A method for constructing a new energy consumption software model based on a simulation platform

Abstract

The present invention relates to the technical field of new energy power systems, and discloses a method for constructing a new energy consumption software model based on a simulation platform. The method first divides the new energy consumption system units and determines their topological structures and operating parameters, and constructs a multi-physics field coupling simulation model; then, the model is solved in a distributed and parallel manner, a global optimization goal is defined, and a joint optimization and network reconstruction model is constructed. Deep reinforcement learning algorithms and graph neural network optimization are respectively adopted, and an optimization scheme is generated through multi-agent collaboration; then, it is verified whether the optimization goal converges. If converged, the control parameters are output, otherwise the optimization variables are updated to continue optimization; finally, the optimization parameters are imported into the calibration model of the digital twin platform integrated with blockchain technology and the hyperparameters are adjusted. This method improves the accuracy of the new energy consumption system model, optimizes system operation, enhances collaborative stability, and promotes the efficient use of new energy and the sustainable development of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy power systems, and in particular to a method for constructing a new energy consumption software model based on a simulation platform. Background Art

[0002] With the growing global demand for clean energy, the proportion of renewable energy in the power system continues to rise. Renewable energy sources such as solar and wind power offer significant advantages in being clean and renewable, and are crucial for alleviating the energy crisis and reducing environmental pollution. However, the large-scale integration of renewable energy sources presents numerous challenges to the stable operation and efficient absorption of power systems.

[0003] Renewable energy sources exhibit significant intermittent and volatile power generation characteristics. For example, photovoltaic power generation directly depends on sunlight intensity, generating high levels of power during abundant daylight hours and completely shutting down at night. Furthermore, weather fluctuations, such as cloud cover, can cause significant fluctuations in photovoltaic power generation. Wind power generation is similarly affected by wind speed, with unstable wind speeds causing erratic fluctuations in turbine power output. This unstable generation significantly conflicts with the relatively stable supply-demand balance of traditional power systems, creating significant challenges for power balance and frequency stability control. If not effectively addressed, these fluctuations will frequently trigger grid frequency fluctuations, impacting power supply quality and even threatening the safe and stable operation of the power grid.

[0004] Energy storage technology is crucial for the absorption of new energy, but current energy storage systems face numerous challenges. For one thing, energy storage equipment is expensive. Whether it's lithium-ion batteries, flow batteries, or supercapacitors, the initial investment cost is high. This significantly limits the large-scale application of energy storage systems, preventing them from fully playing their role in regulating fluctuations in renewable energy output. Furthermore, existing energy storage charge and discharge management strategies are suboptimal and fail to fully account for the lifespan degradation characteristics of energy storage equipment. Irrational charge and discharge strategies accelerate the aging of energy storage equipment, shorten its service life, further increase the overall cost of the energy storage system, and reduce its economic viability within the power system.

[0005] In terms of load management, traditional methods of electricity load management are relatively crude, and users are not very motivated to participate in demand response. The response mechanism for electricity users to price fluctuations is imperfect, and they cannot flexibly adjust their electricity consumption based on real-time price signals. When there is an oversupply of renewable energy, it is difficult to effectively guide users to increase their electricity load; and when there is a shortage of renewable energy, it is difficult to promptly encourage users to reduce non-essential electricity use. This leads to inefficient allocation of electricity resources and low efficiency in the absorption of renewable energy.

[0006] The transmission link in the power grid also faces challenges. With the widespread distribution of renewable energy power generation sites, the grid topology is becoming increasingly complex, making traditional grid planning and operation methods difficult to adapt. Line impedance changes dynamically due to environmental factors such as temperature, leading to increased network losses and reduced transmission efficiency. Furthermore, existing grid topology optimization methods often fail to accurately and in real time address the complex and changing operating conditions of power systems, making it difficult to maximize transmission efficiency while ensuring safe and stable grid operation.

[0007] Existing technologies and models for renewable energy consumption suffer from numerous shortcomings. Some models fail to fully consider the various factors influencing renewable energy generation during construction, resulting in low prediction accuracy and an inability to provide a reliable basis for power system scheduling decisions. Some optimization algorithms suffer from low computational efficiency, making them inefficient and unable to meet the demands of real-time optimization of large-scale power systems. Furthermore, the lack of effective coupling and collaborative optimization mechanisms between different unit models hinders the optimal operation of the renewable energy consumption system as a whole. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for constructing a new energy consumption software model based on a simulation platform to solve the problems raised in the above background technology.

