Micro-grid multi-mode energy optimization scheduling method under zero carbon target

Abstract

The invention relates to the technical field of micro-grid multi-mode energy optimization scheduling under a zero-carbon target, and discloses a micro-grid multi-mode energy optimization scheduling method under the zero-carbon target, which comprises the following steps: establishing a schedulable potential model for various distributed resources, and constructing an equivalent aggregation model in combination with micro-grid topology; the problem that heterogeneous resources are difficult to coordinate in a traditional method is effectively solved, the overall adjustment potential of the system is fully excavated, collaborative optimization is carried out with the operation cost and the minimum carbon emission as multiple targets, it is ensured that economy and environmental protection of the micro-grid are unified under the zero-carbon target, real-time data are input and an optimization algorithm is used for solving, and the optimization efficiency of the micro-grid is improved. And a global optimal scheduling strategy can be quickly generated. The method has the beneficial effects that the operation efficiency and the cooperative control capability of the complex micro-grid are remarkably improved, and reliable technical support is provided for realizing safe, economic and low-carbon energy supply.

Description

Multimodal Energy Optimization Scheduling Method for Microgrids under Zero-Carbon Objective Technical Field

[0001] This invention relates to the field of multimodal energy optimization scheduling technology for microgrids under the zero-carbon objective, and particularly to a multimodal energy optimization scheduling method for microgrids under the zero-carbon objective. Background Technology

[0002] The energy transition led by renewable energy has become an inevitable trend. As a key carrier for integrating distributed energy and improving energy efficiency, microgrids are playing an increasingly important role. However, the optimized scheduling of traditional microgrids faces severe challenges.

[0003] Currently, the internal structure of microgrids is becoming increasingly complex, including various heterogeneous distributed resources such as photovoltaics, wind power, energy storage, and flexible loads. These resources have different characteristics and operating modes, making it difficult for traditional scheduling methods to accurately model and coordinate them in a unified manner. As a result, their adjustable potential has not been fully explored, and the overall operating efficiency of the system is limited. Summary of the Invention

[0004] Therefore, it is necessary to propose a multi-modal energy optimization scheduling method for microgrids under the existing zero-carbon objective, which addresses the existing problem of multi-modal energy optimization scheduling for microgrids.

[0005] A multimodal energy optimization scheduling method for microgrids under a zero-carbon objective is disclosed. The method includes: acquiring historical operating data and resource types of multiple distributed resources within a specified microgrid; generating corresponding schedulable potential calculation target models based on the historical operating data and resource types of each distributed resource; acquiring a topology graph of the specified microgrid and establishing an equivalent aggregation model of the specified microgrid based on the schedulable potential calculation target models and the topology graph; establishing a multi-objective optimization function with the lowest operating cost and lowest carbon emissions of the specified microgrid as optimization objectives, and inputting it into the equivalent aggregation model to obtain an optimized scheduling model; acquiring real-time operating data of each distributed resource and inputting the real-time operating data into the optimized scheduling model, solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy; and optimizing the energy scheduling of the specified microgrid according to the global optimized scheduling strategy.

[0006] Furthermore, the step of generating a corresponding schedulable potential calculation target model based on the historical operating data and resource type of each of the distributed resources includes: obtaining the corresponding inherent physical working principle according to the resource type of each distributed resource; establishing a corresponding schedulable potential calculation initial model according to each inherent physical working principle; calibrating the schedulable potential calculation initial model according to the historical operating data of each distributed resource to obtain the schedulable potential calculation target model.

[0007] Further, the step of acquiring real-time operating data of each of the distributed resources, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy includes: acquiring real-time operating data of each distributed resource; inputting the real-time operating data of each distributed resource and the corresponding initial iteration value into a preset neural network to generate an energy allocation scheme; wherein, the neural network is trained using different real-time operating data and corresponding iteration values ​​as input and the energy allocation scheme as output, and the initial iteration value is a preset value; acquiring the predicted state values ​​of each of the distributed resources in the energy allocation scheme and converting them into state values ​​to obtain a set of state values; calculating the variance of the set of state values ​​and determining whether the variance is greater than a preset variance value; if the variance is greater than the preset variance value, then calculating the iteration value of each distributed resource; wherein, the calculation formula for calculating the iteration value of each distributed resource is as follows: , This represents the value of the (i+1)th iteration. Let V be the variance of the set of state values ​​corresponding to the i-th energy allocation scheme. This represents the variance of the set of state values ​​corresponding to the (i-1)th energy allocation scheme. The preset value is represented; each predicted state value is combined with the corresponding iterative value and input into the preset neural network to regenerate the energy allocation scheme. The predicted state values ​​are repeatedly obtained and energy allocation is performed until the variance of the set of predicted state values ​​of each distributed resource is less than the preset variance value.

[0008] Further, the step of acquiring real-time operating data of each of the distributed resources, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy includes: acquiring real-time operating data of each of the distributed resources and inputting the real-time operating data into the optimized scheduling model to obtain a temporary scheduling model; transforming the multi-objective optimization function into a single-objective function using a linear weighting method; and solving the transformed single-objective function using a mixed-integer linear programming algorithm to generate a global optimized scheduling strategy.

[0009] Furthermore, after the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: sending a strategy confirmation instruction to each communicable distributed resource according to the global optimization scheduling strategy; receiving feedback information from each communicable distributed resource; determining whether there is any rejection feedback information in each feedback information; if there is rejection feedback information, obtaining the corresponding target distributed resource; deleting the target distributed resource from the optimization scheduling model, and regenerating the optimization scheduling strategy.

[0010] Furthermore, after the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: obtaining the actual operating parameters of each of the distributed resources; calculating the deviation between the actual operating parameters and the theoretical parameters in the global optimization scheduling strategy; determining whether the deviation is greater than a preset deviation; if it is greater than the preset deviation, then re-acquiring real-time operating data to regenerate the global optimization scheduling strategy.

