ∑ Pinput =∑(P motor +P loss )+P conveyor +P auxiliary

[0054] Among them, ∑ Pinput is the total input power of the system, which is the total electrical power obtained by the entire system from the power grid or other energy sources. ∑(P motor +P loss ) is the power and loss of the main equipment, which is the total energy consumption of all drive motors (such as crushers and centrifugal pumps), including the effective output power and internal losses. The input power of the crusher motor = output mechanical power (driving the rotor) + stator copper loss + bearing friction loss. P conveyor is the power of the conveyor system, which is the total power consumed by the conveyor operation (including the energy consumption of the drive motor and mechanical losses). The input power of the belt conveyor motor + the friction loss between the belt and the roller. P auxiliary is the power of the auxiliary system, which is the energy consumption of non-direct production equipment, such as control systems, sensors, lighting, etc., and can be the electrical power of the PLC control cabinet, dust monitor, and workshop lighting system.

[0055] For the implementation of co-simulation, time synchronization management can be achieved through hardware-level synchronization. For example, the PTP protocol can be used to align the simulation clocks of device models, with an error < 1ms. It can also be achieved through software-level synchronization by defining a global simulation step size (such as 10ms) in the coordinator to trigger iterative calculations of device models. The data interaction protocol for real-time sensor data can adopt the MQTT protocol, and the data interaction protocol for model state parameters can be implemented through the OPCUA protocol.

[0056] According to the production scheduling plan, electricity prices at different local times, and the charging and discharging costs of the hybrid energy storage system, the start-stop combinations, working periods of power-consuming devices, and the power of the motors of power-consuming devices are simulated through the system-level digital twin model to find the production route with the lowest total electricity price for a single working day. The constraint conditions include the processing volume and start-stop frequency of all power-consuming devices for a single working day.

[0057] The method for finding the production route with the lowest total electricity price for a single working day can be achieved through the following method:

[0058] Divide the low-price period and high-price period of electricity into T and T' periods respectively. For example, T = 8 (0:00 - 8:00), T' = 4 (18:00 - 22:00), and the duration of each period is , = 1 hour, j represents the jth period in the low-price period, k represents the kth period in the high-price period; after numbering the power-consuming devices, they form a set, and m is the total number of devices; let the power of the power-consuming device numbered i in the jth period be ,let the power of the power-consuming device numbered i in the kth period be 。

[0059] , Characterize whether the power-consuming device numbered i works in the jth period. When it means that the power-consuming device i works in the jth period during the low-price period of electricity, and when it means that the power-consuming device i does not work; , Characterize whether the power-consuming device numbered i works in the kth period. When it means that the power-consuming device i works in the kth period during the high-price period of electricity, and when it means that the power-consuming device i does not work;

[0060] Let the low electricity price be C and the power supply cost price of the energy storage system be d; the total daily electricity cost C-low during the low-price period:

[0061]

[0062] The total daily electricity cost C-high during the high-price period:

[0063]

[0064] Establish the objective function \(C_{total}=C_{low}+C_{high}\), where \(C_{total}\) is the total electricity cost, and solve it according to the genetic algorithm. 、 and the combination of to minimize the total electricity cost; determine the on or off state of the power-consuming equipment at each time period through and the value.

[0065] Directly use the total cost \(C_{total}\) as the fitness value, the smaller the better. Add the constraint violation amount to the fitness function. For example, if the processing capacity of equipment 1 is insufficient, add a penalty term. Solve it through the genetic algorithm or particle swarm optimization. The solution steps are as follows: Initialize the population: randomly generate the start-stop state and power value of the equipment. Calculate the fitness. Gradually optimize through selection, crossover, and mutation until the processing capacity is satisfied and the cost is the lowest.

[0066] Optimal solution example

[0067] Low valley period: The equipment operates at full power \( = 50kW\times8\) hours, and the processing capacity is \(0.1\times50\times8 = 40\) tons.

[0068] Peak period: The equipment operates \( = 50kW\times4\) hours, and the processing capacity is \(0.1\times50\times4 = 20\) tons, with a total processing capacity of 60 tons.

[0069] Total cost: \(0.3\times50\times8 + 0.5\times50\times4 = 120 + 100 = 220\) yuan. The cost of the low valley motor is 0.3, and the cost price of the energy storage system power supply is \(d = 0.5\).