[0009] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for constructing a new energy consumption software model based on a simulation platform, the method comprising:

[0010] Step S1: Divide the new energy consumption system into power generation units, energy storage units, load units, and transmission units, and determine the topology and operating parameters of each unit;

[0011] Step S2: Based on real-time meteorological data and equipment operation data, a power generation unit output prediction model, an energy storage unit charge and discharge dynamic model, a load unit demand response model, and a transmission unit network loss model are constructed respectively. Each model is connected through a data interface according to the actual energy flow relationship to form a multi-physics field coupling simulation model of the new energy consumption system;

[0012] Step S3: performing distributed parallel solving on the multi-physics field coupling simulation model to obtain initial simulation results;

[0013] Step S4: defining a global optimization goal for the new energy consumption system, including at least one of minimizing the wind and solar power curtailment rate and maximizing the grid stability index;

[0014] Step S5: Construct a joint optimization model for the power generation unit and the energy storage unit, using power generation scheduling and energy storage charge and discharge rates as optimization variables. Combined with the initial simulation results of step S3 as the feasible solution boundary, a deep reinforcement learning algorithm is used to perform policy iteration to generate a first optimization solution.

[0015] Step S6: Determine whether the load unit has introduced a dynamic electricity price response mechanism. If so, modify the load demand curve based on the first optimization scheme in step S5, and return to step S5. If not, execute step S7.

[0016] Step S7: Construct a transmission unit network reconstruction model, use node voltage deviation and line load rate as constraints, use a graph neural network to dynamically optimize the grid topology, and generate a second optimization solution;

[0017] Step S8: Integrate the optimization solutions of step S5 and step S7, and coordinate the control instructions of each unit through the multi-agent collaborative optimization algorithm;

[0018] Step S9: Verify whether the global optimization goal meets the convergence conditions; if so, output the optimization solution as the simulation platform control parameters; if not, return to step S5 to update the optimization variable constraint range.

[0019] Preferably, in step S1, the power generation unit includes one or more combinations of photovoltaic arrays, wind turbines, and distributed gas turbines; the energy storage unit includes one or more combinations of lithium-ion batteries, flow batteries, and supercapacitors; and the transmission unit includes one or more combinations of AC transmission lines, DC transmission lines, and flexible power electronic devices.

[0020] The topological structure includes the electrical connection relationship and geographical location distribution of each unit; the operating parameters include equipment rated capacity, efficiency curve, aging coefficient and environmental influencing factors.

[0021] Preferably, the power generation unit output prediction model uses a time convolutional network to fuse meteorological satellite data and historical power series to predict power generation fluctuations within the next 24 hours; the energy storage unit charging and discharging dynamic model is based on the electrochemical mechanism equation and the equivalent circuit model to simulate the life attenuation characteristics under different charging and discharging strategies; the load unit demand response model divides user types through cluster analysis, and generates a load adjustment strategy in combination with the electricity price elasticity matrix; the transmission unit network loss model uses an adaptive finite element method to calculate the dynamic characteristics of line impedance as temperature changes.

[0022] Preferably, in step S3, the distributed parallel solution adopts a task scheduling mechanism based on the Kubernetes containerized architecture, splits the simulation model into independent computing nodes, realizes cross-node data synchronization through message middleware, and utilizes GPU to accelerate the solution of partial differential equations.

[0023] Preferably, in step S5, the deep reinforcement learning algorithm adopts a double-delayed deep deterministic policy gradient (TD3) framework, designs a reward function as a weighted combination of wind and solar power curtailment penalties, energy storage life loss, and grid frequency regulation benefits, and dynamically updates the strategy network parameters through an experience replay buffer.

[0024] Preferably, in step S7, the graph neural network constructs a heterogeneous graph structure with grid nodes as vertices and lines as edges, dynamically allocates line weights using a graph attention mechanism, and optimizes the topological connection scheme through a gradient descent method, so that the transmission efficiency improvement rate after network reconstruction is not lower than a preset threshold.