[0011] Furthermore, before the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: establishing a carbon flow distribution model of the designated microgrid based on the topology diagram; obtaining multiple carbon flow distribution branches based on the carbon flow distribution model; calculating the predicted carbon emission intensity of each carbon flow distribution branch according to the global optimization scheduling strategy; determining whether the predicted carbon emission intensity is greater than a preset carbon emission intensity; if the predicted carbon emission intensity is greater than the preset carbon emission intensity, then regenerating the global optimization scheduling strategy until the predicted carbon emission intensity of each carbon flow distribution branch in the regenerated global optimization scheduling strategy is less than or equal to the preset carbon emission intensity.

[0012] Further, the step of obtaining the topology diagram of the specified microgrid and establishing an equivalent aggregation model of the specified microgrid based on the target model calculated from each of the schedulable potentials and the topology diagram includes: obtaining and parsing the electrical connection relationships in the topology diagram to determine the access nodes of each distributed resource; obtaining the distance between each distributed resource; forming a target distributed resource group from distributed resources whose access node distance is less than a preset distance; and linearly superimposing the schedulable potential models of multiple distributed resources accessed to the same access node and the distributed resources in the target distributed resource group to form an equivalent aggregation model of the access node.

[0013] Furthermore, the step of establishing a multi-objective optimization function with the minimum operating cost and minimum carbon emissions of the specified microgrid as optimization objectives, and inputting it into the equivalent aggregation model to obtain the optimized scheduling model, includes: constructing a total operating cost objective function that includes fuel cost, operation and maintenance cost, and interaction cost with the main grid, and constructing a total carbon emission objective function based on the actual output and carbon emission coefficient of each distributed energy source; integrating the total operating cost objective function and the total carbon emission objective function into a single multi-objective optimization function using a linear weighted sum method or an ε-constraint method; and inputting the multi-objective optimization function into the equivalent aggregation model to obtain the optimized scheduling model.

[0014] Furthermore, after the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: obtaining early warning information of the designated microgrid to identify impending extreme scenarios; invoking a reinforcement learning agent pre-matched to the extreme scenario; wherein the agent is obtained through offline training using historical operating data under extreme scenarios; and the reinforcement learning agent outputs a temporary charging and discharging strategy to cope with the extreme scenario, which overrides the execution of the global optimization scheduling strategy.

[0015] The beneficial effects of this invention are as follows: By establishing a schedulable potential model for various distributed resources and combining it with an equivalent aggregation model constructed from the microgrid topology, the problem of coordinating heterogeneous resources using traditional methods is effectively solved, fully exploring the overall regulation potential of the system. Then, by coordinating optimization with multiple objectives of minimizing operating costs and carbon emissions, the economic efficiency and environmental friendliness of the microgrid under the zero-carbon goal are ensured. By inputting real-time data and using optimization algorithms to solve the problem, a globally optimal scheduling strategy can be quickly generated, significantly improving the operating efficiency and collaborative control capability of complex microgrids, and providing reliable technical support for achieving safe, economical, and low-carbon energy supply. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 is a flowchart of a multimodal energy optimization scheduling method for microgrids under a zero-carbon objective in one embodiment. Detailed Implementation

[0018] 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.

[0019] As shown in Figure 1, in one embodiment, a multimodal energy optimization scheduling method for microgrids under a zero-carbon objective is provided. This method can be applied to both terminals and servers; this embodiment illustrates its application to terminals. The specific steps of this multimodal energy optimization scheduling method for microgrids under a zero-carbon objective include: S1: Obtaining historical operating data and resource types of multiple distributed resources within a specified microgrid; S2: Generating corresponding schedulable potential calculation target models based on the historical operating data and resource types of each distributed resource; S3: Obtaining the topology diagram of the specified microgrid, and establishing an equivalent aggregation model of the specified microgrid based on the schedulable potential calculation target models and the topology diagram; S4: Establishing a multi-objective optimization function with the lowest operating cost and lowest carbon emissions of the specified microgrid as optimization objectives, and inputting it into the equivalent aggregation model to obtain an optimized scheduling model; S5: Obtaining real-time operating data of each distributed resource, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy; S6: Optimizing the energy scheduling of the specified microgrid according to the global optimized scheduling strategy.

[0020] As described in step S1 above, historical operating data and resource types of multiple distributed resources within a specified microgrid are acquired. Distributed resources can be wind power, photovoltaic power generation, energy storage systems, and adjustable loads, etc. Historical operating data refers to the operating status of these resources over a past period, including power output, operating time, fault records, and environmental factors (such as temperature, humidity, and radiation). This data can be acquired in real-time through the microgrid's monitoring system, SCADA (Supervisory and Data Acquisition) system, or other distributed control systems, while ensuring a consistent data format for subsequent analysis. Resource types can be determined based on the attributes of the distributed resources, such as functionality (e.g., power generation, energy storage, load); technology type (e.g., renewable energy (wind power, solar power), traditional energy (coal, gas), energy storage (batteries, pumped storage, etc.); and dispatchability (e.g., adjustable and non-adjustable resources). Adjustable resources refer to resources whose output can be adjusted according to demand, such as controllable loads and energy storage devices; non-adjustable resources are generally fixed power generation, such as certain types of small-scale wind power. This application specifically classifies resources according to technology type.