[0070] It is also possible to add the following constraint conditions to solve the objective function \(C_{total}=C_{low}+C_{high}\). Set a threshold for the overall load factor state Lsystem value of the power-consuming equipment, \(L\) system The calculation process of the value includes:

[0071] Based on the real-time simulation operation ability of the quantity and quality process in the digital twin model and the screening and sinking data in the real-time coal quality database, obtain the real-time processing capacity of each device and system, get the optimal raw coal processing capacity of the current system, and calculate the production efficiency of the current coal preparation plant in combination with the real-time raw coal washing volume (real-time raw coal washing volume / optimal raw coal processing capacity). Analyze and calculate the load rate of each device, divide the load levels (low load, optimal load, high load, overloading, and can be assigned values for calculation respectively, such as 1, 2, 3, 4). Further, according to the situation of the devices included in the main production systems (raw coal preparation system, heavy medium separation system, coarse coal slime recovery system, product transportation system, etc.) in each interval (low load, optimal load, high load, overloading), determine the probability distribution model it follows based on the historical load status of the device, collect the failure data and load abnormal data of the device within a certain period of time, such as the number of failures, failure time, average load rate value in the abnormal load interval, abnormal load duration, etc., and calculate the reliability R of a single device, R = R0(1 - F); where R0 is the basic credibility, R0 = e -λt , λ = number of failures / unit operation time, t is the device life; F represents the reliability attenuation factor caused by abnormal load, F = min(△L×p×k, 1), k is the adjustment coefficient, which needs to be set according to experience, △L = (average load rate - rated load rate) / rated load rate, p = abnormal load duration / mean time between failures; the obtained R value can automatically adjust the distribution weight wi of each device, assign the weight wi to each device, and the distribution weight wi is positively correlated with the reliability R, that is, the higher the R, the greater the weight. Thus, calculate the overall load rate status L of the main production system system = ∑(wi×Li) and divide the load level, Li represents the load level of the power-consuming device numbered i, such as 1, 2, 3, 4.

[0072] When the overall load rate status L of the coal preparation plant system is relatively low and there are high-load or overloaded production systems, analyze the load situation and reliability of the devices included in the system. If the reliability is within the credible range (such as above 70%) and the device load is relatively high, the reasons and solutions for its high load (including high load and overloading) can be analyzed in combination with the upstream and downstream relationships of the process and the trained online inference model.

[0073] Reliability not only reflects the stability of equipment performance but also directly affects production efficiency, operation and maintenance costs, and energy consumption optimization. Reliability converts the operating data of equipment (number of failures, load rate, abnormal time) into probability values, intuitively reflecting the current health status of the equipment. Equipment with high reliability (R>0.8) operates stably and is suitable for full-load or over-load operation. Low reliability (R<0.6): The equipment has potential failure risks and needs to reduce load or shut down for maintenance. By calculating reliability in real time, high-risk equipment can be identified in advance. As a weighting factor (wi = Ri / ∑Rj, i∈j), reliability dynamically adjusts the load distribution of equipment in the system. Equipment with high reliability is assigned more tasks to improve utilization.

[0074] Low-reliability equipment reduces load or switches to standby equipment to avoid cascading failures. High load + high reliability allows short-term over-load operation to increase throughput. High load + low reliability, on the other hand, forces load reduction to avoid failures. By quantifying the equipment health status, guiding load distribution, and warning of potential failures, reliability directly supports the improvement of production efficiency and cost optimization. In complex industrial scenarios such as coal preparation plants, combined with digital twin models and real-time data, reliability calculation can significantly improve the system robustness, realize the transformation from "repair after failure" to "predictive maintenance", and ultimately reduce energy consumption and processing cost per ton of coal.

[0075] In a coal preparation plant, the load refers to the workload or power demand that equipment or a system undertakes within a certain period, directly affecting production efficiency, energy consumption, and equipment life. The equipment load refers to the actual workload of a single piece of equipment during operation, usually expressed in throughput (such as tons per hour) or power (such as kW). The load of a crusher can be defined as the number of tons of raw coal processed per hour or the power consumption when the motor is running. The system load is the total workload of the entire production system (such as the dense medium separation system), which is composed of the superposition of the loads of multiple pieces of equipment.