[0025] Preferably, in step S8, the multi-agent collaborative optimization algorithm adopts Nash equilibrium theory to design the game strategy, each unit agent exchanges local optimization information through a distributed consensus protocol, and uses mixed integer linear programming to solve the Pareto optimal solution.

[0026] Preferably, in step S9, the convergence conditions include that the residual of the objective function is less than a set tolerance, the number of iterations reaches an upper limit, or the fluctuation range of the optimization variable is stable within the neighborhood of the historical optimal solution.

[0027] Preferably, step S10 is also included: importing the optimized control parameters into the digital twin platform, calibrating the simulation model in real time through the virtual-reality linkage mechanism, and dynamically adjusting the model hyperparameters using the Bayesian optimization algorithm.

[0028] Preferably, in step S10, the digital twin platform integrates blockchain technology, stores the simulation results and the hash values of the measured data on the chain, verifies the data consistency through smart contracts, and triggers the model reconstruction process under abnormal working conditions.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] The present invention divides the new energy consumption system into power generation, energy storage, load and transmission units in detail, accurately determines the topological structure and operating parameters of each unit, and constructs a multi-physics field coupling simulation model that can highly restore the actual operating state of the system. For example, the power generation unit output prediction model integrates meteorological satellite data and historical power series, and uses a time convolutional network to predict power generation fluctuations. Compared with traditional methods, the prediction accuracy is significantly improved, and a more reliable basis can be provided for power dispatch in advance. The energy storage unit charge and discharge dynamic model is based on electrochemical mechanisms and equivalent circuit models, accurately simulating the life attenuation characteristics under different charge and discharge strategies, which helps to optimize energy storage management strategies, extend the service life of energy storage equipment, and reduce costs.

[0031] By defining the global optimization objectives of the new energy consumption system and constructing a joint optimization model and network reconstruction model, optimization is performed by combining deep reinforcement learning algorithms and graph neural networks. In terms of the joint optimization of power generation and energy storage, the reward function is designed based on the comprehensive considerations of wind and solar power curtailment penalties, energy storage life loss, and grid frequency regulation benefits. The dual-delayed deep deterministic policy gradient (TD3) framework is used for policy iteration, effectively reducing the wind and solar power curtailment rate, improving grid frequency regulation benefits, and achieving the rational allocation of power generation and energy storage resources. The transmission unit network reconstruction model uses node voltage deviation and line load rate as constraints and uses graph neural networks for dynamic optimization, improving transmission efficiency, reducing network losses, and making grid operation more economical and efficient.

[0032] The multi-agent collaborative optimization algorithm, based on Nash equilibrium theory, enables each unit agent to exchange information through a distributed consensus protocol and solve the Pareto optimal solution using mixed integer linear programming, thus achieving collaborative control among the units. This process not only improves the overall performance of the system but also enhances its stability under complex operating conditions. For example, when renewable energy generation experiences significant fluctuations, the various units can quickly and collaboratively respond. The energy storage unit quickly adjusts power, the load unit adjusts electricity demand according to the dynamic electricity price response mechanism, and the transmission unit optimizes the network topology, ensuring the stable power supply of the power system.

[0033] The application of a digital twin platform and blockchain technology provides strong support for continuous model optimization and data management. Optimized control parameters are imported into the digital twin platform, and the simulation model is calibrated in real time through a virtual-real linkage mechanism. This, combined with a Bayesian optimization algorithm, dynamically adjusts model hyperparameters, enabling the model to adapt to dynamic changes in the power system. Blockchain technology stores hash values of simulation results and measured data on-chain, and uses smart contracts to verify data consistency, ensuring data authenticity and reliability. Under abnormal operating conditions, the model reconstruction process can be triggered promptly, further improving model reliability and the safety of power system operations.