[0021] As described in S2 above, a corresponding schedulable potential calculation target model is generated based on the historical operating data and resource type of each distributed resource. By analyzing the operating characteristics of each resource under different conditions, a mathematical model can be established to reflect the schedulable capability of each resource in a specific context. The design of the schedulable potential model usually needs to consider multiple factors, including historical output power, load curves, meteorological conditions, etc. The modeling method can be time series analysis, regression analysis, or machine learning algorithms, etc. The goal of the model is to accurately predict the output potential of each resource in a certain period of time in the future, thereby providing a basis for decision-making. In addition, the schedulable potential model can be corrected by combining weather forecast data to improve the prediction accuracy.

[0022] As described in step S3 above, the topology diagram of the specified microgrid is obtained, and an equivalent aggregation model of the specified microgrid is established based on the target models for calculating dispatchable potential and the topology diagram. The topology diagram describes the connection relationships between various resources, including the electrical connections between power generation units, energy storage devices, and loads. This diagram not only reflects the spatial distribution of resources but also reveals the paths of power flow and potential constraints. Then, based on this topology diagram and combined with the target models for calculating dispatchable potential generated in the previous steps, an equivalent aggregation model is established. The core of this model is to reasonably aggregate various distributed resources within the microgrid to form a more simplified, large-scale overall model. During the aggregation process, the parallel or series relationships between resources should be considered, and idealization should be performed through electrical characteristics such as equivalent impedance and equivalent voltage sources to improve computational efficiency. The equivalent aggregation model refers to a simplified model formed by linearly superimposing the dispatchable potential models of multiple distributed resources based on topology relationships.

[0023] As described in step S4 above, a multi-objective optimization function is established with the goal of minimizing the operating cost and carbon emissions of the specified microgrid. This function is then input into the equivalent aggregation model to obtain the optimized scheduling model. In other words, based on the previously established equivalent aggregation model, the multi-objective optimization function is constructed and solved. The optimization objectives mainly include reducing the operating cost and carbon emissions of the microgrid. First, it is necessary to clarify the calculation methods for various costs and carbon emissions. Typically, operating costs include equipment maintenance costs, fuel costs, and grid interaction costs. Carbon emissions are determined by considering emission coefficients related to the type of energy used. In the multi-objective optimization, the weights of costs and carbon emissions are dynamically adjusted based on real-time electricity prices or carbon policies. The constructed multi-objective optimization function is a mathematical formula, usually expressed as minimizing costs and carbon emissions. Using strategies such as the weighting method or the ε-constraint method, the two objectives are transformed into one optimization problem and incorporated into the equivalent aggregation model.

[0024] As described in step S5 above, real-time operating data of each of the distributed resources is acquired and input into the optimized scheduling model. A preset optimization algorithm is used to solve the optimized scheduling model to generate a global optimized scheduling strategy. Real-time operating data of each distributed resource within the microgrid is acquired through methods such as direct access from a monitoring system or sensor readings. This real-time data includes information such as the resource's immediate output, load demand, and energy storage status, effectively reflecting the actual operating conditions of the microgrid. This real-time data is then input into the previously established optimized scheduling model, and a preset optimization algorithm (such as genetic algorithm, particle swarm optimization, or mixed integer linear programming) is used to solve the model. The selection of this optimization algorithm should consider both computational complexity and solution efficiency to generate a scheduling strategy in a short time. Finally, a global optimized scheduling strategy is output, specifically specifying the power generation, energy storage, and load regulation plans for each distributed resource within a specified time period.

[0025] As described in step S6 above, the energy of the designated microgrid is optimized and scheduled according to the global optimization scheduling strategy. The previously obtained scheduling strategy is implemented in detail, adjusting the output, storage, and consumption of each resource according to plan. The microgrid control system issues instructions based on the global scheduling strategy to adjust the operating status of each distributed resource, including starting, shutting down, and adjusting power. During implementation, the control system needs to monitor the microgrid's operating status in real time to ensure that all indicators meet expectations. Simultaneously, the scheduling is continuously adjusted and optimized to cope with potential load changes, market price fluctuations, and equipment failures. Furthermore, this process must ensure compliance with environmental standards, especially striving to increase the proportion of renewable energy use while keeping carbon emissions under control. Through this step, the microgrid achieves an optimal balance between overall operating efficiency and carbon emissions, contributing to sustainable development while achieving the zero-carbon goal.

[0026] In one embodiment, step S2, which generates a corresponding schedulable potential calculation target model based on the historical operating data and resource type of each of the distributed resources, includes: S201: obtaining the corresponding inherent physical working principle according to the resource type of each distributed resource; S202: establishing a corresponding schedulable potential calculation initial model based on each inherent physical working principle; S203: calibrating the schedulable potential calculation initial model based on the historical operating data of each distributed resource to obtain the schedulable potential calculation target model.

[0027] As described in step S201 above, the inherent physical working principle is obtained according to the resource type of each distributed resource. An in-depth analysis is conducted on each type of distributed resource to obtain its inherent physical working principle. These resource types include photovoltaic power generation, wind power generation, gas turbines, energy storage systems (such as batteries and supercapacitors), and controllable loads. By consulting relevant literature, technical manuals, and standards, and combining practical engineering experience, the physical characteristics, output features, and scheduling response of each resource are described in detail. For example, the working principle of a photovoltaic power generation system is mainly based on the photoelectric effect, and its output power is closely related to environmental factors such as light intensity and temperature. Wind power generation, on the other hand, is affected by wind speed and turbine design, and has a rated output power range. In obtaining the inherent physical working principle, detailed mathematical models also need to be established to describe these principles, including nonlinear relationships and dynamic change characteristics.

