Digital twin cooperative power system user side energy optimization regulation method

Abstract

The application discloses a digital twin cooperative power system user side energy optimization regulation method and belongs to the technical field of power system regulation. The application constructs a digital twin model which is in real-time synchronization with the physical entity of the user side energy system, adopts a four-layer architecture of a geometric mapping layer, a physical mapping layer, a behavior mapping layer and a rule mapping layer, realizes full-dimensional depiction of equipment space layout, operation mechanism, behavior trend and constraint rules, and ensures that the model is in real-time synchronization with the physical entity. Meanwhile, by establishing a unified cooperative decision framework, relying on a digital twin cooperative bus to break data islands and greatly improving the orderliness of multi-device cooperative regulation and resource allocation efficiency, the application realizes balanced optimization of multi-dimensional targets through an optimal solution set screening mechanism, combines scenario-based linkage rules and dynamic adjustment strategies, improves the adaptation capability to complex scenes, dynamically corrects model parameters and constraint conditions, ensures that the regulation strategy continuously adapts to the actual operation state, and improves the regulation precision and stability.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a digital twin-based collaborative method for optimizing energy control on the user side of a power system. Background Technology

[0002] Currently, in the field of user-side energy regulation in power systems, static optimization models based on historical operating data or traditional centralized regulation methods are mainly used to achieve dispatch and management of user-side distributed energy resources, energy storage devices, and various electricity loads. Although the concept of digital twins has been introduced, it is limited to the digital mapping of single devices or local scenarios, failing to achieve dynamic mapping of all elements and processes of the user-side energy system. During regulation, there is a lack of effective data interaction and collaborative decision-making mechanisms between diverse user-side energy devices and between the user side and the grid side. Optimization objectives are often based on grid operational stability or the energy cost of individual users, generating regulation strategies through preset fixed rules or simple algorithms. This makes it difficult to adapt to complex and ever-changing source-load fluctuations, environmental changes, and user energy demands, and the following problems remain:

[0003] 1. Existing technologies rely on static parameters and historical data to build control models, which cannot reflect dynamic factors such as fluctuations in distributed energy output, degradation of energy storage equipment status, and random changes in flexible loads in real time. This results in insufficient model accuracy and a mismatch between control decisions and actual scenarios.

[0004] 2. The lack of a digital twin mapping system for all elements of the user-side energy system has resulted in inconsistent data interaction formats among various user-side devices, delays in data transmission between the user side and the power grid side, and a lack of a real-time collaborative decision-making mechanism based on digital twins, leading to low efficiency in the allocation of regulatory resources.

[0005] 3. Existing control schemes mostly focus on a single objective, failing to take into account multiple dimensions such as energy efficiency, environmental benefits, and user energy comfort, making it difficult to achieve multi-objective synergistic optimization; control rules are mostly preset fixed modes, which result in delayed response when facing complex scenarios such as extreme weather, sudden load growth, and changes in power grid dispatch instructions, making it impossible to dynamically adjust control parameters and strategies, and resulting in insufficient control accuracy and robustness. Summary of the Invention

[0006] The purpose of this invention is to provide a digital twin-based collaborative method for optimizing and controlling energy on the user side of a power system, in order to solve the problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A digital twin-based collaborative method for user-side energy optimization and control in power systems includes the following steps:

[0009] Step 1: Digital twin modeling of user-side energy system: Obtain energy data from the user side of the power system and preprocess it to obtain the effective energy data after preprocessing. Combine the physical topology of the user-side energy system to construct a digital twin model of the user-side energy system.

[0010] Step 2: Real-time acquisition of multi-source data and synchronization of digital twin: Build a timestamp alignment engine to accurately align the preprocessed data with the timeline of the digital twin model, driving the real-time synchronization of the geometric state, physical operating state, behavior prediction results and physical entities of the user-side energy digital twin model;

[0011] Step 3: Construction of a collaborative control mechanism: Establish a collaborative decision-making framework within the user side and between the user side and the power grid side; build a linkage rule library for each twin unit in conjunction with the user-side energy digital twin model; clarify the linkage triggering conditions; and define the collaborative response priority of each twin unit under different operating scenarios.

[0012] Step 4: Generation of multi-objective optimization control strategy: Based on the real-time state simulated by the digital twin model and combined with the collaborative decision-making framework, a multi-objective optimization control system is determined, a multi-objective optimization model is constructed, the optimal solution set that satisfies the constraints is obtained based on the multi-objective optimization model, and the optimal control strategy is obtained by selecting from the optimal solution set according to the priority of collaborative response.

[0013] Step 5: Control Strategy Execution: Based on the optimal control strategy, control commands are sent to the distributed power controller, energy storage charge and discharge manager, flexible load regulator, and grid interaction terminal in the user-side energy system to drive each physical device to perform corresponding control operations.

[0014] Furthermore, the acquisition and preprocessing of energy data from the user side of the power system specifically includes:

[0015] Sensors and smart metering devices are deployed at key locations such as the output end of distributed energy equipment, the charging and discharging interface of energy storage equipment, the control end of electrical equipment, grid interaction nodes, and the user side to collect real-time operating data.

[0016] Extract data features from real-time operational data, and classify the collected real-time operational data based on the data features to obtain sub-operational datasets. Perform corresponding preprocessing on each sub-operational dataset based on the data features to generate standardized effective energy data.

[0017] Furthermore, the construction of the user-side energy digital twin model also includes:

[0018] Obtain the physical entity information of each twin unit, and perform layered modeling for each twin unit according to a four-layer mapping architecture of geometric mapping layer, physical mapping layer, behavioral mapping layer and rule mapping layer, so that each twin unit has the ability to map to the corresponding physical entity in all dimensions.

[0019] Based on the requirements of hierarchical modeling, a four-layer mapping architecture is adopted to construct the geometric mapping layer, physical mapping layer, behavioral mapping layer and behavioral mapping layer of each twin unit;

[0020] Among them, the geometric mapping layer is used to restore the spatial layout, installation location and energy transmission link topology of the corresponding equipment; the physical mapping layer is used to establish the equipment operation mechanism model based on the law of conservation of energy and the principle of electromagnetic induction; the behavior mapping layer is used to integrate historical and real-time data and use machine learning algorithms to build a predictive model of equipment operation behavior; and the rule mapping layer is used to embed power grid security rules, energy trading rules and user energy preference rules.

[0021] Furthermore, the specific modeling process for each twin unit includes:

[0022] A geometric mapping layer is constructed based on the installation location of the load equipment and the network topology to restore the spatial layout of flexible and rigid loads and the power link topology.

[0023] Based on the law of energy conservation and load energy consumption characteristics, a physical mapping layer is constructed to establish a flexible load power regulation model and a rigid load constant power model.

[0024] By integrating historical energy consumption data with real-time environmental data, a behavior mapping layer is constructed to generate a load power prediction model;

[0025] A rule mapping layer is constructed by embedding user energy consumption preferences and grid constraint rules, and binding user energy consumption time period thresholds, comfort constraints and grid voltage and frequency limit rules;

[0026] A geometric mapping layer is constructed based on the installation coordinates and grid connection point of the distributed power source to reconstruct the spatial deployment of the distributed power source and its connection topology with the power grid.

[0027] A physical mapping layer is constructed based on the principle of electromagnetic induction and the correlation with environmental parameters to establish photovoltaic power output model and wind turbine power output model;

[0028] By integrating historical power output data and meteorological forecast data, a behavior mapping layer is constructed to generate a distributed power generation output prediction model;

[0029] An embedded grid connection rule and energy trading rule are used to construct a rule mapping layer, which binds grid connection voltage / frequency constraints, peak-valley electricity price rules and ancillary service compensation rules.

[0030] A geometric mapping layer is constructed based on the cabinet layout and charging / discharging interface topology of the energy storage battery pack to restore the connection links between the energy storage device and the load, distributed power source, and power grid.

[0031] A physical mapping layer is constructed based on electrochemical principles and charge-discharge characteristics, and an energy storage charge-discharge model considering charge-discharge efficiency and remaining capacity constraints is established.

[0032] By integrating historical charging and discharging data with source-load prediction data, a behavior mapping layer is constructed to generate an energy storage charging and discharging strategy optimization model;

[0033] An embedded energy storage operation safety rules and trading rules are used to construct a rule mapping layer, which binds the upper and lower limits of the remaining energy storage capacity, the charging and discharging time limit, and the peak-valley electricity price arbitrage rules.

[0034] A geometric mapping layer is constructed based on the connection topology between the user-side grid connection point and the distribution network, and the installation location of metering equipment, to restore the interaction link between the user side and the power grid and the layout of metering nodes;

[0035] A physical mapping layer is constructed based on the power transmission principle and the power grid impedance characteristics to establish a power grid interaction calculation model.

[0036] By integrating historical interactive power data with power grid dispatch instructions, a behavior mapping layer is constructed to generate a power grid interactive power prediction model;

[0037] An embedded rule mapping layer is constructed using power grid dispatch rules and transaction rules, binding power grid dispatch command constraints, power purchase and sale price rules, and power grid safety and stability operation rules.