[0034] By optimizing the operation of new energy consumption systems, this invention effectively reduces wind and solar power curtailment, improves the utilization rate of new energy, reduces dependence on traditional energy sources, and reduces energy waste and environmental pollution. Furthermore, rational energy storage management strategies extend the life of energy storage equipment and reduce the cost of energy storage systems. Optimized grid topology reduces line losses and lowers power transmission costs. In the long run, these factors will help promote the sustainable development of the clean energy industry, promote the transformation and upgrading of the energy structure, and contribute positively to achieving global low-carbon energy goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a working principle diagram of the method for constructing a new energy consumption software model based on a simulation platform according to the present invention;

[0036] Figure 2 This is a diagram of the implementation steps for distributed parallel solving;

[0037] Figure 3 Flowchart of the joint optimization algorithm for power generation and energy storage. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] See also Figure 1-3 This paper provides a method for constructing a new energy consumption software model based on a simulation platform, aiming to efficiently solve complex problems in the new energy consumption process and improve the utilization efficiency and stability of new energy in the power system. The overall implementation process is as follows:

[0040] Step S1: Detailedly divide the new energy consumption system into power generation units, energy storage units, load units, and transmission units. For each unit, accurately determine its topology and operating parameters. The topology includes the electrical connections and geographic distribution of each unit. Operating parameters include equipment rated capacity, efficiency curves, aging coefficients, and environmental impact factors. These parameters provide key foundational data for subsequent model construction.

[0041] Step S2: Based on real-time meteorological data and equipment operation data, a generation unit output forecast model, an energy storage unit charge and discharge dynamic model, a load unit demand response model, and a transmission unit network loss model are constructed. These models are connected through data interfaces according to actual energy flow relationships to form a multi-physics coupled simulation model of the new energy consumption system, thereby realistically simulating the energy interactions and physical processes between various units within the system.

[0042] Step S3: Distributed parallel solving technology is used to solve the multi-physics coupled simulation model and obtain initial simulation results. The task scheduling mechanism based on the Kubernetes containerized architecture is used to split the simulation model into independent computing nodes. Cross-node data synchronization is achieved through message middleware, and GPU acceleration is used to solve the partial differential equations, improving solution efficiency.

[0043] Step S4: Clearly define the global optimization goals of the new energy consumption system, such as minimizing the wind and solar power curtailment rate and maximizing the grid stability index, to provide a clear direction and measurement standard for subsequent optimization work.

[0044] Step S5: Construct a joint optimization model for the power generation unit and energy storage unit, using power generation scheduling and energy storage charge and discharge rates as optimization variables. Combined with the initial simulation results from step S3, the feasible solution boundary is determined. A deep reinforcement learning algorithm is used for policy iteration to generate the first optimization solution, achieving coordinated optimization of power generation and energy storage.

[0045] Step S6: Determine whether the load unit has introduced a dynamic electricity price response mechanism. If so, modify the load demand curve based on the first optimization scheme in step S5, and return to step S5 for re-optimization; if not, proceed to step S7.

[0046] Step S7: Construct a transmission unit network reconstruction model, use the node voltage deviation and line load rate as constraints, use the graph neural network to dynamically optimize the grid topology, generate a second optimization solution, and optimize the grid transmission performance.

[0047] Step S8: Integrate the optimization solutions of step S5 and step S7, coordinate the control instructions of each unit through the multi-agent collaborative optimization algorithm, and realize the optimal control of the entire system.

[0048] Step S9: Verify that the global optimization objective meets convergence conditions. Convergence conditions include the objective function residual being less than the set tolerance, the number of iterations reaching the upper limit, or the optimization variable fluctuation range being stable within the neighborhood of the historical optimal solution. If so, the optimization solution is output as the simulation platform control parameters. If not, return to step S5 to update the optimization variable constraints and continue optimization.

[0049] Step S10: Import the optimized control parameters into the digital twin platform. The simulation model is calibrated in real time through a virtual-real linkage mechanism, and the model hyperparameters are dynamically adjusted using a Bayesian optimization algorithm. The digital twin platform integrates blockchain technology, storing hash values of simulation results and measured data on-chain. Smart contracts verify data consistency and trigger model reconstruction under abnormal operating conditions to ensure model accuracy and reliability.

[0050] The implementation of the present invention will be further described below in conjunction with Examples 1 to 4.

[0051] Example 1:

[0052] The power generation unit output prediction model uses a temporal convolutional network (TCN) to fuse meteorological satellite data with historical power series to predict power generation fluctuations within the next 24 hours. Taking the photovoltaic array power generation prediction as an example, meteorological satellite data contains information such as solar radiation intensity and cloud cover, while the historical power series records the actual power generation of the photovoltaic array over the past period of time. After preprocessing, these data are input into the temporal convolutional network. The structure of the temporal convolutional network consists of multiple convolutional layers and pooling layers, and the size and step size of its convolution kernel are optimized according to the characteristics of the data. The convolutional layer is used to extract features from the data to capture the time series features in the data, such as the relationship between the daily variation of solar radiation intensity and the power generation of the photovoltaic array. The output of the prediction model is the predicted value of photovoltaic power generation per hour in the next 24 hours. , , the prediction formula can be expressed as: ,in is the temporal convolutional network model, for Weather satellite data at all times, for The historical power series data corresponding to the time.