[0028] As described in step S202 above, an initial model for calculating the dispatchable potential is established based on the inherent physical working principles of each resource. The purpose of this model is to quantify the dispatchable capacity of each resource, providing a basis for subsequent scheduling and optimization. The model should consider the dynamic characteristics, operational limitations, and adjustment capabilities of each resource, while simplifying it as much as possible to improve computational efficiency. Different modeling methods can be used for each type of resource. For photovoltaic and wind power resources, time-series models based on weather forecasts and historical data are typically used to capture the time-varying characteristics of their output power. For energy storage systems, a dynamic model can be established based on their charging and discharging patterns and state of charge to reflect changes in their energy storage capacity. Similarly, for controllable loads, a demand-response-based adjustment model can be established to respond quickly to changes in demand.

[0029] As described in step S203 above, the initial model for calculating schedulable potential is calibrated based on the historical operating data of each distributed resource to obtain the target model for calculating schedulable potential. The calibration process involves adjusting model parameters and verifying the input-output relationship. By comparing historical data with model output, statistical analysis methods (such as least squares, genetic algorithms, or machine learning) are used to optimize model parameters, minimizing the deviation between the optimized model output and the actual operating data. This can be quantified using error metrics (such as root mean square error, absolute error, etc.), thereby ensuring the accuracy and reliability of the model.

[0030] In one embodiment, step S5, which involves acquiring real-time operating data of each of the distributed resources, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy, includes: S501: acquiring real-time operating data of each distributed resource; S502: inputting the real-time operating data of each distributed resource and its corresponding initial iteration value into a preset neural network to generate an energy allocation scheme; wherein the neural network is trained using different real-time operating data and corresponding iteration values ​​as input and the energy allocation scheme as output, and the initial iteration value is a preset value; S503: acquiring the predicted state values ​​of each of the distributed resources in the energy allocation scheme and converting them into state values ​​to obtain a set of state values; S504: calculating the variance of the set of state values ​​and determining whether the variance is greater than a preset variance value; S505: if the variance is greater than the preset variance value, calculating the iteration value of each distributed resource; wherein the formula for calculating the iteration value of each distributed resource is... , This represents the value of the (i+1)th iteration. Let V be the variance of the set of state values ​​corresponding to the i-th energy allocation scheme. This represents the variance of the set of state values ​​corresponding to the (i-1)th energy allocation scheme. S506: Input each of the predicted state values ​​and the corresponding iterative values ​​into the preset neural network to regenerate the energy allocation scheme, repeatedly obtain the predicted state values ​​and perform energy allocation, until the variance of the set of predicted state values ​​of each distributed resource is less than the preset variance value.

[0031] As described in step S501 above, real-time operating data of each distributed resource is acquired. The acquisition of real-time operating data can be achieved through various monitoring devices and sensors, which can be integrated into the microgrid management system, such as SCADA (Supervisory and Data Acquisition System). Real-time data may include output power, operating status, environmental conditions (such as temperature and illuminance), and other relevant information.

[0032] As described in step S502 above, the real-time operating data and corresponding initial iteration values ​​of each distributed resource are input into a preset neural network to generate an energy allocation scheme. The neural network, through learning from historical data, has the ability to identify the complex relationship between input data and energy allocation. The input data typically includes real-time power output, load demand, and other relevant influencing factors. Meanwhile, the "initial iteration value" refers to a pre-set feedback value, usually used to help the model converge quickly during the iteration process. Based on the input real-time data and initial iteration value, the neural network calculates the optimal energy allocation strategy and provides it to the microgrid's dispatcher. The neural network is a multilayer perceptron, using historical operating data and iteration values ​​as input, and is trained using a backpropagation algorithm.

[0033] As described in step S503 above, the predicted state values ​​of each distributed resource in the energy allocation scheme are obtained and converted into state values ​​to obtain a set of state values. The predicted state values ​​are obtained from the generated energy allocation scheme and converted to form a set of state values. Predicted state values ​​refer to the energy output or performance indicators that each resource may achieve in a future time period under a specific energy allocation scheme. This process may involve modeling and predicting variables such as predicted load demand and future weather conditions, thereby achieving more accurate scheduling.

[0034] As described in step S504 above, the variance of the set of state values ​​is calculated, and it is determined whether the variance is greater than a preset variance value. The variance of the set of state values ​​obtained in step S503 is calculated and compared with a preset variance value. Variance is a statistic used to measure the dispersion of a set of data distributions and can reflect the volatility between predicted state values. By analyzing the variance of the state values, the stability and consistency of the current energy allocation scheme can be effectively evaluated. Specifically, if the variance is large, it indicates that the output states of each distributed resource differ significantly, which may lead to unstable system scheduling; conversely, a smaller variance indicates that the system states are more consistent, and the scheduling scheme is close to optimal. The process of comparing the variance with the preset value is crucial because it provides a basis for judgment for subsequent iterations and optimizations. If the variance is greater than the preset value, it indicates that the current energy allocation scheme needs further optimization; while if the variance is less than or equal to the preset value, the current scheme can be considered acceptable, and subsequent scheduling can be implemented based on this scheme. The preset variance value is a pre-set value, for example, set to 9.

[0035] As described in step S505 above, if the variance is greater than a preset variance value, then the iterative value for each distributed resource is calculated. The new iterative value calculated according to the formula provides a basis for subsequent adjustments. The calculation logic in the formula explains that if the variance of the current set of state values ​​is large, the new iterative value will be adjusted to a certain extent so that the variance can be significantly reduced in the next iteration, improving the stability of the scheme. In this way, the energy output between distributed resources can be more accurately coordinated, achieving better overall performance. Therefore, the importance of this process lies in optimizing resource scheduling through iterative calculation to ensure the operating efficiency and stability of the microgrid. The target variance value is set based on historical data or system stability requirements.