[0038] Furthermore, a physical mapping layer is constructed based on electrochemical principles and charge-discharge characteristics, and an energy storage charge-discharge model considering charge-discharge efficiency and remaining capacity constraints is established, including:

[0039] Piecewise fitting is used to obtain the open-circuit voltage measurement data of the energy storage battery in each state of charge range, which was collected in advance through a piecewise constant current static calibration experiment, and to obtain a table of correspondence between open-circuit voltage and state of charge covering the entire state of charge range.

[0040] The electrochemical impedance spectroscopy data collected in advance through offline pulse charge-discharge calibration experiments under multiple temperature ranges and multiple charge-discharge rates are processed for equivalent circuit parameter identification. Combined with the corresponding relationship table, the ohmic internal resistance, polarization internal resistance and polarization capacitance parameter values ​​under each temperature range and each charge-discharge rate combination are extracted respectively, and an electrochemical equivalent circuit parameter lookup table indexed by temperature range and charge-discharge rate is obtained.

[0041] Temperature correction modeling is performed on the variation of each impedance parameter with temperature in the electrochemical equivalent circuit parameter lookup table. Each impedance parameter is expressed as a continuous variation function with respect to the current battery temperature, and a temperature adaptive equivalent circuit model is obtained that can dynamically output the current equivalent circuit parameters based on the real-time temperature.

[0042] The real-time synchronized charging and discharging current time series and battery surface temperature data are input into the temperature adaptive equivalent circuit model. The battery terminal voltage response and the current state of charge are jointly recursively estimated. The estimation results are corrected in parallel based on the energy loss difference corresponding to the charging and discharging processes, and a dynamic correspondence table of charging and discharging efficiency is obtained with the current charging and discharging rate, current temperature and current state of charge as indexes.

[0043] The dynamic correspondence table of charging and discharging efficiency is superimposed with the upper and lower limits of the state of charge that the energy storage device is allowed to operate, the upper limit of the maximum charging and discharging power allowed to be output under each operating condition, and the maximum energy throughput allowed in a single scheduling cycle. The compliance judgment of the charging and discharging operation under each operating condition is carried out, and infeasible operating conditions that exceed the boundary are eliminated to obtain the compliance and feasibility domain dataset of charging and discharging operation under each operating condition.

[0044] For the compliant and feasible domain dataset, a battery life decay mapping relationship is introduced, with the single charge-discharge cycle depth and charge-discharge rate as inputs and the corresponding battery capacity decay rate as output. The life loss cost corresponding to each charge-discharge operation scheme in the compliant and feasible domain is quantitatively labeled, and an energy storage charge-discharge model that simultaneously includes operating efficiency information, remaining capacity safety boundary information, and life loss cost information is obtained.

[0045] Furthermore, the energy storage twin unit also includes:

[0046] By integrating the battery characteristic model of energy storage devices with real-time operating data, a model for simulating the charge and discharge characteristics of energy storage devices, predicting SOC, and assessing their lifespan is constructed.

[0047] Key features of the target operation data of energy storage equipment are extracted, and the feature data are clustered to obtain voltage operation data groups and current operation data groups of energy storage equipment under different operating conditions.

[0048] Based on the clustering results, the parameters of the energy storage twin unit model under different operating conditions are calibrated differently.

[0049] Furthermore, based on the multi-objective optimization model, an optimal set of solutions satisfying the constraints is obtained. The optimal control strategy is then selected from this optimal set according to the priority of the coordinated response, including:

[0050] Obtain the original calculated value of each candidate solution in the optimal solution set under each objective function, and normalize the original calculated value to obtain the normalized value in each objective function dimension.

[0051] Using the candidate solution index as the row index and the objective function index as the column index, all normalized values ​​are filled according to their row and column positions to construct a normalized objective function matrix;

[0052] Construct a feature vector for the current operating scenario based on real-time system operating status data collected from the digital twin platform.

[0053] The label of the current running scenario is determined based on the feature vector of the current running scenario and the pre-stored center vectors of each typical scenario;

[0054] Based on the current running scenario tags and linkage rule base, determine the collaborative response priority coefficient of each objective function;

[0055] Based on the coordinated response priority coefficient and the basic weights of each objective function preset by historical control experience, the dynamic weights of each objective function under the previous running scenario are calculated.

[0056] The comprehensive score of each candidate control strategy is calculated based on the normalized objective function matrix, dynamic weights, and constraint violation degree of each candidate solution.

[0057] The candidate solution with the highest comprehensive score and the constraint violation degree that meets the preset conditions is selected as the optimal control strategy.

[0058] Furthermore, the method of driving each physical device to perform corresponding control operations also includes:

[0059] Based on the current operating status of user-side energy equipment, analyze energy regulation projects to determine the current regulation information of energy regulation projects, including the output adjustment progress of distributed power sources, the execution progress of energy storage charging and discharging, the completion degree of flexible load transfer, and the power matching degree of grid interaction;

[0060] Based on the current control information of the energy control project, an energy control quality analysis is conducted to obtain control quality evaluation factors. Based on the control quality evaluation factors, control quality assessment data is calculated to obtain control quality analysis data.

[0061] The regulation quality analysis data is standardized and analyzed in conjunction with the preset energy regulation standards to determine whether the regulation quality analysis data meets the preset energy regulation standards and to obtain the energy regulation quality analysis results on the user side.

[0062] When the user-side energy regulation quality analysis result is that the regulation quality analysis data does not meet the preset energy regulation standard, the cause analysis is performed based on the regulation quality analysis data to determine the cause of the regulation quality analysis data not meeting the preset energy regulation standard, and the cause of the non-standard energy regulation is obtained.

[0063] Based on the reasons for irregular energy regulation, and in conjunction with the energy regulation planning scheme, we analyze whether the reasons for irregular energy regulation can be corrected during the regulation implementation process, and obtain the results of the cause analysis and judgment.

[0064] Furthermore, obtaining the cause analysis and judgment results also includes:

[0065] When the cause analysis results indicate that the irregularities in energy regulation can be corrected during the regulation execution process, regulation correction information is generated based on the causes of the irregularities in energy regulation.

[0066] Simultaneously, based on the progress deviation of the energy regulation task and the real-time monitoring of equipment operating status data, corresponding regulation operation instructions are generated. The regulation correction information is used to correct and adjust the corresponding regulation operation instructions, and the energy equipment is guided to adjust its working strategy according to the corrected and adjusted regulation operation instructions.

[0067] When the cause analysis results indicate that the non-standard energy regulation cannot be corrected during the regulation execution process, a corresponding regulation operation instruction is generated based on the progress deviation of the energy regulation task and the real-time monitored equipment operating status data.

[0068] Simultaneously, the difference between the control quality analysis data and the preset energy control standard is determined, and a danger warning information is generated based on the difference data. While guiding the energy equipment to adjust its working strategy according to the control operation instructions, a warning prompt is issued based on the danger warning information.

[0069] Furthermore, the implementation of the control strategy also includes:

[0070] During the control and control process, the execution feedback data of each device is collected in real time, and the feedback data is transmitted to the digital twin model and subjected to deviation analysis with the model prediction data;

[0071] If the deviation value is greater than the preset deviation threshold, the parameters of the physical mapping layer and behavioral mapping layer of the digital twin model and the constraints of the optimized model are corrected based on the feedback data, and the corrected control strategy is regenerated.

[0072] If the deviation value is less than or equal to the preset deviation threshold, the current control strategy remains unchanged, and the operating parameters of the digital twin model are updated through an incremental update algorithm.

[0073] Compared with the prior art, the beneficial effects of the present invention are:

[0074] 1. This invention constructs a digital twin model that is synchronized in real time with the physical entity of the user-side energy system. It adopts a four-layer architecture of geometric mapping layer, physical mapping layer, behavioral mapping layer and rule mapping layer to achieve a full-dimensional characterization of equipment spatial layout, operation mechanism, behavior trend and constraint rules, ensuring that the model is synchronized with the physical entity in real time. This solves the problem that static models in the prior art cannot reflect dynamic characteristics, and achieves accurate characterization of source and load equipment status, energy use scenario and grid interaction relationship.

[0075] 2. This invention establishes a unified collaborative decision-making framework, standardizes the data interaction format, command transmission path, and response time limit of each twin unit, and clarifies the triggering conditions and response priorities by constructing a linkage rule base, thereby achieving efficient collaboration within the user side and between the user side and the power grid side. Relying on the digital twin collaborative bus, it breaks down data silos, solves the defects of untimely data interaction and disordered linkage in existing technologies, and significantly improves the orderliness of multi-device collaborative control and resource allocation efficiency.

[0076] 3. This invention constructs a multi-objective optimization system that includes maximizing energy utilization efficiency, minimizing grid interaction costs, optimizing energy supply and demand balance, and maximizing the lifespan of energy storage devices. It also incorporates multiple constraints such as source-load prediction error and energy storage SOC safety. Through an optimal solution set screening mechanism, it achieves balanced optimization of multi-dimensional objectives, solving the problem of single optimization objectives in existing technologies. Combined with scenario-based linkage rules and dynamic adjustment strategies, it significantly improves the adaptability to complex scenarios such as extreme weather and sudden loads.