[0053] The energy storage unit charge and discharge dynamic model is based on the electrochemical mechanism equation and the equivalent circuit model to simulate the life attenuation characteristics under different charge and discharge strategies. Taking lithium-ion batteries as an example, according to the electrochemical mechanism, the battery charge and discharge process involves chemical reactions, and its capacity change is related to the ion diffusion, charge transfer and other processes inside the battery. The equivalent circuit model equates the battery to a combination circuit of a resistor, capacitor and voltage source. By combining the electrochemical mechanism equation and the equivalent circuit model equation, the battery life at different charge and discharge currents is obtained. Voltage changes under and capacity changes Relationship: , ,in is the open circuit voltage of the battery, is the internal resistance of the battery, is a constant related to battery characteristics, is the number of reaction electrons, is the Faraday constant. By simulating the changes in these parameters under different charge and discharge strategies, the battery life attenuation can be evaluated, such as calculating the battery Remaining capacity after 10 charge and discharge cycles .

[0054] The load unit demand response model divides user types through cluster analysis and generates load adjustment strategies based on the electricity price elasticity matrix. First, a large amount of user electricity consumption data is collected, including information such as electricity consumption time and electricity consumption. Cluster analysis algorithms, such as the K-means clustering algorithm, are used to divide users into different types such as industrial users, commercial users, and residential users. For each user type, an electricity price elasticity matrix is constructed based on historical electricity consumption data and electricity price changes. The elements of the electricity price elasticity matrix are: Indicates the Class user in The calculation formula of the electricity price elasticity coefficient for a period of time is: ,in For the Class user in the Time period due to changes in electricity prices The load change caused by For the Class user in the The initial load of the period, For the The initial electricity price for the time period. When the electricity price changes, the load adjustment amount for each type of user is calculated based on the electricity price elasticity matrix, thereby generating a load adjustment strategy to achieve demand response.

[0055] The transmission unit network loss model uses the adaptive finite element method to calculate the dynamic characteristics of line impedance as temperature changes. The transmission line is divided into multiple finite element units, and the impedance of each unit is and temperature According to transmission line theory, line impedance consists of resistance and reactance ,and With the temperature change, the relationship is as described in Example 1. Through the adaptive finite element method, the density of the finite element grid is automatically adjusted according to the change of the line temperature distribution to improve the calculation accuracy. When calculating the network loss, the current of each line is first calculated. , according to the power loss formula Calculate the loss of each line and then sum the losses of all lines to get the loss of the entire transmission network ,in Indicates the line number.

[0056] Through the above precise model construction method, each unit model can more accurately reflect the actual operation situation, providing a powerful tool for the simulation and optimization of the new energy consumption system.

[0057] Example 2: This example introduces in detail the application of distributed parallel solution technology in solving multi-physics field coupling simulation models of new energy consumption systems.

[0058] When performing distributed parallel solutions for multi-physics coupled simulation models, a task scheduling mechanism based on the Kubernetes containerized architecture is employed. Kubernetes is an open-source container orchestration platform that automates the deployment, scaling, and management of containerized applications. First, the simulation model is split into multiple independent computing nodes, each corresponding to a container. For example, the power generation unit model, energy storage unit model, load unit model, and transmission unit model are encapsulated in separate containers. These containers run in a Kubernetes cluster, where Kubernetes dynamically assigns tasks to each container based on the available resources (such as CPU, memory, and GPU) of each node, ensuring optimal task scheduling.

[0059] Cross-node data synchronization is achieved through message-based middleware. Kafka, a high-throughput distributed messaging system, is used as the messaging middleware. During the simulation process, data needs to be exchanged between unit models. For example, output data from the power generation unit needs to be transmitted to the load unit and energy storage unit models. Kafka acts as a message queue, receiving data from one unit model and accurately delivering it to other unit model containers that need the data, ensuring reliable data transmission between different nodes.