[0036] As described in step S506 above, each predicted state value, combined with its corresponding iterative value, is input into the preset neural network to regenerate the energy allocation scheme. This process of acquiring predicted state values ​​and allocating energy is repeated until the variance of the set of predicted state values ​​for each distributed resource is less than the preset variance value. The goal is to input the predicted state value of each distributed resource, combined with its corresponding iterative value, into the preset neural network to regenerate the energy allocation scheme. By integrating the iterative values ​​calculated in the previous step into the new input data, the neural network helps improve the accuracy and effectiveness of the scheduling scheme. By repeating this process, a dynamic energy allocation adjustment mechanism can be achieved. Each time an input-output cycle is completed, a new predicted state value is obtained, and the variance is calculated again. This process is repeated until the variance of the set of state values ​​is less than the set preset variance value. This iterative mechanism ensures that the scheduling strategy can fully adapt to the dynamically changing operating environment and uncertainties, contributing to the stable and economical operation of the entire microgrid. Finally, the microgrid will be scheduled according to the corrected energy allocation scheme to achieve the predetermined economic and environmental goals.

[0037] In another embodiment, step S5, which involves acquiring real-time operational data of each of the distributed resources, inputting the real-time operational data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy, includes: S511: acquiring real-time operational data of each of the distributed resources and inputting the real-time operational data into the optimized scheduling model to obtain a temporary scheduling model; S512: transforming the multi-objective optimization function into a single-objective function using a linear weighting method; S513: solving the transformed single-objective function using a mixed-integer linear programming algorithm to generate a global optimized scheduling strategy.

[0038] As described in step S511 above, real-time operational data of various distributed resources in the microgrid is acquired through a monitoring system. This data includes the instantaneous output of photovoltaic and wind power generation, the charging and discharging status of energy storage systems, and real-time changes in various load demands. Real-time data acquisition generally relies on smart meters, sensors, and data acquisition systems (such as SCADA), and the accuracy and timeliness of the data must be guaranteed. The key to this step is ensuring that the acquired data comprehensively reflects the current operating status of the system, providing a solid foundation for subsequent optimization model establishment. This data is then input into the optimization scheduling model to generate a temporary scheduling model. The temporary scheduling model is initially generated based on current real-time data and existing model parameters, with the aim of providing a rapid preliminary solution for the upcoming detailed optimization.

[0039] As described in step S512 above, the multi-objective optimization function is transformed into a single-objective function using a linear weighting method. This step is an important process in multi-objective optimization. Originally, the multi-objective optimization function might contain multiple different objectives, such as reducing operating costs, reducing carbon emissions, and improving system reliability. Since these objectives may conflict, direct optimization becomes relatively complex. Therefore, the linear weighting method integrates multiple objectives into a single objective function. By introducing weights, the optimization problem can be transformed into minimizing the absolute value of a weighted sum, such as: F = w1 * f1 + w2 * f2 + ... + wn * fn, where F is the objective function, fi are the different objectives, and wi are the corresponding weights. The weights are set according to priority or a specific strategy to ensure that the relative influence of each objective is appropriately reflected during the solution process. The advantage of the linear weighting method lies in its simplicity and effectiveness; it allows for flexible adjustment of the optimization direction without losing the characteristics of mutual interference between objectives.

[0040] As described in step S513 above, the transformed single-objective function is solved using a mixed-integer linear programming (MILP) algorithm to generate a global optimal scheduling strategy. MILP is a powerful tool in operations research, widely used for optimization problems involving combinations of decision variables and linear constraints. In microgrid scheduling, many decision variables (such as generation power, energy storage power, and load regulation) may be constrained, and some variables are integers (e.g., yes / no decisions), while others are continuous (e.g., power values). Therefore, the MILP method is well-suited for this scenario. During the solution process, the solver combines the input linear objective function with constraints and uses mathematical optimization techniques (such as the simplex method and branch-and-bound method) for efficient solution, finding the optimal combination of decision variables that minimizes the objective function. Through this solution process, a global optimal scheduling strategy can be generated, providing the optimal scheduling scheme for each distributed resource. Under this strategy, the microgrid can more effectively coordinate its resources, achieving the goals of cost reduction and carbon emission reduction.

[0041] Specifically, the optimization algorithm can be selected as an alternative to the neural network iterative method or the mixed integer linear programming method, depending on the system requirements.

[0042] In one embodiment, after step S6 of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: S701: sending a strategy confirmation instruction to each communicable distributed resource according to the global optimization scheduling strategy; S702: receiving feedback information from each communicable distributed resource; S703: determining whether there is any rejection feedback information among the feedback information; S704: if there is rejection feedback information, obtaining the corresponding target distributed resource; S705: deleting the target distributed resource from the optimization scheduling model and regenerating the optimization scheduling strategy.

[0043] As described in step S701 above, based on the previously generated global optimization scheduling strategy, the control system needs to send strategy confirmation commands to each distributed resource within the microgrid. To ensure that all resources understand and accept the scheduling strategy and perform actual energy scheduling according to the new strategy, strategy confirmation commands are typically sent in the form of digital signals or command text. The content includes key information such as the scheduling plan for each resource, the expected power output, charging or discharging status, and response time. To ensure the effectiveness and accuracy of command transmission, the sending mechanism may employ a secure communication protocol to ensure that information is not lost or tampered with. Furthermore, for non-communicable resources, predictive models are used instead of feedback.

[0044] As described in step S702 above, feedback information is received from each of the communicable distributed resources. This feedback information consists of status reports generated by the resources during real-time operation after the scheduling strategy is executed. These reports typically relate to the actual output and operational status of the resources, such as power output, operating status, fault alarms, and other key performance indicators. The feedback information can be received via smart meters, sensors, or a direct communication interface.

[0045] As described in step S703 above, it is determined whether there is any rejection feedback among the various feedback messages. The control system needs to judge the received feedback messages to detect whether there is any rejection feedback. Rejection feedback typically refers to a situation where, after executing a scheduling instruction, the actual operating state of a distributed resource deviates significantly from the expected target, or the resource refuses to execute the instruction due to a self-protection mechanism.