[0076] In a coal preparation plant, the load is usually divided into four categories according to the ratio of the actual workload to the rated capacity, as shown in the following table:

[0077]

[0078] High / over-load, mechanical wear intensifies (such as bearing overheating, belt breakage). Exceeding the motor power limit causes insulation aging, the failure rate (λ) increases, and the reliability (R) decreases. Low-load equipment starts and stops frequently or runs idly, increasing ineffective energy consumption and shortening the life cycle.

[0079] For the determination of the load level, an algorithm model can also be introduced for automatic judgment. For the thickening system, the load level judgment of the thickening system includes:

[0080] Obtain the target data of the concentration system, where the target data includes the current solid content of the thickener feed, the turbidity of the thickener overflow, the torque of the thickener rake, the height of the slime layer in the thickener, the underflow concentration of the thickener, the model specifications of the thickener, the current number of operating filter presses, and the discharge amount of the filter presses.

[0081] Based on historical data, perform data cleaning on the target data. Steps for data cleaning: handling missing values, deleting records with a missing rate exceeding 30%, and filling the remaining missing values using linear interpolation or KNN interpolation; detecting outliers, identifying outliers using box plots or the 3σ principle, and correcting or removing abnormal data in combination with process knowledge (e.g., a 10-fold sudden increase in rake torque may be due to a sensor failure); data standardization, perform feature importance analysis on the cleaned data, and use supervised learning algorithms to give the system load level. Perform Z-score standardization on continuous variables (such as turbidity and torque), and perform one-hot encoding on categorical variables (thickener models).

[0082] Conduct feature importance analysis, retain key features (such as underflow concentration, rake torque, and slime layer height), and eliminate redundant features (such as model specifications, which may have less impact due to equipment standardization).

[0083] Construct a supervised learning model. Gradient boosting trees are suitable for high-dimensional data and non-linear relationships and support classification tasks. Divide the data into a training set and a test set at a ratio of 7:3. Use grid search to optimize hyperparameters (such as learning rate and tree depth), and 5-fold cross-validation to prevent overfitting.

[0084] The load levels are defined as shown in the following table:

[0085]

[0086] According to the optimal production route, determine the start and stop times of power-consuming equipment and the power of the motors. Based on the working hours of the power-consuming equipment and the power of the motors, determine the total peak power consumption W during a working day.

[0087]

[0088] The energy storage capacity E of the hybrid energy storage system is E = W×a / (η×DOD), where DOD is the depth of discharge. For lead-acid batteries, DOD = 0.5, and for lithium batteries, DOD = 0.8. η is the charge-discharge efficiency, and the charging efficiency ηc and discharge efficiency ηd need to be considered. Generally, η = ηc×ηd. a is the redundancy coefficient, usually a≥1.2, to reserve additional capacity to cope with prediction errors or sudden demands.

[0089] Predict the peak load period at the start and stop moments of the power-consuming device according to its working period. The peak starting current of the motor can reach 5 to 7 times the rated value. The battery management system turns on the supercapacitor array during the peak load period to perform high-frequency power compensation for the power-consuming device, and the lithium battery array supplies power to the power-consuming device during the periods other than the peak load of the peak electricity. Through peak load prediction and coordinated control of hybrid energy storage, the peak load pressure on the lithium battery can be significantly reduced, its service life can be extended, and the dynamic response performance of the system can be improved. Using supercapacitors to compensate for instantaneous high-frequency demands and combining lithium batteries to provide continuous power supply forms an energy management strategy with both economy and reliability.

[0090] In some embodiments of the present application, the device-level digital twin model can also be used to optimize the motor frequency of a single device.

[0091] Establish the correlation between the throughput and frequency of a single power-consuming device. The throughput is usually linearly related to the frequency (such as the conveyor belt speed). Input parameters: the minimum throughput of the production plan, the maximum throughput of the device, the rated frequency and power, and the inverter efficiency table. Calculate the minimum frequency according to the minimum throughput and determine the frequency scanning range. Establish an objective function for optimizing unit energy consumption and solve for the optimal motor frequency.