[0060] Use GPU to accelerate the solution of partial differential equations. In the new energy consumption system model, there are some partial differential equations that describe physical processes, such as Maxwell's equations that describe the power transmission process. GPU (graphics processing unit) has powerful parallel computing capabilities and can significantly improve the speed of solving partial differential equations compared to traditional CPUs. Taking the finite element method to solve partial differential equations as an example, the finite element calculation task is assigned to the GPU for execution. On the GPU, by solving the equations of multiple finite element units in parallel, the solution time is greatly shortened. Assuming the original CPU solution time is , the solution time after GPU acceleration is , speedup ratio ,Through actual tests, the acceleration ratio in this system can reach above [X], ,effectively improving the efficiency of solving the simulation model.

[0061] Through the collaborative application of task scheduling based on the Kubernetes containerized architecture, data synchronization with message middleware, and GPU acceleration technology, efficient distributed parallel solution of multi-physics field coupling simulation models is achieved, providing a guarantee for quickly obtaining initial simulation results.

[0062] Example 3: This example focuses on the application of deep reinforcement learning algorithms in the joint optimization model of power generation units and energy storage units, as well as the application of graph neural networks in the network reconstruction model of transmission units. Through these optimization algorithms, the optimized operation of various parts of the system is achieved, and the efficiency of new energy consumption and grid performance are improved.

[0063] When constructing the joint optimization model of power generation unit and energy storage unit, the double-delayed deep deterministic policy gradient (TD3) framework is used for optimization. Taking a regional new energy power generation and energy storage joint system as an example, the optimization variable is power generation scheduling. and energy storage charge and discharge rates , , Design a reward function to optimize the time interval. It is a weighted combination of wind and solar power curtailment penalties, energy storage life loss, and grid frequency regulation benefits. The formula is: ,in 、 、 is the weight coefficient, and Penalties for curtailing wind and solar power and curtailed wind and solar power Related, , is the penalty coefficient; energy storage life loss The life attenuation is determined based on the calculation of the energy storage unit charge and discharge dynamic model, such as , is the coefficient related to life loss, is the energy storage capacity attenuation; grid frequency regulation benefit Frequency deviation from the grid and FM power related, , is the profit coefficient.

[0064] Dynamically update the policy network parameters through the experience replay buffer. During the simulation, the state of each step ,action ,award and the next state The data is stored in the experience replay buffer. When the buffer reaches a certain amount, a batch of data is randomly sampled from the buffer to train the policy network. The policy network updates its parameters based on the sampled data, enabling it to learn a more optimal strategy and generate a first optimization solution. This achieves coordinated optimization of the power generation unit and energy storage unit, reduces wind and solar power curtailment, extends energy storage life, and improves grid frequency regulation benefits.

[0065] When constructing the transmission unit network reconstruction model, the graph neural network is used to dynamically optimize the power grid topology. Taking a city's distribution network as an example, the graph neural network constructs a heterogeneous graph structure with grid nodes as vertices and lines as edges. In the graph neural network, the graph attention mechanism is used to dynamically assign line weights. and the nodes connected to it , by calculating the attention coefficient To determine the edge The weight is calculated as follows: ,in is the weight matrix, 、 、 is the feature vector of the node, For nodes The neighbor node set is optimized by gradient descent method to optimize the topological connection scheme with node voltage deviation. and line load factor The goal is to ensure that the transmission efficiency improvement rate after network reconstruction is not lower than the preset threshold. Transmission efficiency improvement rate The calculation formula is: ,in is the transmission power after network reconstruction, This is the transmission power before network reconstruction. By continuously optimizing topological connections, a second optimization plan is generated to improve the transmission efficiency of the power grid, reduce line losses, and enhance the operational stability of the power grid.

[0066] Through the effective application of these two optimization algorithms, the coordinated optimization of power generation and energy storage in the new energy consumption system and the improvement of grid transmission performance are achieved, providing key technical support for the overall optimized operation of the system.