[0077] 4. This invention collects equipment execution feedback data in real time, compares model predictions with actual values, and dynamically corrects model parameters and constraints. For problems with substandard control quality, it can quickly locate the cause and generate correction instructions or trigger early warning prompts. This solves the defects of existing technologies, such as lack of closed-loop correction, response lag, and insufficient robustness, ensuring that the control strategy continuously adapts to the actual operating state and improving control accuracy and stability. Attached Figure Description

[0078] Figure 1 This is a flowchart of the digital twin collaborative power system user-side energy optimization and control method of the present invention;

[0079] Figure 2 This is a flowchart of the four-layer modeling and operating condition calibration process for the energy storage twin unit of the present invention. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] Please see Figures 1-2 The present invention provides the following technical solutions:

[0082] A digital twin-based collaborative method for user-side energy optimization and control in power systems includes the following steps:

[0083] Step 1: Digital Twin Modeling of User-Side Energy System: Acquire and preprocess energy data from the user side of the power system. This energy data includes user-side load data, distributed power generation output data, energy storage device operation data, grid interaction power data, and environmental sensing data. Obtain the preprocessed effective energy data and, combined with the physical topology of the user-side energy system, construct a digital twin model of the user-side energy system. The digital twin model includes load twin units, distributed power generation twin units, energy storage twin units, and grid interaction twin units. Each twin unit achieves data interaction through a digital twin collaborative bus.

[0084] Step 2: Real-time acquisition of multi-source data and synchronization of digital twin: Construct a timestamp alignment engine to accurately align the preprocessed data with the timeline of the digital twin model. Use an incremental update algorithm to update only the changed parameters to drive the geometric state, physical operating state, and behavior prediction results of the user-side energy digital twin model to synchronize with the physical entity in real time. Control the synchronization delay to no more than 1 second to ensure that the digital twin can mirror the operating state of the physical entity in real time.

[0085] Step 3: Construction of Collaborative Control Mechanism: Establish a collaborative decision-making framework within the user side and between the user side and the grid side, standardize the data interaction format, command transmission path, and response time limit requirements between various twin units, and ensure the orderliness and consistency of multi-unit linkage control; build a linkage rule base for each twin unit in conjunction with the user-side energy digital twin model, clarify the linkage triggering conditions between the load twin unit and the distributed power generation twin unit, energy storage twin unit, and grid interaction twin unit, and define the collaborative response priority of each twin unit under different operating scenarios;

[0086] Step 4: Generation of Multi-Objective Optimization Control Strategies: Based on the real-time state simulated by the digital twin model and combined with the collaborative decision-making framework, a multi-objective optimization control system is determined, including maximizing user-side energy utilization efficiency, minimizing grid interaction costs, optimizing energy supply and demand balance, and maximizing the lifespan of energy storage devices; a multi-objective optimization model is constructed, including source-load prediction error constraints, grid interaction power constraints, energy storage SOC safety constraints, and equipment operation physical constraints; based on the multi-objective optimization model, the optimal solution set that satisfies the constraints is obtained, and the optimal control strategies are selected from the optimal solution set according to the priority of collaborative responses, including distributed power generation output adjustment schemes, energy storage charging and discharging control schemes, flexible load transfer schemes, and grid interaction power allocation schemes;

[0087] Step 5: Control Strategy Execution: Based on the optimal control strategy, control commands are sent to the distributed power controller, energy storage charge and discharge manager, flexible load regulator, and grid interaction terminal in the user-side energy system to drive each physical device to perform corresponding control operations.

[0088] In this embodiment, acquiring and preprocessing energy data from the user side of the power system specifically includes:

[0089] Sensors and smart metering devices are deployed at key locations such as the output end of distributed energy equipment, the charging and discharging interface of energy storage equipment, the control end of electrical equipment, grid interaction nodes, and the user side to collect real-time operating data.

[0090] Extract data features from real-time operational data, and classify the collected real-time operational data based on the data features to obtain sub-operational datasets. Perform corresponding preprocessing on each sub-operational dataset based on the data features, including data cleaning, data standardization, and data association and integration, to remove invalid data, standardize data format, and generate standardized and effective energy data.

[0091] In this embodiment, by constructing a user-side energy digital twin model containing multiple types of twin units and relying on a digital twin collaborative bus to realize data interaction among the units, collaborative perception and linkage control of all aspects of the user-side energy system are achieved. This effectively solves the technical challenges brought about by the fluctuation of distributed power output and the uncertainty of load demand to user-side energy optimization. By deploying sensors and smart metering devices at key locations to collect real-time data and combining it with physical topology modeling, the digital twin model can accurately map the operating status of the physical system, providing a high-precision decision-making basis for subsequent energy optimization and control. Compared with control methods that rely on offline data or experience models, this significantly improves the real-time performance, accuracy, and effectiveness of the control strategy, thereby achieving the technical effects of improving user-side energy utilization efficiency, reducing grid interaction costs, and optimizing energy supply and demand balance.

[0092] In this embodiment, constructing a user-side energy digital twin model further includes layered modeling of each twin unit according to a four-layer mapping architecture: geometric mapping layer, physical mapping layer, behavioral mapping layer, and rule mapping layer. This enables each twin unit to have full-dimensional mapping capabilities with the corresponding physical entity. The geometric mapping layer is used to restore the spatial layout, installation location, and energy transmission link topology of the corresponding equipment; the physical mapping layer is used to establish a model of the equipment's operating mechanism based on the law of conservation of energy and the principle of electromagnetic induction; the behavioral mapping layer is used to integrate historical and real-time data and use machine learning algorithms to construct a predictive model of the equipment's operating behavior; and the rule mapping layer is used to embed power grid security rules, energy trading rules, and user energy preference rules.

[0093] In this embodiment, the specific modeling process for each twin unit includes:

[0094] For load twin units: a geometric mapping layer is constructed based on the installation location and network topology of load equipment to restore the spatial layout and power link topology of flexible and rigid loads; a physical mapping layer is constructed based on the law of conservation of energy and load energy consumption characteristics to establish a power regulation model for flexible loads and a constant power model for rigid loads; a behavioral mapping layer is constructed by integrating historical energy consumption data and real-time environmental data to generate a load power prediction model; and a rule mapping layer is constructed by embedding user energy consumption preferences and grid constraint rules to bind user energy consumption time thresholds, comfort constraints, and grid voltage and frequency limit rules.

[0095] For distributed generation twin units: a geometric mapping layer is constructed based on the installation coordinates and grid connection point of the distributed generation to reconstruct the spatial deployment of the distributed generation and its connection topology with the grid; a physical mapping layer is constructed based on the principle of electromagnetic induction and the correlation of environmental parameters to establish photovoltaic power output models and wind turbine power output models; a behavioral mapping layer is constructed by integrating historical power output data and meteorological forecast data to generate a distributed generation power output prediction model; and a rule mapping layer is constructed by embedding grid connection rules and energy trading rules to bind grid connection voltage / frequency constraints, peak-valley electricity price rules, and ancillary service compensation rules.

[0096] For energy storage twin units: A geometric mapping layer is constructed based on the cabinet layout and charging / discharging interface topology of the energy storage battery pack to restore the connection links between the energy storage device and the load, distributed power source, and grid; a physical mapping layer is constructed based on electrochemical principles and charging / discharging characteristics to establish an energy storage charging / discharging model that considers charging / discharging efficiency and remaining capacity constraints; a behavioral mapping layer is constructed by integrating historical charging / discharging data and source-load prediction data to generate an energy storage charging / discharging strategy optimization model; and a rule mapping layer is constructed by embedding energy storage operation safety rules and trading rules to bind upper and lower limits of remaining energy storage capacity, charging / discharging duration limits, and peak-valley electricity price arbitrage rules.

[0097] For the power grid interaction twin unit: a geometric mapping layer is constructed based on the connection topology between the user-side grid connection point and the distribution network, and the installation location of metering equipment, to restore the interaction link between the user side and the power grid and the layout of metering nodes; a physical mapping layer is constructed based on the power transmission principle and the power grid impedance characteristics to establish a power grid interaction power calculation model; a behavioral mapping layer is constructed by integrating historical interaction power data and power grid dispatch instructions to generate a power grid interaction power prediction model; and a rule mapping layer is constructed by embedding power grid dispatch rules and trading rules to bind power grid dispatch instruction constraints, power purchase and sale price rules, and power grid safe and stable operation rules.

[0098] In this embodiment, a physical mapping layer is constructed based on electrochemical principles and charge-discharge characteristics, and an energy storage charge-discharge model considering charge-discharge efficiency and remaining capacity constraints is established, including:

[0099] Piecewise fitting is used to obtain a table showing the correspondence between open-circuit voltage and state of charge of the energy storage battery in each state of charge range, which is obtained by pre-collecting open-circuit voltage measurement data of the energy storage battery in each state of charge range through piecewise constant current static calibration experiments.

[0100] The experiment was performed on single-cell samples under a constant temperature environment (the calibration temperature in this embodiment was set to 25°C): First, the sample cells were fully charged to the cutoff voltage using a small current of 0.1C and allowed to stand for 2 hours to reach electrochemical equilibrium. Then, they were continuously discharged at a constant current of 0.1C. Discharge was paused every 5% of the rated capacity and allowed to stand for 40 minutes. The open-circuit voltage at the end of the standing period was recorded. This process was repeated until the discharge cutoff voltage was reached, resulting in approximately 20 open-circuit voltage measurements corresponding to different states of charge. The experiment was repeated three times, and the average value was taken to eliminate random measurement errors, resulting in the original discrete calibration data set.