[0092] In some embodiments of the present application, the system-level digital twin model further includes a prediction model based on a Transformer-TCN hybrid architecture. The prediction model predicts the electric energy required during the peak electricity period of the next working day according to the input data. The input data includes the power of the power-consuming device, the raw coal screening and sinking data, the sorting process timing data, the working temperature and humidity of the power-consuming device, the vibration signal, the health assessment index, the production shift, the output target, and the device start-stop plan. The Transformer-TCN hybrid architecture includes a Transformer layer and a TCN layer. The Transformer layer is used to capture the long-term dependencies of the load data, and the TCN layer extracts local temporal features through causal convolution and dilated convolution. The non-stationary load signal of the input data is decomposed by variational mode decomposition to extract modal components in different frequency bands, reduce noise interference, and enhance the interpretability of features. A temporal feature matrix is generated through Z-score normalization and sliding window technology. The model is trained using historical data, and transfer learning is combined to improve the generalization ability in small-sample scenarios. Adversarial training (GAN) is introduced to generate extreme working condition data (such as equipment failures and coal quality mutations) to enhance the robustness of the model. When the deviation between the actual load and the prediction exceeds the threshold, the LSTM error correction module is triggered to adjust the parameters of the prediction model online. For example, when the coal quality mutation during the production process of the previous day causes a sudden increase in the load of the separator, the digital twin immediately generates simulation data, updates the prediction result, and adjusts the stored electric energy for the next working day.

[0093] The prediction model based on the Transformer-TCN hybrid architecture realizes high-precision peak electricity demand prediction by integrating long-term dependencies and local time-series features, and combining VMD signal decomposition with dynamic feature engineering. The model deeply integrates device health and process data, providing a full-chain solution from electricity prediction to scheduling optimization for coal preparation plants, significantly reducing energy consumption and operating costs, and promoting intelligent upgrading.

[0094] The hybrid energy storage system stores electrical energy during low electricity price periods and releases excess electrical energy into the power grid during peak periods; the grid power electronic interface also includes a virtual synchronous generator, which is used to simulate the rotor of the synchronous generator in the power grid system. The operating state of the power grid is obtained in real time, and it is judged whether the change in the power grid frequency reaches a threshold. When the set threshold is reached, the virtual synchronous generator simulates the inertial response of the synchronous generator in the power grid system, dynamically adjusts the inertia of the virtual generator to delay the frequency change, calculates the power of the virtual generator according to the frequency deviation and the adjusted inertia, and adjusts the power distribution of the virtual synchronous generator corresponding to the battery array according to the health status of the battery array in the hybrid energy storage system.

[0095] The dynamic variable inertia VSG control solves the problem that fixed inertia cannot balance fast response and stability by adjusting the virtual inertia J(t) in real time. According to the frequency deviation Δf(t) and its change rate dΔf / dt, the inertia is adjusted in two stages:

[0096]

[0097] Among them, is the minimum inertia (typical value ), which is used to ensure the basic inertia, is the maximum inertia (typical value ), which is used to limit the response delay, is the adjustment coefficient (calibrated through HIL tests, recommended ), example: when is detected and is detected, .

[0098] Dynamic variable inertia VSG control is a control strategy used to improve the stability of the power grid system. It monitors the change of the power grid frequency in real time and dynamically adjusts the inertia value of the system according to the deviation of the power grid frequency and its rate of change. This control method can solve the problem that a fixed inertia value cannot ensure both fast response and system stability at the same time. Glossary: 1. Dynamic variable inertia: It refers to dynamically adjusting the inertia value according to real-time data during the operation of the system. Inertia is a physical quantity that measures the resistance of a system to changes in speed. In the power system, the magnitude of inertia affects the response speed and stability of the system. 2. VSG control: The abbreviation of Virtual Synchronous Generator Control, which is a control strategy used to simulate the behavior of traditional synchronous generators to improve the stability and controllability of distributed generation systems. 3. Power grid frequency deviation: It refers to the difference between the actual operating frequency of the power grid and the standard frequency (such as 50Hz or 60Hz). An excessive power grid frequency deviation may affect the stable operation of the power system. 4. Inertia value J(t): In dynamic variable inertia VSG control, the inertia value J(t) is calculated in real time according to the power grid frequency deviation and changes with the operating state of the power grid. 5. Regulation coefficient The regulation coefficients are used to control the calculation method of the inertia value J(t), and they are calibrated through experiments or simulations (HIL tests) to achieve the best control effect.