[0046] As described in step S704 above, if there is a rejection feedback, the corresponding target distributed resource is acquired. Acquiring the target distributed resource typically involves checking the identity of each running resource and its feedback content, which requires the use of an existing resource database or operational status monitoring system. At this time, the system records the specific content of the rejection feedback, including detailed information such as the reason, status, and timestamp. For rejected distributed resources, the control system may consider adopting temporary alternatives, adjusting scheduling targets, or troubleshooting. In this way, the microgrid can quickly respond to various operational anomalies, ensuring the overall stability and security of the system.

[0047] As described in step S705 above, the target distributed resource is deleted from the optimized scheduling model, and a new optimized scheduling strategy is generated. The purpose is to remove resources that are unavailable or cannot function properly according to the scheduling strategy, thus freeing up space and time for resources that are operating normally. The regenerated optimized scheduling strategy should consider new operating conditions and parameters, and be solved again using a suitable optimization algorithm to ensure optimal operating efficiency and minimum carbon emissions. This process forms a continuous adjustment mechanism, ensuring that the microgrid can still respond flexibly under adverse conditions and always maintain system stability and security.

[0048] In one embodiment, after step S6 of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: S711: obtaining the actual operating parameters of each of the distributed resources; S712: calculating the deviation between the actual operating parameters and the theoretical parameters in the global optimization scheduling strategy; S713: determining whether the deviation is greater than a preset deviation value; S714: if it is greater than the preset deviation value, then re-acquiring real-time operating data to regenerate the global optimization scheduling strategy.

[0049] As described in step S711 above, the actual operating parameters of each of the distributed resources are obtained. These parameters typically include the resource's output power, operating status, energy storage and release, environmental conditions (such as temperature, humidity, and light intensity), fault status, and other indicators affecting resource operation. Obtaining these parameters usually relies on sensors and monitoring devices embedded in each distributed resource, which periodically send data to the microgrid's control system.

[0050] As described in step S712 above, the deviation between the actual operating parameters and the theoretical parameters in the global optimization scheduling strategy is calculated. The calculation of the deviation typically involves simple mathematical operations, such as obtaining the deviation by taking the difference between the actual output and the theoretical output. This deviation calculation helps identify potential operational problems, such as decreased equipment efficiency, energy waste, or unmet demand. By quantitatively analyzing the deviation, the system can gain a clearer understanding of its operating status and generate actionable insights.

[0051] As described in step S713 above, it is determined whether the deviation value is greater than a preset deviation value. The key to this determination is to set an appropriate preset deviation value (usually based on experience and historical data), such as 20% based on electrical impedance experience, in order to effectively assess the consistency between actual operation and scheduling objectives. If the deviation in actual operation is too large, it may indicate potential problems in the system, such as equipment failure, improper operation, or reduced operating efficiency due to changes in the external environment.

[0052] As described in step S714 above, if the deviation exceeds a preset value, real-time operating data is reacquired to regenerate the global optimized scheduling strategy. This aims to further verify and update the current operating status to obtain more accurate and up-to-date data, thus providing a basis for regenerating the global optimized scheduling strategy. Reacquiring real-time data can be achieved by calling the aforementioned monitoring devices or sensors to ensure the extraction of the latest energy output and operating status information. Once the new data acquisition is complete, the control system analyzes the new data and combines it with previous feedback to adjust the current scheduling model. This involves updating the model's input parameters, optimizing weights, or incorporating new environmental factors into the scheduling algorithm.

[0053] In one embodiment, before step S7 of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: S601: establishing a carbon flow distribution model of the designated microgrid based on the topology diagram; S602: obtaining multiple carbon flow distribution branches based on the carbon flow distribution model; S603: calculating the predicted carbon emission intensity of each carbon flow distribution branch according to the global optimization scheduling strategy; S604: determining whether the predicted carbon emission intensity is greater than a preset carbon emission intensity; S605: if the predicted carbon emission intensity is greater than the preset carbon emission intensity, regenerating the global optimization scheduling strategy until the predicted carbon emission intensity of each carbon flow distribution branch in the regenerated global optimization scheduling strategy is less than or equal to the preset carbon emission intensity.

[0054] As described in step S601 above, a carbon flow distribution model for the specified microgrid is established based on the topology diagram. This carbon flow distribution model is a tool used to analyze and calculate the carbon emissions of various parts of the microgrid. The topology diagram details the connections between various distributed resources (such as photovoltaics, wind turbines, energy storage devices, and load devices) in the microgrid, as well as their locations and functions within the entire grid. The modeling process involves considering multiple factors, such as the carbon emission factors of different energy types and the actual operating status of the resources. Furthermore, it may involve comparing the carbon emissions of renewable energy sources with those of traditional energy sources (such as coal-fired power and natural gas power). Through this process, the carbon emissions of each distributed resource under specific operating conditions can be quantified, providing fundamental data for subsequent optimization.

[0055] As described in step S602 above, multiple carbon flow distribution branches are obtained based on the carbon flow distribution model. These branches represent the output routes and energy flow paths of various distributed resources in the microgrid. Specifically, they can be the power supply path from the power generation unit to the load center or the charging and discharging path of the energy storage device. The process of obtaining carbon flow distribution branches not only helps to understand the flow mode of carbon emissions, but also reveals the responsibilities of different nodes and branches in terms of carbon emissions.

[0056] As described in step S603 above, the predicted carbon emission intensity of each carbon flow distribution branch is calculated according to the global optimization scheduling strategy. The carbon emission intensity is usually the amount of carbon dioxide emitted per unit of energy (e.g., kilowatt-hour). This value can effectively reflect the environmental impact of each branch. The calculation method may include multiplying the energy output of each branch by its corresponding carbon emission factor to obtain the comprehensive predicted carbon emission intensity.