[0101] The original discrete data set was piecewise fitted to obtain a table showing the correspondence between open-circuit voltage and state of charge (SOC) covering the entire SOC range. The specific implementation of piecewise fitting was as follows: the SOC range from 0% to 100% was divided into several sub-intervals. In the plateau region unique to lithium iron phosphate batteries (SOC between approximately 20% and 80%, where the open-circuit voltage changes very smoothly with SOC), a low-order polynomial fitting with fewer nodes was used to avoid overfitting and introducing spurious fluctuations. In the non-plateau regions where SOC is below 20% and above 80%, the open-circuit voltage changes more significantly with SOC, and a piecewise cubic polynomial fitting with denser nodes was used to ensure accuracy. The fitting results were stored in a lookup table with a SOC step size of 0.5%. The table contained 201 interpolation nodes, and linear interpolation was used to read intermediate values ​​between adjacent nodes.

[0102] The electrochemical impedance spectroscopy data collected in advance through offline pulse charge-discharge calibration experiments under multiple temperature ranges and multiple charge-discharge rates are processed for equivalent circuit parameter identification. Combined with the corresponding relationship table, the ohmic internal resistance, polarization internal resistance and polarization capacitance parameter values ​​under each temperature range and each charge-discharge rate combination are extracted to obtain an electrochemical equivalent circuit parameter lookup table indexed by temperature range and charge-discharge rate.

[0103] The experiment was conducted in a temperature-controlled chamber at five temperature levels: -10℃, 0℃, 10℃, 25℃, and 40℃. At each temperature level, pulse excitation sequences with four charge / discharge rates of 0.2C, 0.5C, 1C, and 2C were applied to the sample batteries. At each target state of charge node (10 nodes were selected in 10% increments), a constant current charging pulse lasting 10 seconds was applied, followed by an open-circuit rest for 60 seconds, then a constant current discharging pulse lasting 10 seconds was applied, followed by another 60-second rest, and this cycle was repeated. The transient response curve of the battery terminal voltage and the charging / discharging current waveform were recorded throughout the process at a sampling frequency of at least 1000Hz.

[0104] The transient response curves of the terminal voltage obtained from the above pulse experiment were processed for equivalent circuit parameter identification to obtain an electrochemical equivalent circuit parameter lookup table indexed by temperature range and charge / discharge rate. This embodiment uses a second-order RC equivalent circuit model, which consists of an ohmic internal resistance, two parallel RC circuits (corresponding to the electrochemical polarization process and concentration polarization process, respectively), and an open-circuit voltage source in series. It includes five parameters to be identified: ohmic internal resistance, first polarization internal resistance, first polarization capacitance, second polarization internal resistance, and second polarization capacitance. The identification method is a time-domain curve fitting method: the initial value of the ohmic internal resistance is directly estimated by the voltage drop at the first sampling point after the pulse ends (the ohmic voltage drop can be approximated by the product of the ohmic internal resistance and the pulse current). Then, nonlinear least-squares fitting is performed on the relaxation recovery curve of the terminal voltage during the resting period, using a second-order exponential decay function as the fitting model. The two time constants and their corresponding amplitudes are solved, and the polarization internal resistance value is then calculated by inversely using the product of the time constant and the capacitance. The identification process is performed independently on the data of all 20 state-of-charge nodes. The average value of the identification results of each node is taken as the representative parameter value of the temperature-rate combination. A total of 5×4=20 sets of parameter records are generated and organized into a two-dimensional parameter lookup table with temperature range row index and charge / discharge rate column index. Each cell in the table stores the five equivalent circuit parameter values ​​under the corresponding operating condition.

[0105] Temperature correction modeling is performed on the variation of each impedance parameter with temperature in the electrochemical equivalent circuit parameter lookup table. Each impedance parameter is expressed as a continuous variation function with respect to the current battery temperature, and a temperature adaptive equivalent circuit model is obtained that can dynamically output the current equivalent circuit parameters based on the real-time temperature.

[0106] Using the calibration data from five temperature ranges in the parameter lookup table as fitting samples, temperature correction functions were established for the three impedance parameters: ohmic internal resistance, first polarization internal resistance, and second polarization internal resistance. The polarization capacitance parameter has relatively low temperature sensitivity and was processed using piecewise linear interpolation. The method for establishing the temperature correction functions is as follows: Observing the changing trends of each impedance parameter at the five temperature ranges, the impedance parameters showed a monotonically decreasing trend with increasing temperature in the range of -10℃ to 25℃, with the rate of decrease being steeper in the low-temperature region than in the high-temperature region. The change tended to be gentler in the range of 25℃ to 40℃, exhibiting an overall non-linear monotonically decreasing law. Based on this, a piecewise piecewise linear fitting method was used to establish the temperature correction function for each impedance parameter: a line segment with a larger slope was used in the low-temperature region of -10℃ to 10℃, and a line segment with a smaller slope was used in the normal to high-temperature region of 10℃ to 40℃, with the inflection point set at 10℃; the slope of each line segment was determined by performing linear regression on the calibration values ​​of adjacent temperature ranges. The resulting temperature-adaptive equivalent circuit model can receive the battery surface temperature value reported by the BMS in real time as input, and query the output of the real-time equivalent circuit parameters at the current operating temperature through the temperature correction function. This parameter output will be directly called in the next step.

[0107] It should be noted that this embodiment uses a piecewise linear function as the temperature correction function, mainly because it is easy to implement on the real-time controller, does not require floating-point exponential operations, and is easy to deploy on embedded platforms with limited computing power. In cases where computing resources are sufficient, a smooth curve function can be used to obtain a more continuous temperature interpolation effect. The modeling logic of the two implementation methods is the same and does not affect the execution flow of subsequent steps.

[0108] The real-time synchronized charging and discharging current time series and battery surface temperature data are input into the temperature adaptive equivalent circuit model. The battery terminal voltage response and the current state of charge are jointly recursively estimated. The estimation results are corrected in sync with the energy loss difference between the charging and discharging processes. A dynamic correspondence table of charging and discharging efficiency is obtained with the current charging and discharging rate, current temperature and current state of charge as indexes.

[0109] Using the real-time charging and discharging current time series reported by the BMS as the excitation input, and the real-time equivalent circuit parameters output by the temperature adaptive equivalent circuit model at the current temperature as the model parameters, the dual extended Kalman filter algorithm is used to perform joint recursive estimation of the battery terminal voltage response and the current state of charge.

[0110] The state vector of the dual extended Kalman filter consists of the current state of charge and the polarization voltages of the two RC links; the observations are the measured terminal voltages reported by the BMS; the state equations are established based on the integral relationship between the state of charge and the charging / discharging current and the voltage relaxation dynamics of the two RC links; the observation equations are established based on the terminal voltage calculation relationship of the second-order RC equivalent circuit, where the open-circuit voltage is obtained by looking up the corresponding relationship table. The algorithm executes a prediction-update loop once in each control sampling period (100ms in this embodiment), and outputs the current state of charge estimate and the estimation error covariance in real time.

[0111] While performing joint estimation of terminal voltage and state of charge, differentiated corrections are made for energy loss during charging and discharging: During charging, a portion of the electrical energy flowing into the battery is dissipated as heat in the impedance links of the equivalent circuit, and the effective energy actually stored in the battery is less than the input energy at the charging end; the charging efficiency is defined as the ratio of the two. During discharging, the effective energy released by the battery is greater than the net electrical energy output to the load side; the difference is the heat dissipation in the impedance links, and the discharging efficiency is defined as the ratio of the output electrical energy to the released effective energy. Both types of efficiency values ​​are calculated by the product of the estimated terminal voltage and current in real time, and are dynamically updated based on the current charge / discharge rate, temperature estimate, and state of charge estimate.

[0112] The system categorizes and archives the real-time estimation results according to three dimensions: charge / discharge rate (divided into four levels: 0.2C, 0.5C, 1C, and 2C), temperature (divided into five intervals, corresponding to five calibration temperature levels), and state of charge (divided into ten intervals with a step size of 10%). A moving average method is used to smooth historical estimates falling within the same index cell, continuously updating the efficiency statistics for the corresponding index cell. In the initial stage of system operation, when data accumulation is insufficient, offline measurement efficiency values ​​from calibration experiments are used as initial fillers; as the operating time increases, online estimates gradually replace the initial values. The final dynamic table of charge / discharge efficiency is stored in the form of a three-dimensional index lookup table, reflecting the charge / discharge efficiency status under the current operating conditions in real time.

[0113] The dynamic correspondence table of charging and discharging efficiency is superimposed with the upper and lower limits of the allowed state of charge of the energy storage device, the upper limit of the maximum allowed charging and discharging power output under each operating condition, and the maximum allowed total energy throughput within a single scheduling cycle. The compliance of the charging and discharging operations under each operating condition is judged, and infeasible operating conditions that exceed the boundaries are eliminated to obtain a compliant and feasible domain dataset for charging and discharging operations under each operating condition.

[0114] The following three constraints are checked one by one for all indexed operating points in the dynamic mapping table of charge and discharge efficiency:

[0115] State of charge (SOC) safety boundary constraints. In this embodiment, the lower limit of the SOC of the lithium iron phosphate battery is set at 10%, and the upper limit is set at 95%, to avoid irreversible capacity loss due to over-discharge and the risk of lithium plating caused by overcharging. For operating points where the estimated SOC falls outside the range of [10%, 95%], the corresponding charge / discharge operations are marked as infeasible and removed from the corresponding table.