[0099] For the small deviation stage (≤0.5Hz), gradually increase the inertia to suppress frequency fluctuations and improve stability. For the large deviation stage (>0.5Hz), dynamically reduce the inertia through the rate of change to accelerate the power response and avoid frequency collapse. It is simple and easy to implement, only requiring the frequency deviation and its rate of change as inputs, with low computational complexity and suitable for real-time control.

[0100] Deep charge and discharge can increase revenue but damage battery life. The State of Charge (SOC) and State of Health (SOH) of a battery are two important parameters for evaluating battery performance and life. By regularly detecting the SOH of the battery and adjusting the SOC usage range according to the detection results, the battery life and revenue can be optimized. Specifically, if the SOH of the battery is below 80%, the depth of discharge of the battery needs to be reduced, that is, a higher minimum SOC value is set to protect the battery. Glossary: 1. Battery SOC: SOC refers to the current charge state of the battery, usually expressed as a percentage. It reflects the remaining charge of the battery and is one of the key parameters in the battery management system. 2. Battery SOH: SOH refers to the health state of the battery, which reflects the performance and aging degree of the battery. A decrease in SOH means a reduction in battery capacity and a shortening of battery life. 3. Deep charge and discharge: It refers to the battery being charged and discharged to near full charge or fully discharged. Although deep charge and discharge can increase the usage revenue of the battery in the short term, it will accelerate battery aging and reduce battery life in the long term.

[0101] Harmonic suppression is achieved through a virtual resistor. The specific steps are to measure the system resonance point offline (such as FFT analysis), design a virtual resistor transfer function with band-pass characteristics, and inject the virtual resistor transfer function into the current loop feed-forward of the VSG. Harmonics are suppressed by increasing the damping, but this will reduce the efficiency of the system. It includes several key steps: first, measure the resonance point of the system, then design a specific virtual resistor transfer function, and finally inject this function into the current loop of the Voltage Source Inverter (VSG). The transfer function is as follows:

[0102]

[0103] Among them, Rvirtual(s) represents the dynamic impedance value of the virtual resistor in the complex frequency domain, is the target resonance frequency (such as 250 Hz), ζ is the damping ratio (recommended 0.7 - 1.0), K is the gain coefficient (determined by sweep frequency test), and s is the complex frequency variable. Associated keyword analysis: 1. Virtual resistor: A virtual resistor is a way to simulate the resistance characteristics in a power electronics system, used to control or suppress specific electrical phenomena such as harmonics. 2. Harmonic suppression: Harmonic suppression refers to the technology of reducing or eliminating the distortion of current or voltage waveforms in a power system caused by nonlinear loads. 3. FFT analysis: Fast Fourier Transform (FFT) is an algorithm used to transform a signal from the time domain to the frequency domain to analyze the frequency components in the signal. 4. Band-pass characteristic: The band-pass characteristic means that a system or device has good transmission ability for signals within a certain frequency range, while having significant attenuation for signals outside this range. 5. VSG current loop feed-forward injection: VSG refers to the virtual synchronous generator, and current loop feed-forward injection means directly adding a control signal to the current control loop to improve the dynamic response and stability of the system.

[0104] Based on the same inventive concept, an embodiment of the present application also discloses a coal plant energy storage method, which uses the above-mentioned digital twin-based coal preparation plant energy storage system.

[0105] Corresponding to the above method, an embodiment of the present disclosure also provides an electronic device, such as Figure 4 shown, which is a schematic structural diagram of the electronic device 400 provided by an embodiment of the present disclosure, including: a processor 41, a memory 42, and a bus 43; the memory 42 is used to store execution instructions, including an internal memory 421 and an external memory 422; here, the internal memory 421 is also called the main memory, which is used to temporarily store the operation data in the processor 41 and the data exchanged with the external memory 422 such as a hard disk. The processor 41 exchanges data with the external memory 422 through the internal memory 421. When the electronic device 400 runs, the processor 41 communicates with the memory 42 through the bus 43, so that the processor 41 executes Figure 1 the steps of the coal plant energy storage method. The communication interface is used for communication between the above-mentioned electronic device and other devices.