[0067] Example 4:

[0068] When integrating the optimization solutions from steps S5 and S7, a multi-agent collaborative optimization algorithm is used to coordinate the control instructions of each unit. For example, the power generation unit, energy storage unit, and transmission unit in a new energy consumption system are considered as an agent. Game strategies are designed using Nash equilibrium theory, and the agents in each unit exchange local optimization information through a distributed consensus protocol. Within the distributed consensus protocol, each agent broadcasts its optimization solution (such as the power generation scheduling plan for the power generation unit, the charging and discharging strategy for the energy storage unit, and the network reconfiguration plan for the transmission unit) to other agents. Each agent adjusts its optimization solution based on the information received from other agents and its own objectives and constraints. Mixed integer linear programming is used to solve Pareto optimal solutions to achieve collaborative optimization among the units. For example, the power generation unit aims to maximize power generation revenue, the energy storage unit aims to minimize its own losses, and the transmission unit aims to reduce network losses and ensure power supply reliability. Using mixed integer linear programming, an optimal set of control instructions is found that balances the interests of each unit, thereby improving overall system performance.

[0069] In step S9, it is verified whether the global optimization objective meets the convergence conditions. The convergence conditions include the objective function residual being less than the set tolerance, the number of iterations reaching the upper limit, or the fluctuation range of the optimization variable being stable within the neighborhood of the historical optimal solution. Assume that the objective function is the comprehensive cost function of the new energy consumption system , which includes multiple parts such as power generation cost, energy storage cost, wind and solar curtailment cost, etc. In each iteration, the objective function value of the current iteration is calculated The objective function value of the previous iteration The residual .when Less than the set tolerance When , the residual error of the objective function meets the convergence condition. At the same time, the maximum number of iterations is set , if the number of iterations achieve Regardless of whether the objective function has converged, iterations cease. Furthermore, the fluctuation range of optimization variables (such as power generation scheduling and energy storage charge and discharge rates) is monitored. Convergence conditions are also considered met when the change in the optimization variables over multiple iterations is less than a certain threshold and remains stable within the neighborhood of the historical optimal solution. If the convergence conditions are met, the optimization solution is output as the simulation platform control parameters. If not, the process returns to step S5 to update the optimization variable constraints and continue optimization.

[0070] In step S10, the optimized control parameters are imported into the digital twin platform. The digital twin platform integrates blockchain technology and stores the simulation results and the hash values of the measured data on the chain. Taking a certain power system as an example, in the digital twin platform, the actual power generation of the power generation unit, the charge and discharge status of the energy storage unit, the actual power consumption of the load unit, and the line loss of the transmission unit are collected. At the same time, the simulation model is run to obtain the corresponding simulation results. The hash function (such as the SHA-256 algorithm) is used to perform hash calculations on the measured data and the simulation results to obtain the hash value. and These hash values are stored on the blockchain. The decentralized and tamper-proof nature of the blockchain ensures the security and credibility of the data. Data consistency is verified through smart contracts, and verification rules are pre-set in the smart contracts. and If the values are equal, the data is considered consistent; if they are not, the model reconstruction process for abnormal operating conditions is triggered. During the model reconstruction process, the causes of abnormal data, such as equipment failure and parameter changes, are analyzed, and the model of the new energy consumption system is adjusted and retrained. The Bayesian optimization algorithm is used to dynamically adjust the model hyperparameters. By constructing a probabilistic model of the objective function and predicting the next optimal hyperparameter value based on the existing data, the Bayesian optimization algorithm improves model training efficiency while ensuring model accuracy, thereby enabling the simulation model to better reflect the actual system operation.

[0071] Through the above embodiments, the specific implementation methods of the new energy consumption software model construction method in each key link of the present invention are elaborated in detail. From unit parameter determination, model construction, solution technology, optimization algorithm application to system optimization verification and digital twin application, each part works closely together to realize the efficient simulation, optimization and operation management of the new energy consumption system, providing strong technical support for the widespread application of new energy in the power system.