[0057] As described in step S604 above, it is determined whether the predicted carbon emission intensity is greater than the preset carbon emission intensity. The preset carbon emission intensity value is a value set based on policy objectives, environmental standards, or sustainable development requirements, for example, set to 1000 grams per kilowatt-hour. If the predicted carbon emission intensity of any branch is greater than the preset value, it indicates that the operation mode of that branch needs to be adjusted to reduce its carbon emissions.

[0058] As described in step S605 above, if the predicted carbon emission intensity is greater than the preset carbon emission intensity, a global optimal scheduling strategy is regenerated until the predicted carbon emission intensity of each carbon flow distribution branch in the regenerated global optimal scheduling strategy is less than or equal to the preset carbon emission intensity. Regenerating the global optimal scheduling strategy typically involves reassessing the operating parameters of each distributed resource, updating the optimization algorithm, and introducing new constraints. The system needs to iterate and solve the problem multiple times until the predicted carbon emission intensity of all carbon flow distribution branches is reduced to or maintained within a preset reasonable range. Through continuous optimization and regeneration of the scheduling strategy, the microgrid can achieve effective control over carbon emissions to meet the dual goals of economic and sustainable development.

[0059] In one embodiment, step S3, which involves obtaining the topology diagram of the specified microgrid and calculating the target model based on each schedulable potential and establishing the equivalent aggregation model of the specified microgrid based on the topology diagram, includes: S301: obtaining and parsing the electrical connection relationships in the topology diagram to determine the access nodes of each distributed resource; S302: obtaining the distances between each distributed resource; S303: forming a target distributed resource group by grouping distributed resources whose access node distances are less than a preset distance; S304: linearly superimposing the schedulable potential models of multiple distributed resources accessing the same access node and the distributed resources in the target distributed resource group to form the equivalent aggregation model of the access node.

[0060] As described in step S301 above, the electrical connection relationships in the topology diagram are obtained and parsed to determine the access nodes of each distributed resource. The topology diagram is an important descriptive tool for microgrids. Through this diagram, the electrical connection relationships between various distributed resources (such as photovoltaic, wind power, energy storage batteries and loads) can be intuitively understood, thereby identifying the location of access nodes. These nodes are the points in the power network that connect multiple distributed resources.

[0061] As described in step S302 above, the distance between each distributed resource is obtained. The definition of distance can be not only physical distance, such as the cable length or straight-line distance between distributed resources, but may also include electrical distance (such as impedance). The process of obtaining the distance between each resource can be carried out by various means, such as using GIS (Geographic Information System) data, electrical design drawings, or real-time measurement through smart sensors.

[0062] As described in step S303 above, distributed resources with access node distances less than a preset distance are grouped into target distributed resource groups. Based on a set preset distance threshold (which can be set to 5km based on empirical values ​​of electrical impedance), resources with access node distances less than this threshold are selected and aggregated into a group. This grouping strategy ensures that resources with close proximity can be effectively coordinated and operated through electrical connections, achieving optimized scheduling. The core logic of forming target distributed resource groups lies in the fact that resources with close proximity typically face similar operating conditions, and faster response times can improve the efficiency of the entire system during actual scheduling. For example, placing resources with close electrical connections in the same group can reduce electrical interference and harmony issues.

[0063] As described in step S304 above, multiple distributed resources connected to the same access node, and the schedulable potential models of the distributed resources in the target distributed resource group, are linearly superimposed to form an equivalent aggregation model of the access node. Linear superposition means combining the schedulable potential models of different resources in a certain way to form a unified model. Specifically, this involves weighted summation of the schedulable potential of each resource to reflect the common capabilities of all resources. This method can consider various factors, such as different resource types, output characteristics, and operation and maintenance costs. Through this process, the system can generate a simplified and representative equivalent aggregation model, enabling scheduling optimization to be performed from a global perspective. In one embodiment, step S4, which establishes a multi-objective optimization function with the minimum operating cost and minimum carbon emissions of the specified microgrid as optimization objectives and inputs it into the equivalent aggregation model to obtain the optimized scheduling model, includes: S401: Constructing a total operating cost objective function that includes fuel cost, operation and maintenance cost, and interaction cost with the main grid, and constructing a total carbon emission objective function based on the actual output and carbon emission coefficient of each distributed energy source; S402: Integrating the total operating cost objective function and the total carbon emission objective function into a single multi-objective optimization function using a linear weighted sum method or an ε-constraint method; S403: Inputting the multi-objective optimization function into the equivalent aggregation model to obtain the optimized scheduling model.

[0064] As described in step S401 above, a total operating cost objective function is constructed, which includes fuel costs, operation and maintenance costs, and interaction costs with the main grid. A total carbon emission objective function is also constructed based on the actual output and carbon emission coefficient of each distributed energy source. The total operating cost objective function typically includes multiple components, such as fuel costs, operation and maintenance costs (O&M), and interaction costs with the main grid. The total carbon emission objective function is constructed based on the actual output of each distributed energy source and its corresponding carbon emission coefficient. This function helps to identify the gap between the goal of achieving zero carbon emissions and the goal of zero carbon emissions by calculating the total carbon dioxide emissions generated by each energy source during operation.