[0106] The memory may include a Random Access Memory (RAM), or may also include a Non-Volatile Memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far from the aforementioned processor.

[0107] Embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the energy storage method described in the above method embodiments. Among them, the storage medium may be a volatile or non-volatile computer-readable storage medium.

[0108] Embodiments of the present disclosure also provide a computer program product, which includes computer instructions. When the computer instructions are executed by a processor, they can execute the steps of the UAV intelligent airport deployment method described in the above method embodiments. For details, refer to the above method embodiments and will not be elaborated here.

[0109] Those skilled in the art should understand that the embodiments in the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the embodiments in the embodiments of the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments in the embodiments of the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) that contain computer-usable program codes.

[0110] The embodiments in the embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0111] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device realizes the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the functions specified in the process Figure 1 one process or a plurality of processes and / or blocks Figure 1 steps for implementing the functions specified in one block or a plurality of blocks.

Claims

1. A coal preparation plant energy storage system based on digital twin, comprising a hybrid energy storage system, a battery management system, a high-voltage DC protection system, a pre-charge / discharge circuit, a high-voltage AC power distribution system, a centralized PCS cluster, and a grid power electronic interface. The hybrid energy storage system is connected to the power-consuming equipment in the coal preparation plant, and it includes a lithium battery array and a supercapacitor array. It is characterized in that, It further includes: Constructing a device-level digital twin model based on the dynamic data of power-consuming devices and performing coupling to generate a system-level digital twin model, where the dynamic data includes electrical parameters, operating parameters, and environmental parameters of the power-consuming devices under various coal quality processes; According to the production plan, electricity prices at different local times, and the charge and discharge costs of the hybrid energy storage system, simulating and emulating the start-stop combination, working hours of the power-consuming devices, and the power of the motors of the power-consuming devices through the system-level digital twin model to find the production route with the lowest total electricity price for a single working day, and the constraint conditions include the processing volume and start-stop frequency of all power-consuming devices for a single working day; Determining the start-stop time period of the power-consuming devices and the power of the motors according to the optimal production route; Determining the energy storage capacity of the hybrid energy storage system according to the start-stop time period of the power-consuming devices and the power of the motors; predicting the peak load time period at the moment of start-stop of the power-consuming devices according to the start-stop time period of the power-consuming devices, and the battery management system turning on the supercapacitor array to perform high-frequency power compensation for the power-consuming devices during the peak load time period, and the lithium battery array supplying power to the power-consuming devices during the time period outside the peak electricity peak load.

2. The coal preparation plant energy storage system based on digital twin according to claim 1, wherein: The method for finding the production route with the lowest total electricity price for a single working day includes: The time periods of low electricity price and high electricity price are divided into T and T' time periods respectively, and the duration of each time period is , where j represents the j-th time period in the low electricity price period and k represents the k-th time period in the high electricity price period; Numbering the power-consuming devices to form a set, and m is the total number of devices; Let the power of the power-consuming device numbered i in the j-th period be expressed as , and let the power of the power-consuming device numbered i in the k-th period be expressed as , , characterizes whether the power-consuming device numbered i works in the j time period. When , it means that the power-consuming device i works in the j time period during the low electricity price period. When , it means that the power-consuming device i does not work; , characterizes whether the power-consuming device numbered i works in the k time period. When , it means that the power-consuming device i works in the k time period during the high electricity price period. When , it means that the power-consuming device i does not work; Let the low valley electricity price be C and the power supply cost price of the energy storage system be d; the total daily electricity cost C-low during the low valley period: The total electricity cost C-high during the peak period: Establish the objective function C-total = C-low + C-high, where C-total is the total electricity cost, and solve it according to the genetic algorithm , and combinations to minimize the total electricity cost; determine the on or off state of the power-consuming equipment at each time period through and values 3. The coal preparation plant energy storage system based on digital twin according to claim 2, characterized in that: The calculation method for the energy storage capacity of the hybrid energy storage system includes: Determining the start-stop time period of the power-consuming devices and the power of the motors according to the optimal production route, and determining the total electricity consumption W during the peak electricity period of a working day according to the working time period and motor power of the power-consuming devices; The energy storage capacity E of the hybrid energy storage system = W×a / (η×DOD), where DOD is the depth of discharge, η is the charge and discharge efficiency, and a is the redundancy factor.