[0072] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0073] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a new energy consumption software model based on a simulation platform, characterized in that: The method comprises: Step S1: Divide the new energy consumption system into power generation units, energy storage units, load units, and transmission units, and determine the topology and operating parameters of each unit; Step S2: Based on real-time meteorological data and equipment operation data, a power generation unit output prediction model, an energy storage unit charge and discharge dynamic model, a load unit demand response model, and a transmission unit network loss model are constructed respectively. Each model is connected through a data interface according to the actual energy flow relationship to form a multi-physics field coupling simulation model of the new energy consumption system; Step S3: performing distributed parallel solving on the multi-physics field coupling simulation model to obtain initial simulation results; Step S4: defining a global optimization goal for the new energy consumption system, including at least one of minimizing the wind and solar power curtailment rate and maximizing the grid stability index; Step S5: Construct a joint optimization model for the power generation unit and the energy storage unit, using power generation scheduling and energy storage charge and discharge rates as optimization variables. Combined with the initial simulation results of step S3 as the feasible solution boundary, a deep reinforcement learning algorithm is used to perform policy iteration to generate a first optimization solution. Step S6: Determine whether the load unit has introduced a dynamic electricity price response mechanism. If so, modify the load demand curve based on the first optimization scheme in step S5, and return to step S5. If not, execute step S7. Step S7: Construct a transmission unit network reconstruction model, use node voltage deviation and line load rate as constraints, use a graph neural network to dynamically optimize the grid topology, and generate a second optimization solution; Step S8: Integrate the optimization solutions of step S5 and step S7, and coordinate the control instructions of each unit through the multi-agent collaborative optimization algorithm; Step S9: Verify whether the global optimization target meets the convergence condition; if so, output the optimization solution as the simulation platform control parameter; if not, return to step S5 to update the optimization variable constraint range; In step S2, the power generation unit output prediction model uses a time convolutional network to fuse meteorological satellite data with historical power series to predict power generation fluctuations within the next 24 hours; the energy storage unit charge and discharge dynamic model simulates the life attenuation characteristics under different charge and discharge strategies based on electrochemical mechanism equations and equivalent circuit models; the load unit demand response model divides user types through cluster analysis and generates load adjustment strategies in combination with the electricity price elasticity matrix; the transmission unit network loss model uses an adaptive finite element method to calculate the dynamic characteristics of line impedance as it changes with temperature; In step S3, the distributed parallel solution uses a task scheduling mechanism based on the Kubernetes containerized architecture to split the simulation model into independent computing nodes, synchronizes data across nodes through message middleware, and utilizes GPU acceleration to solve partial differential equations. In step S5, the deep reinforcement learning algorithm adopts the dual-delay deep deterministic policy gradient TD3 framework, designs the reward function as a weighted combination of wind and solar power curtailment penalties, energy storage life loss, and grid frequency regulation benefits, and dynamically updates the policy network parameters through the experience replay buffer.

2. The method according to claim 1, characterized in that In step S1, the power generation unit includes one or more combinations of photovoltaic arrays, wind turbines, and distributed gas turbines; the energy storage unit includes one or more combinations of lithium-ion batteries, flow batteries, and supercapacitors; and the transmission unit includes one or more combinations of AC transmission lines, DC transmission lines, and flexible power electronic devices; The topological structure includes the electrical connection relationship and geographical location distribution of each unit; the operating parameters include equipment rated capacity, efficiency curve, aging coefficient and environmental influencing factors.

3. The method according to claim 1, characterized in that In step S7, the graph neural network constructs a heterogeneous graph structure with grid nodes as vertices and lines as edges, dynamically allocates line weights using the graph attention mechanism, and optimizes the topological connection scheme through the gradient descent method, so that the transmission efficiency improvement rate after network reconstruction is not lower than the preset threshold.

4. The method according to claim 1, wherein In step S8, the multi-agent collaborative optimization algorithm adopts Nash equilibrium theory to design the game strategy, each unit agent exchanges local optimization information through a distributed consensus protocol, and uses mixed integer linear programming to solve the Pareto optimal solution.

5. The method according to claim 1, wherein In step S9, the convergence conditions include that the residual of the objective function is less than a set tolerance, the number of iterations reaches an upper limit, or the fluctuation range of the optimization variable is stable within the neighborhood of the historical optimal solution.

6. The method according to claim 1, characterized in that The method also includes step S10: importing the optimized control parameters into the digital twin platform, calibrating the simulation model in real time through the virtual-reality linkage mechanism, and dynamically adjusting the model hyperparameters using the Bayesian optimization algorithm.

7. The method according to claim 6, characterized in that In step S10, the digital twin platform integrates blockchain technology, stores the simulation results and the hash values of the measured data on the chain, verifies the data consistency through smart contracts, and triggers the model reconstruction process under abnormal working conditions.

Images