[0116] Maximum charge / discharge power constraints. The upper limit of the maximum allowable charge / discharge power under each operating condition is determined jointly based on the battery pack's rated power and the thermal management status reported by the BMS in real time: Under normal operating conditions at room temperature, the upper limit of the maximum charge / discharge power is taken as 100% of the rated power; when the highest single-cell temperature reported by the BMS exceeds 45°C, the system automatically reduces the maximum allowable power to 70% of the rated power to prevent thermal runaway; when the highest single-cell temperature exceeds 50°C, it is further reduced to 40% of the rated power and a temperature alarm is triggered. Operating points exceeding the current maximum allowable power are marked as infeasible.

[0117] Maximum energy throughput constraint per scheduling cycle. To delay battery cycle life degradation, the total energy throughput of charge and discharge operations within a single scheduling cycle (set to 24 hours in this embodiment) must not exceed 1.5 times the rated capacity (i.e., the equivalent cycle depth does not exceed 1.5 complete cycles). At the beginning of each scheduling cycle, the system resets the energy throughput counter to zero and accumulates the actual throughput as the scheduling progresses; when the remaining available throughput is insufficient to support the complete execution of the operation corresponding to a certain operating point, that operating point is marked as infeasible due to excess capacity.

[0118] After the above three constraint checks, all compliant operating condition points are retained and organized into a compliant feasible domain dataset. This dataset uses the operating condition triplet (charge / discharge rate, temperature, state of charge) as an index, and the corresponding charge / discharge efficiency value, maximum allowable power, and remaining available throughput as values, forming a dynamically updated three-dimensional sparse data structure. This structure provides quantitative boundary inputs for the energy storage-side inequality constraints to the multi-objective optimization model in step four in real time. Whenever the BMS reports new operating status data, the compliant feasible domain dataset is updated incrementally synchronously, with a delay not exceeding one BMS reporting cycle (100ms).

[0119] For the compliant and feasible domain dataset, a battery life decay mapping relationship is introduced, with the single charge-discharge cycle depth and charge-discharge rate as inputs and the corresponding battery capacity decay rate as output. The life loss cost corresponding to each charge-discharge operation scheme in the compliant and feasible domain is quantitatively labeled, and an energy storage charge-discharge model that simultaneously includes operating efficiency information, remaining capacity safety boundary information, and life loss cost information is obtained.

[0120] The battery life degradation mapping relationship was established through offline accelerated aging experiments. The experiments performed cyclic charging and discharging on sample batteries under various combinations of cycle depths (10%, 30%, 50%, 80%, 100%) and charge / discharge rates (0.2C, 0.5C, 1C, 2C). Capacity calibration was performed every 50 cycles, and the capacity degradation relative to the initial capacity was recorded. A two-dimensional lookup table was constructed, with the average capacity degradation rate per cycle on the vertical axis and the cycle depth and charge / discharge rate combination on the horizontal axis. The experimental results of this embodiment show that the greater the cycle depth and the higher the charge / discharge rate, the greater the corresponding capacity degradation rate per cycle. The effect of charge / discharge rate is more significant in the high-rate region (rate exceeding 1C), while the effect of cycle depth is more significant in the deep-cycle region (cycle depth exceeding 50%).

[0121] For each compliant operating condition point in the compliant feasible domain dataset, based on the actual cycle depth (estimated by the absolute value of the difference between the current state of charge and the target state of charge) and the charge / discharge rate involved in the corresponding charge / discharge operation scheme, a two-dimensional lookup table of decay rate is consulted to obtain the single-cycle capacity decay rate corresponding to the operation scheme; the decay rate is multiplied by the current calibrated capacity of the battery pack to convert it into the capacity loss amount of this operation in ampere-hours (Ah); further, combined with the battery replacement unit price and the remaining available capacity, the capacity loss amount is converted into the economic cost of life loss in monetary terms, and written as the life loss cost label value of this operating condition point into the corresponding index cell of the compliant feasible domain dataset.

[0122] After the lifetime loss is quantitatively labeled, the compliant and feasible domain dataset is upgraded to an energy storage charging and discharging model that simultaneously includes the following three types of information: First, operating efficiency information, namely the real-time charging and discharging efficiency value corresponding to each operating point, which is used to quantify the energy loss during the charging and discharging process in the economic cost sub-objective of the multi-objective optimization in step four; Second, remaining capacity safety boundary information, namely the compliance label and allowable operating power limit of each operating point, which is used to construct the quantitative boundary of the energy storage side inequality constraint conditions in the multi-objective optimization model in step four; Third, lifetime loss cost information, namely the economic cost of lifetime loss corresponding to each operating point, which is used to construct the energy storage lifetime minimization sub-objective in the multi-objective optimization objective function in step four, so that the optimization and control process can take into account the full life cycle cost of energy storage equipment while pursuing economic benefits.

[0123] The energy storage charge-discharge model is persistently stored in a structured database and incrementally refreshed during each scheduling cycle as the data reported by the BMS is updated. The compliance and feasible domain boundary in the model is updated in real time (the refresh cycle follows the BMS reporting cycle, which is 100ms), the efficiency information is updated smoothly using a moving average method (the update step is performed once every 10 newly accumulated sampling points under the same operating conditions), and the lifetime loss cost is updated at a frequency of recalibration after each scheduling cycle to match the time scale of the battery capacity slowly decaying with the number of cycles.

[0124] Finally, the compliance and feasible domain data of the energy storage charging and discharging model is pushed to the multi-objective optimization model in step four in real time via an interface, serving as the quantitative boundary input for the inequality constraints on the energy storage side; the quantification result of lifetime loss cost is passed to the multi-objective optimization objective function in step four as objective function parameters, serving as the basis for constructing the sub-objective of minimizing energy storage lifetime; the upper limit of allowed operating power and the safety boundary of state of charge are passed to the energy storage charging and discharging manager in step five as command constraint parameters, serving as hard boundaries for power and timing constraints when issuing control commands, ensuring that the control commands are always within the safe operating range of the energy storage device at the physical level.

[0125] In this embodiment, the energy storage twin unit further includes:

[0126] By integrating the battery characteristic model of energy storage devices with real-time operating data, a model for simulating the charge and discharge characteristics of energy storage devices, predicting SOC, and assessing their lifespan is constructed.

[0127] Extract key features from the target operating data of energy storage devices, including voltage fluctuation amplitude, current stability, SOC change rate, and temperature gradient;

[0128] The K-means clustering algorithm was used to cluster the feature data to obtain voltage operation data groups and current operation data groups of energy storage equipment under different operating conditions;

[0129] Based on the clustering results, the parameters of the energy storage twin unit model under different operating conditions are calibrated differently to improve the simulation and prediction accuracy of the model under complex operating conditions and provide data support for the refined charging and discharging control of energy storage devices.

[0130] In this embodiment, an optimal set of solutions satisfying the constraints is obtained based on a multi-objective optimization model. The optimal control strategy is then selected from the optimal set of solutions according to the priority of the coordinated response, including:

[0131] Obtain each candidate solution in the optimal solution set Original calculated values ​​under each objective function The original calculated values ​​are normalized to obtain normalized values ​​for each dimension of the objective function. :

[0132] ;

[0133] in, For candidate solutions In the Normalized values ​​on each objective function dimension; for The maximum value obtained statistically within the Pareto optimal solution set; for The minimum value obtained statistically within the Pareto optimal solution set;

[0134] By candidate solution number Row index, objective function number As column indexes, fill all normalized values ​​according to row and column positions to construct a normalized objective function matrix;

[0135] Construct a feature vector for the current operating scenario based on real-time system operating status data collected from the digital twin platform.

[0136] The label of the current running scenario is determined based on the feature vector of the current running scenario and the pre-stored center vectors of each typical scenario;

[0137] Based on the current running scenario label In conjunction with the rule base, determine the priority coefficients of the collaborative responses for each objective function. ;

[0138] Based on the collaborative response priority coefficient The basic weights of each objective function preset with historical regulation experience The dynamic weights of each objective function in the running scenario before calculation. :

[0139] ;

[0140] in, For the previous running scenario Next Dynamic weights of each objective function; Traverse the index for the target function. For the first The base weights of the objective function for each iteration For the first The priority coefficient of the cooperative response of the objective function of each iteration; For the previous running scenario Next The basic weights of each objective function; For the previous running scenario Next The priority coefficients of the collaborative responses of each objective function;

[0141] Based on the normalized objective function matrix, dynamic weights, and constraint violation degrees of each candidate solution. Calculate the overall score for each candidate regulation strategy:

[0142] ;

[0143] in, For the first The overall score of each candidate regulation strategy For the first Dynamic weights of each objective function For the first The candidate solution at the th... Normalized values ​​for each objective dimension. To constrain the penalty coefficient for violation ( (Preset by the safety margin requirements for regulation) For the first The degree of constraint violation of each candidate solution;

[0144] The candidate solution with the highest comprehensive score and the constraint violation degree that meets the preset conditions is selected as the optimal control strategy.