4. The coal preparation plant energy storage system based on digital twin according to claim 1, wherein: The constraint condition further includes the overall load rate state of the power-consuming device Lsystem value reaches the set threshold, and the L system value calculation process includes: Obtain the load rate situation of power-consuming devices, divide the load levels, based on the load levels of power-consuming devices, determine the probability distribution model that the load of power-consuming devices follows according to the historical load status of the devices, determine the reliability R of the power-consuming devices. The reliability R is the probability that the power-consuming devices can complete the expected functions without failure under specific conditions and within a specified time. Adjust the load distribution weight wi of each power-consuming device according to the obtained R value, and calculate the overall load rate status of all power-consuming devices Lsystem =∑(wi × Li), where Li represents the load level of the power-consuming device numbered i.

5. The coal preparation plant energy storage system based on digital twin according to claim 4, wherein: The reliability R = R0(1 - F), where R0 is the basic credibility, and R0 = e -λt , λ = number of failures / unit operation time, and t is the equipment life; F represents the reliability decay factor caused by abnormal load, F = min(△L×p×k, 1), k is the adjustment coefficient, △L = (average load rate - rated load rate) / rated load rate, p = abnormal load duration / mean time between failures.

6. The coal preparation plant energy storage system based on digital twin according to claim 5, wherein: The allocated weight wi is positively correlated with the reliability R, that is, the higher the R, the greater the weight.

7. The coal preparation plant energy storage system based on digital twin according to claim 4, wherein: The power-consuming devices include a thickening system, and the load level judgment of the thickening system includes: Obtaining the target data of the thickening system, where the target data includes the current solid content of the feed of the thickener, the turbidity of the overflow of the thickener, the torque of the rake of the thickener, the height of the coal slime layer of the thickener, the bottom flow concentration of the thickener, the model specifications of the thickener, the current number of opened filter presses, and the discharge amount of the filter presses; According to historical data, cleaning the target data, performing feature importance analysis on the cleaned data, and using a supervised learning algorithm to give the system load level.

8. The coal preparation plant energy storage system based on digital twin according to claim 1, characterized in that: The system-level digital twin model further includes a prediction model based on a Transformer-TCN hybrid architecture, and the prediction model predicts the electric energy required during the peak electricity period of the next working day according to the input data; the input data includes the power of the power-consuming devices, raw coal screening and floating-sinking data, sorting process time series data, working temperature and humidity of the power-consuming devices, vibration signals, health assessment indicators, production shifts, production targets, and equipment start-stop plans. The Transformer-TCN hybrid architecture includes a Transformer layer and a TCN layer. The Transformer layer is used to capture the long-term dependencies of load data, and the TCN layer extracts local temporal features through causal convolution and dilated convolution. The input data of the non-stationary load signal is decomposed by variational mode decomposition to extract modal components in different frequency bands, reduce noise interference and enhance feature interpretability. A temporal feature matrix is generated through Z-score normalization and sliding window technology.

9. The coal preparation plant energy storage system based on digital twin according to claim 1, characterized in that: The hybrid energy storage system stores electrical energy during the low electricity price period and injects the excess electrical energy into the power grid during the peak period. The grid power electronic interface further includes a virtual synchronous generator, which is used to simulate the rotor of the synchronous generator in the power grid system. The operating state of the power grid is obtained in real time to determine whether the change in the power grid frequency reaches a threshold. When the set threshold is reached, the virtual synchronous generator simulates the inertial response of the synchronous generator in the power grid system, dynamically adjusts the inertia of the virtual generator to delay the frequency change, calculates the power of the virtual generator according to the frequency deviation and the adjusted inertia, and adjusts the power distribution of the virtual synchronous generator corresponding to the battery array according to the health status of the battery array in the hybrid energy storage system.

10. A coal preparation plant energy storage method, characterized in that: The coal preparation plant energy storage system based on digital twin as described in any one of claims 1 to 9 is adopted.

Images