[0145] The working principle and beneficial effects of the above technical solution are as follows:

[0146] Since the original calculated values ​​of each candidate solution in the Pareto optimal solution set obtained by multi-objective optimization have different physical dimensions and numerical magnitudes in different objective function dimensions (for example, energy utilization efficiency is represented by a dimensionless ratio while grid interaction cost is represented by monetary units), if the original values ​​are directly weighted and superimposed, the objective function with a larger numerical magnitude will naturally dominate the comprehensive scoring result, causing a systematic suppression of other objectives and making multi-objective decision-making lose objectivity. To this end, this invention first performs range normalization on the original calculated values ​​of each candidate solution in each objective function dimension. Using the difference between the statistical maximum and minimum values ​​of the objective function in the Pareto optimal solution set as the denominator, the original calculated values ​​are mapped to the interval [0,1]. Regardless of whether the original optimization direction is maximization or minimization, the normalized values ​​satisfy the unified semantic that "the larger the value, the better the performance of that dimension," thereby eliminating dimensional differences and unifying the optimization direction, creating a reliable comparison basis for subsequent weighted superposition. The normalized values ​​are then filled into a matrix with the candidate solution index as the row and the objective function index as the column, so that subsequent scoring calculations can be efficiently completed in batches in the form of matrix-vector multiplication, avoiding the need to repeatedly query the objective function value for each candidate solution.

[0147] In the weight determination stage, this invention abandons the traditional fixed-weight method. The fixed-weight method implicitly assumes that the importance of each control objective remains constant across different operating periods. However, in reality, the user side of the power system exhibits significant time-varying operational scenarios: when peak photovoltaic output and peak load coincide, the urgency of the grid absorption efficiency target far outweighs the cost target; during periods of tight grid supply and demand, the priority of the supply-demand balance target needs to be significantly increased. Fixed weights cannot capture these scenario dependencies, inevitably leading to a significant deviation between the weight allocation in some control cycles and actual operational needs. This invention collects system operation status data in real time through a digital twin platform. Key state variables that comprehensively reflect the current system operation status, such as renewable energy output rate, load rate, energy storage charge status, and grid supply and demand tension, are organized into scene feature vectors. Then, the distance between this feature vector and the center vectors of each typical scenario (determined offline by cluster analysis of historical operation data, representing several typical operating conditions that have repeatedly occurred in the system's history) pre-stored in the linkage rule base is calculated. The typical scenario number corresponding to the smallest distance is taken as the current operating scenario label. This nearest neighbor matching mechanism is mathematically equivalent to merging the current operating state to the most similar historical typical operating condition, giving the scene identification results an interpretable physical correspondence. The linkage rule base is queried using the scene label as an index to obtain the control strategies developed by domain experts based on various typical scenarios. A pre-defined collaborative response priority coefficient is used to quantitatively characterize the additional urgency amplification factor of each control objective relative to its basic importance under a specific operating scenario. Then, the collaborative response priority coefficient is multiplied by the basic weights of each objective function preset by historical control experience and normalized to calculate the dynamic weights of each objective function under the current operating scenario. This normalization step ensures that the sum of the dynamic weights is always 1, which meets the consistency requirement of the probabilistic interpretation of weighted scoring. At the same time, it organically integrates the prior experience information carried by the basic weights with the scenario-related priority adjustment information in a product form. The basic weights provide robust priors across scenarios and prevent the priority coefficients from degenerating into single objectives due to excessively large values ​​in individual extreme scenarios. The priority coefficients, on the other hand, endow the weights with the ability to dynamically respond to scenarios and prevent the fixed basic weights from becoming mismatched under special operating conditions.

[0148] In the comprehensive scoring calculation stage, this invention performs an inner product between the normalized row vector of each candidate solution in the normalized objective function matrix and the dynamic weight vector to obtain a weighted score reflecting the multi-objective comprehensive performance of the candidate solution. This weighted score is then subtracted from the constraint violation penalty term, which is proportional to the constraint violation penalty coefficient, to form the comprehensive score. The constraint violation degree measures the degree to which a candidate solution violates system operating constraints (such as equipment power limits, energy storage charge state boundaries, line transmission capacity, etc.). The penalty coefficient ranges from [10, 100], preset according to the regulatory safety margin requirements: a larger value is used when the system safety margin requirements are strict, so that the comprehensive score of a candidate solution with slight violations is significantly lower than that of a fully feasible solution, thus creating strong inhibition of constraint violations at the decision-making level; a smaller value is used when the safety margin requirements are relatively lenient, allowing slightly non-compliant solutions with excellent comprehensive performance to retain a moderate chance in the competition, balancing regulatory benefits and safety margin. Finally, this invention selects the solution with the maximum comprehensive score and constraint violation degree that meets the preset conditions (satisfying all constraints, i.e....). Extract the corresponding distributed power generation output regulation scheme, energy storage charging and discharging control scheme, flexible load transfer scheme and grid interaction power allocation scheme from the candidate solutions of the problem, and use them as the optimal control strategy output to ensure that the selected strategy achieves the optimization of multi-objective cooperative performance under the current operating scenario while meeting the system safety constraints.

[0149] For example, taking the control cycle of a user-side integrated energy system in an industrial park at 10:00 AM on a certain day as an example, the complete numerical calculation process of the optimal control strategy decision-making method described in this invention is illustrated. The system sets four objective functions: maximizing energy utilization efficiency (…). (Dimensionless, larger is better), minimizing grid interaction costs ( (yuan / cycle, the smaller the better), minimizing supply and demand balance deviation ( (kWh, the smaller the better), minimizing carbon emission intensity ( (kg / kWh, the smaller the better). The Pareto optimal solution set output by the multi-objective optimization solver during this regulation cycle contains a total of There are 10 candidate solutions.

[0150] The raw computational values ​​of all five candidate solutions across the four objective function dimensions are retrieved from the Pareto optimal solution set. The raw values ​​of the five candidate solutions are as follows: of to The prices are 0.923 yuan, 386.4 yuan, 52.3 kWh, and 0.418 kg / kWh, respectively. The values ​​are 0.876, 298.5 yuan, 27.4 kWh, and 0.361 kg / kWh. The prices are 0.841, 247.3 yuan, 14.2 kWh, and 0.312 kg / kWh, respectively. The prices are 0.813, 220.4 yuan, 9.1 kWh, and 0.281 kg / kWh, respectively. The prices are 0.801, 214.7 yuan, 8.6 kWh, and 0.271 kg / kWh.

[0151] Calculate the extreme values ​​for each dimension of the objective function within the Pareto optimal solution set:

[0152] ;

[0153] ;

[0154] ;

[0155] ;

[0156] With candidate solutions Example to demonstrate the normalization calculation process:

[0157] ;

[0158] ;

[0159] ;

[0160] ;

[0161] The normalized objective function matrix is ​​calculated sequentially for all five candidate solutions. Each line is as follows: Behavior [0, 0, 0, 0]; Behavior [0.385, 0.512, 0.570, 0.388]; Behavior [0.672, 0.810, 0.872, 0.721]; Behavior [0.902, 0.966, 0.991, 0.932]; Behavior [1, 1, 1, 1]. The corresponding extreme solution in the Pareto front has the highest energy efficiency but also the highest cost, with all normalized values ​​being 0. The extreme solution, which corresponds to the lowest cost but also the lowest energy efficiency, has a normalized value of 1 for all of them; the other candidate solutions fall between the two, reflecting the trade-offs between the objectives of the Pareto solution set.

[0162] The digital twin platform reads the system's real-time operating status data at 10:00: actual photovoltaic output power 387kW, output rate... The actual load on the user side is 423kW, and the load factor is... Energy storage state of charge The power grid supply and demand tension index is normally set to 0. Construct the feature vector for the current operating scenario: ;

[0163] The linkage rule base pre-stores 6 typical scenario center vectors, which are as follows: Scenario 1 (cloudy day, low light, high load). Scenario 2 (High output and medium load on sunny days) Scenario 3 (High output, low load at noon) Scenario 4 (Peak load and grid strain) Scenario 5 (High nighttime load and strained power grid) Scenario 6 (Medium output during transition period) .

[0164] calculate Euclidean distances between the vectors and the center vectors of each typical scene: , , , , , Of the 6 distance values To find the minimum value, the current running scenario label is determined. This corresponds to the typical scenario of "high output and medium load on sunny days". The physical meaning is that the current photovoltaic output is relatively sufficient, the load is at a medium level, the energy storage is half charged, and the grid is operating normally. In this scenario, priority should be given to the efficiency of renewable energy consumption and carbon emission control.

[0165] With scene tags Query the linkage rule base and read the priority coefficients of the collaborative responses of each objective function corresponding to scenario 2: (Energy utilization efficiency is significantly improved due to the high pressure on energy consumption during sunny, high-output scenarios.) (Grid interaction costs: Currently, the grid is operating normally and has sufficient output, resulting in relatively low cost pressure; therefore, the priority should be appropriately reduced.) (There is a supply-demand imbalance, and a certain mismatch between output and load. Maintain a medium priority.) (Carbon emission intensity; when renewable energy output is high, carbon emission control benefits are prominent and priority is high).

[0166] The basic weights of the four objective functions preset based on historical regulation experience are: , , , Calculate the numerator of the dynamic weights for each objective function. :

[0167]

[0168] ;

[0169]

[0170] ;

[0171] The denominator is the sum of the numerators:

[0172]

[0173] Dynamic weights for each objective function: .

[0174] Compared with the basic weights, the weight of energy utilization efficiency increased from 0.30 to 0.403, the weight of carbon emission intensity increased from 0.20 to 0.224, while the weight of grid interaction cost decreased from 0.25 to 0.149, reflecting the regulation logic of prioritizing the consumption of renewable energy while taking into account carbon emission control in the scenario of "high output and medium load on sunny days".

[0175] This embodiment pre-sets a constraint violation penalty coefficient. This corresponds to a moderate safety margin requirement. The constraint violation degree of each candidate solution is calculated synchronously by the multi-objective optimization solver when outputting the Pareto solution set. , , , , . There is a minor constraint violation. This indicates that the candidate solution has a slight out-of-bounds error in the energy storage charging and discharging power boundary constraints, and the out-of-bounds error is converted to a per-unit value of 0.012.

[0176] The comprehensive score for each candidate solution is calculated using the following formula:

[0177] ;

[0178] The dot product of the normalized row vector of each candidate solution and the dynamic weight vector [0.403, 0.149, 0.224, 0.224] is then subtracted, and the calculation is performed one by one as follows: , , , , The maximum overall score is Corresponding candidate solutions Its degree of constraint violation The pre-defined fully feasible requirement is met. Note the candidate solutions. Although they are close in normalization performance However, due to the violation of the constraint penalty item The deduction caused the overall score to drop from 0.939 (excluding penalties) to 0.339, significantly lower than the previous score. The value of 0.749 reflects the penalty coefficient. It has an effective deterrent effect on behaviors that violate constraints.

[0179] Therefore, candidate solutions are selected. The corresponding distributed power generation output adjustment scheme (photovoltaic full output of 387kW, no reduction), energy storage charging and discharging control scheme (energy storage discharges at 68kW power, and the state of charge decreases from 0.61 to 0.60), flexible load transfer scheme (transferring 36kW interruptible air conditioning load to the second peak period), and grid interaction power allocation scheme (purchasing 16kW from the grid to compensate for the supply and demand gap) are issued and implemented as the optimal control strategy for this control cycle.

[0180] This invention eliminates the dimensional differences and directional inconsistencies of the original values ​​of multiple objectives through range normalization, ensuring the objective comparability of the comprehensive score. Through a scenario-label-driven dynamic weighting mechanism, it achieves real-time adaptive adjustment of control priorities according to the operating scenario, overcoming the systematic mismatch defect of fixed weights in time-varying operating scenarios. By explicitly incorporating constraint violation penalties into the comprehensive score, a quantitative penalty mechanism for safety constraints is established at the decision-making level, avoiding the potential boundary default risks that may remain if only optimization layer constraints are used. The synergistic effect of these three aspects enables the final selected control strategy to simultaneously possess multi-objective balance, scenario adaptability, and operational safety, significantly improving the intelligent control level of the user-side integrated energy system.

[0181] In this embodiment, driving each physical device to perform corresponding control operations further includes:

[0182] Based on the current operating status of user-side energy equipment, analyze energy regulation projects to determine the current regulation information of energy regulation projects, including the output adjustment progress of distributed power sources, the execution progress of energy storage charging and discharging, the completion degree of flexible load transfer, and the power matching degree of grid interaction;

[0183] Based on the current regulation information of energy regulation projects, energy regulation quality analysis is conducted to obtain regulation quality evaluation factors, including energy utilization efficiency compliance rate, grid interaction cost control accuracy, energy storage SOC maintenance stability, and voltage / frequency fluctuation amplitude. Regulation quality assessment data is calculated based on the aforementioned regulation quality evaluation factors to obtain regulation quality analysis data.

[0184] By combining the preset energy regulation standards, the regulation quality analysis data is standardized and analyzed to determine whether the regulation quality analysis data meets the preset energy regulation standards, and the user-side energy regulation quality analysis results are obtained.

[0185] When the user-side energy regulation quality analysis result is that the regulation quality analysis data does not meet the preset energy regulation standard, the cause analysis is performed based on the regulation quality analysis data to determine the reasons why the regulation quality analysis data does not meet the preset energy regulation standard. The reasons for the non-standard energy regulation are obtained, including excessive deviation in the output prediction of distributed power sources, attenuation of energy storage charging and discharging efficiency, lag in the response of flexible loads, and changes in grid interaction constraints.

[0186] Based on the reasons for irregular energy regulation, and in conjunction with the energy regulation planning scheme, we analyze whether the reasons for irregular energy regulation can be corrected during the regulation implementation process, and obtain the results of the cause analysis and judgment.

[0187] In this embodiment, obtaining the cause analysis and judgment result also includes:

[0188] When the cause analysis results indicate that the irregularities in energy regulation can be corrected during the regulation execution process, regulation correction information is generated based on the causes of the irregularities in energy regulation.

[0189] Simultaneously, based on the progress deviation of the energy regulation task and the real-time monitoring of equipment operating status data, corresponding regulation operation instructions are generated. The regulation correction information is used to correct and adjust the corresponding regulation operation instructions, and the energy equipment is guided to adjust its working strategy according to the corrected and adjusted regulation operation instructions.

[0190] When the cause analysis results indicate that the non-standard energy regulation cannot be corrected during the regulation execution process, a corresponding regulation operation instruction is generated based on the progress deviation of the energy regulation task and the real-time monitored equipment operating status data.

[0191] Simultaneously, the difference between the control quality analysis data and the preset energy control standard is determined, and a danger warning information is generated based on the difference data. While guiding the energy equipment to adjust its working strategy according to the control operation instructions, a warning prompt is issued based on the danger warning information.

[0192] In this embodiment, the execution of the control strategy further includes:

[0193] During the control and control process, the execution feedback data of each device is collected in real time, and the feedback data is transmitted to the digital twin model and subjected to deviation analysis with the model prediction data;

[0194] If the deviation value is greater than the preset deviation threshold, the parameters of the physical mapping layer and behavioral mapping layer of the digital twin model and the constraints of the optimized model are corrected based on the feedback data, and the corrected control strategy is regenerated.

[0195] If the deviation value is less than or equal to the preset deviation threshold, the current control strategy remains unchanged, and the operating parameters of the digital twin model are updated through an incremental update algorithm to achieve continuous optimization and dynamic adaptation of the control strategy.

[0196] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A digital twin-based collaborative method for user-side energy optimization and control in power systems, characterized in that, Includes the following steps: Step 1: Acquire energy data from the user side of the power system and preprocess it to obtain the effective energy data after preprocessing. Combine the physical topology of the user-side energy system to construct a digital twin model of the user-side energy. Step 2: Build a timestamp alignment engine to accurately align the preprocessed data with the timeline of the digital twin model, driving the geometric state, physical operating state, behavior prediction results of the user-side energy digital twin model to synchronize with the physical entity in real time; Step 3: Establish a collaborative decision-making framework within the user side and between the user side and the power grid side. Combine the user-side energy digital twin model to build a linkage rule base for each twin unit, clarify the linkage triggering conditions, and define the collaborative response priority of each twin unit under different operating scenarios. Step 4: Based on the real-time state simulated by the digital twin model, combined with the collaborative decision-making framework, determine the multi-objective optimization control system, construct the multi-objective optimization model, obtain the optimal solution set that satisfies the constraints based on the multi-objective optimization model, and select the optimal control strategy from the optimal solution set according to the priority of collaborative response. Step 5: Based on the optimal control strategy, send control commands to the distributed power controller, energy storage charge and discharge manager, flexible load regulator and grid interaction terminal in the user-side energy system to drive each physical device to perform the corresponding control operation.

2. The digital twin-based collaborative power system user-side energy optimization and control method as described in claim 1, characterized in that, The acquisition and preprocessing of energy data from the user side of the power system specifically includes: Sensors and smart metering devices are deployed at key locations such as the output end of distributed energy equipment, the charging and discharging interface of energy storage equipment, the control end of electrical equipment, grid interaction nodes, and the user side to collect real-time operating data. Extract data features from real-time operational data, and classify the collected real-time operational data based on the data features to obtain sub-operational datasets. Perform corresponding preprocessing on each sub-operational dataset based on the data features to generate standardized effective energy data.

3. The digital twin-based collaborative power system user-side energy optimization and control method as described in claim 1, characterized in that, The construction of the user-side energy digital twin model also includes: Obtain the physical entity information of each twin unit, perform layered modeling for each twin unit, and based on the layered modeling requirements, construct the geometric mapping layer, physical mapping layer, behavioral mapping layer and behavioral mapping layer of each twin unit using a four-layer mapping architecture; The geometry mapping layer is used to restore the spatial layout, installation location, and energy transmission link topology of the corresponding equipment. The physical mapping layer is used to establish a model of the device's operating mechanism based on the law of conservation of energy and the principle of electromagnetic induction. The behavior mapping layer is used to fuse historical and real-time data and employ machine learning algorithms to build a predictive model of device operating behavior. The rule mapping layer is used to embed power grid security rules, energy trading rules, and user energy preference rules.

4. The power system user-side energy optimization and control method based on digital twin collaboration as described in claim 3, characterized in that, The specific modeling process for each twin unit includes: A geometric mapping layer is constructed based on the installation location of the load equipment and the network topology to restore the spatial layout of flexible and rigid loads and the power link topology. Based on the law of energy conservation and load energy consumption characteristics, a physical mapping layer is constructed to establish a flexible load power regulation model and a rigid load constant power model. By integrating historical energy consumption data with real-time environmental data, a behavior mapping layer is constructed to generate a load power prediction model; A rule mapping layer is constructed by embedding user energy consumption preferences and grid constraint rules, and binding user energy consumption time period thresholds, comfort constraints and grid voltage and frequency limit rules; A geometric mapping layer is constructed based on the installation coordinates and grid connection point of the distributed power source to reconstruct the spatial deployment of the distributed power source and its connection topology with the power grid. A physical mapping layer is constructed based on the principle of electromagnetic induction and the correlation with environmental parameters to establish photovoltaic power output model and wind turbine power output model; By integrating historical power output data and meteorological forecast data, a behavior mapping layer is constructed to generate a distributed power generation output prediction model; An embedded grid connection rule and energy trading rule are used to construct a rule mapping layer, which binds grid connection voltage / frequency constraints, peak-valley electricity price rules and ancillary service compensation rules. A geometric mapping layer is constructed based on the cabinet layout and charging / discharging interface topology of the energy storage battery pack to restore the connection links between the energy storage device and the load, distributed power source, and power grid. A physical mapping layer is constructed based on electrochemical principles and charge-discharge characteristics, and an energy storage charge-discharge model considering charge-discharge efficiency and remaining capacity constraints is established. By integrating historical charging and discharging data with source-load prediction data, a behavior mapping layer is constructed to generate an energy storage charging and discharging strategy optimization model; An embedded energy storage operation safety rules and trading rules are used to construct a rule mapping layer, which binds the upper and lower limits of the remaining energy storage capacity, the charging and discharging time limit, and the peak-valley electricity price arbitrage rules. A geometric mapping layer is constructed based on the connection topology between the user-side grid connection point and the distribution network, and the installation location of metering equipment, to restore the interaction link between the user side and the power grid and the layout of metering nodes; A physical mapping layer is constructed based on the power transmission principle and the power grid impedance characteristics to establish a power grid interaction calculation model. By integrating historical interactive power data with power grid dispatch instructions, a behavior mapping layer is constructed to generate a power grid interactive power prediction model; An embedded rule mapping layer is constructed using power grid dispatch rules and transaction rules, binding power grid dispatch command constraints, power purchase and sale price rules, and power grid safety and stability operation rules.

5. The power system user-side energy optimization and control method based on digital twin collaboration as described in claim 4, characterized in that, A physical mapping layer is constructed based on electrochemical principles and charge-discharge characteristics. An energy storage charge-discharge model considering charge-discharge efficiency and remaining capacity constraints is established, including: Piecewise fitting is used to obtain the open-circuit voltage measurement data of the energy storage battery in each state of charge range, which was collected in advance through a piecewise constant current static calibration experiment, and to obtain a table of correspondence between open-circuit voltage and state of charge covering the entire state of charge range. The electrochemical impedance spectroscopy data collected in advance through offline pulse charge-discharge calibration experiments under multiple temperature ranges and multiple charge-discharge rates are processed for equivalent circuit parameter identification. Combined with the corresponding relationship table, the ohmic internal resistance, polarization internal resistance and polarization capacitance parameter values ​​under each temperature range and each charge-discharge rate combination are extracted respectively, and an electrochemical equivalent circuit parameter lookup table indexed by temperature range and charge-discharge rate is obtained. Temperature correction modeling is performed on the variation of each impedance parameter with temperature in the electrochemical equivalent circuit parameter lookup table. Each impedance parameter is expressed as a continuous variation function with respect to the current battery temperature, and a temperature adaptive equivalent circuit model is obtained that can dynamically output the current equivalent circuit parameters based on the real-time temperature. The real-time synchronized charging and discharging current time series and battery surface temperature data are input into the temperature adaptive equivalent circuit model. The battery terminal voltage response and the current state of charge are jointly recursively estimated. The estimation results are corrected in parallel based on the energy loss difference corresponding to the charging and discharging processes, and a dynamic correspondence table of charging and discharging efficiency is obtained with the current charging and discharging rate, current temperature and current state of charge as indexes. The dynamic correspondence table of charging and discharging efficiency is superimposed with the upper and lower limits of the state of charge that the energy storage device is allowed to operate, the upper limit of the maximum charging and discharging power allowed to be output under each operating condition, and the maximum energy throughput allowed in a single scheduling cycle. The compliance judgment of the charging and discharging operation under each operating condition is carried out, and infeasible operating conditions that exceed the boundary are eliminated to obtain the compliance and feasibility domain dataset of charging and discharging operation under each operating condition. For the compliant and feasible domain dataset, a battery life decay mapping relationship is introduced, with the single charge-discharge cycle depth and charge-discharge rate as inputs and the corresponding battery capacity decay rate as output. The life loss cost corresponding to each charge-discharge operation scheme in the compliant and feasible domain is quantitatively labeled, and an energy storage charge-discharge model that simultaneously includes operating efficiency information, remaining capacity safety boundary information, and life loss cost information is obtained.

6. The digital twin-based collaborative power system user-side energy optimization and control method as described in claim 1, characterized in that, The energy storage twin unit also includes: By integrating the battery characteristic model of energy storage devices with real-time operating data, a model for simulating the charge and discharge characteristics of energy storage devices, predicting SOC, and assessing their lifespan is constructed. Key features of the target operation data of energy storage equipment are extracted, and the feature data are clustered to obtain voltage operation data groups and current operation data groups of energy storage equipment under different operating conditions. Based on the clustering results, the parameters of the energy storage twin unit model under different operating conditions are calibrated differently.

7. The digital twin-based collaborative power system user-side energy optimization and control method as described in claim 1, characterized in that, The optimal solution set satisfying the constraints is obtained based on a multi-objective optimization model. The optimal control strategy is then selected from the optimal solution set according to the priority of the coordinated response, including: Obtain the original calculated value of each candidate solution in the optimal solution set under each objective function, and normalize the original calculated value to obtain the normalized value in each objective function dimension. Using the candidate solution index as the row index and the objective function index as the column index, all normalized values ​​are filled according to their row and column positions to construct a normalized objective function matrix; Construct a feature vector for the current operating scenario based on real-time system operating status data collected from the digital twin platform. The label of the current running scenario is determined based on the feature vector of the current running scenario and the pre-stored center vectors of each typical scenario; Based on the current running scenario tags and linkage rule base, determine the collaborative response priority coefficient of each objective function; Based on the coordinated response priority coefficient and the basic weights of each objective function preset by historical control experience, the dynamic weights of each objective function under the previous running scenario are calculated. The comprehensive score of each candidate control strategy is calculated based on the normalized objective function matrix, dynamic weights, and constraint violation degree of each candidate solution. The candidate solution with the highest comprehensive score and the constraint violation degree that meets the preset conditions is selected as the optimal control strategy.

8. The power system user-side energy optimization and control method based on digital twin collaboration as described in claim 1, characterized in that, The method of driving each physical device to perform corresponding control operations also includes: Analyze energy regulation projects based on the current operating status of user-side energy equipment to determine the current regulation information for energy regulation projects; Based on the current control information of energy control projects, conduct energy control quality analysis, obtain control quality evaluation factors, calculate control quality assessment data, and obtain control quality analysis data; The regulation quality analysis data is standardized and analyzed in conjunction with the preset energy regulation standards to determine whether the regulation quality analysis data meets the preset energy regulation standards and to obtain the energy regulation quality analysis results on the user side. When the user-side energy regulation quality analysis result is that the regulation quality analysis data does not meet the preset energy regulation standard, the cause analysis is performed based on the regulation quality analysis data to determine the cause of the non-standard energy regulation. Based on the reasons for irregular energy regulation, and in conjunction with the energy regulation planning scheme, we analyze whether the reasons for irregular energy regulation can be corrected during the regulation implementation process, and obtain the results of the cause analysis and judgment.

9. The power system user-side energy optimization and control method based on digital twin collaboration as described in claim 8, characterized in that, The obtained cause analysis and judgment results also include: When the cause analysis results indicate that the irregularities in energy regulation can be corrected during the regulation execution process, regulation correction information is generated based on the causes of the irregularities in energy regulation. Simultaneously, based on the progress deviation of the energy regulation task and the real-time monitoring of equipment operating status data, corresponding regulation operation instructions are generated, and the corresponding regulation operation instructions are corrected and adjusted using regulation correction information. When the cause analysis results indicate that the non-standard energy regulation cannot be corrected during the regulation execution process, a corresponding regulation operation instruction is generated based on the progress deviation of the energy regulation task and the real-time monitored equipment operating status data. Simultaneously, the difference between the control quality analysis data and the preset energy control standard is determined, and a danger warning information is generated based on the difference data. While guiding the energy equipment to adjust its working strategy according to the control operation instructions, a warning prompt is issued based on the danger warning information.

10. The digital twin-based collaborative power system user-side energy optimization and control method as described in claim 1, characterized in that, The implementation of the control strategy also includes: During the control and control process, the execution feedback data of each device is collected in real time, and the feedback data is transmitted to the digital twin model and subjected to deviation analysis with the model prediction data; If the deviation value is greater than the preset deviation threshold, the parameters of the physical mapping layer and behavioral mapping layer of the digital twin model and the constraints of the optimized model are corrected based on the feedback data, and the corrected control strategy is regenerated. If the deviation value is less than or equal to the preset deviation threshold, the current control strategy remains unchanged, and the operating parameters of the digital twin model are updated through an incremental update algorithm.

Images