[0065] As described in step S402 above, the total operating cost objective function and the total carbon emission objective function are integrated into a single multi-objective optimization function using either a linear weighted sum method or an ε-constraint method. The integration process typically employs two common methods: the linear weighted sum method and the ε-constraint method. Linear weighted sum method: This method assigns weights to each objective function, summing them into a single weighted objective function. The weights reflect the decision-maker's emphasis on different objectives (economic and environmental). For example, weights can be set as w1 and w2, then the combined objective function can be expressed as: F = w1 * (total operating cost) + w2 * (total carbon emissions). The weights can be flexibly adjusted as needed to optimize different objectives. ε-constraint method: In this method, one objective function (such as total operating cost) is transformed into an optimization objective, while another objective function (such as total carbon emissions) is used as a constraint. By setting a certain value as a constraint, it is ensured that the optimization process does not exceed a predetermined carbon emission limit.

[0066] As described in step S403 above, the multi-objective optimization function is input into the equivalent aggregation model to obtain the optimized scheduling model. The equivalent aggregation model carries the operating capacity of the entire microgrid and can reflect different resource characteristics, topological relationships, and electrical network limitations. After the multi-objective optimization function is input into the model, the control system can iteratively solve the current operating state and scheduling objectives. The solution process may involve optimization algorithms, such as linear programming, integer programming, or heuristic algorithms, and update resource scheduling and output strategies based on the multi-objective results.

[0067] In one embodiment, after step S6 of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: S721: obtaining early warning information of the designated microgrid to identify impending extreme scenarios; S722: invoking a reinforcement learning agent pre-matched to the extreme scenario; wherein the agent is obtained through offline training using historical operating data under extreme scenarios; S723: the reinforcement learning agent outputs a temporary charging and discharging strategy to cope with the extreme scenario, which overrides the execution of the global optimization scheduling strategy.

[0068] As described in step S721 above, the early warning information of the designated microgrid is obtained to identify the extreme scenarios that are about to occur. Extreme scenarios may include load surges caused by weather changes, equipment failures, natural disasters (such as floods or storms), sudden fluctuations in power demand, etc. These situations may have a significant impact on the stability and economy of the microgrid. The process of obtaining early warning information generally relies on multiple information sources, such as meteorological departments, network monitoring systems, equipment status monitoring, or historical operation data analysis. When the system detects abnormal changes in certain key indicators (such as temperature, wind speed, and power load) or reaches a preset threshold, it will trigger the corresponding early warning.

[0069] As described in step S722 above, a reinforcement learning agent pre-matched with the extreme scenario is invoked; wherein, the agent is obtained through offline training using historical operation data under extreme scenarios, and the reinforcement learning agent is obtained through offline training based on historical operation data under extreme scenarios. Using reinforcement learning technology, the agent can learn how to optimize resource scheduling and energy management under specific conditions.

[0070] The reinforcement learning agent described uses the Deep Q-Network (DQN) algorithm and is trained offline based on historical extreme scenario data to maximize the reward for microgrid stability. The training process of the reinforcement learning agent typically includes multiple stages. In the initial stage, an environmental model is built by collecting historical data, and coping strategies are learned by simulating extreme situations. The agent learns the best strategy through a reward mechanism. As training progresses, its ability to make efficient decisions in extreme scenarios continuously improves. This mechanism enables the agent to exhibit strong adaptability and flexibility in new and unknown extreme scenarios.

[0071] As described in step S723 above, the reinforcement learning agent outputs temporary charging and discharging strategies to cope with the extreme scenarios, and overrides the execution of the global optimization scheduling strategy. The algorithm calculates the optimal charging and discharging operations for each distributed resource under specific extreme situations to ensure the system maintains balance and safety during potential crises. The temporary charging and discharging strategies output by the reinforcement learning agent can include specific operations such as charging and discharging regulation through energy storage devices, complementary use between wind and photovoltaic power generation, and load shedding. These strategies are flexible and updated in real time, designed to quickly adapt to the environmental changes and constraints faced by the microgrid. The overriding mechanism is designed based on a trade-off between short-term risks and overall long-term optimization. Although the global optimization scheduling strategy provides long-term economic and environmental goals, in extreme situations, adopting flexible temporary strategies will better reduce risks and ensure system security.

[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0073] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0074] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A multi-modal energy optimization scheduling method for microgrids under a zero-carbon objective, characterized in that, The method includes: acquiring historical operating data and resource types of multiple distributed resources within a specified microgrid; generating corresponding schedulable potential calculation target models based on the historical operating data and resource types of each distributed resource; acquiring a topology diagram of the specified microgrid and establishing an equivalent aggregation model of the specified microgrid based on the schedulable potential calculation target models and the topology diagram; establishing a multi-objective optimization function with the lowest operating cost and lowest carbon emissions of the specified microgrid as optimization objectives, and inputting it into the equivalent aggregation model to obtain an optimized scheduling model; acquiring real-time operating data of each distributed resource and inputting the real-time operating data into the optimized scheduling model, solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy; and optimizing the energy scheduling of the specified microgrid according to the global optimized scheduling strategy.

2. The microgrid multimodal energy optimization scheduling method under the zero-carbon objective as described in claim 1, characterized in that, The step of generating a corresponding schedulable potential calculation target model based on the historical operating data and resource type of each distributed resource includes: obtaining the corresponding inherent physical working principle according to the resource type of each distributed resource; establishing a corresponding schedulable potential calculation initial model according to each inherent physical working principle; calibrating the schedulable potential calculation initial model according to the historical operating data of each distributed resource to obtain the schedulable potential calculation target model.

3. The microgrid multimodal energy optimization scheduling method under the zero-carbon objective as described in claim 1, characterized in that, The steps of acquiring real-time operating data of each of the distributed resources, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy include: acquiring real-time operating data of each distributed resource; inputting the real-time operating data of each distributed resource and the corresponding initial iteration value into a preset neural network to generate an energy allocation scheme; wherein the neural network is trained using different real-time operating data and corresponding iteration values ​​as input and the energy allocation scheme as output, and the initial iteration value is a preset value; acquiring the predicted state values ​​of each of the distributed resources in the energy allocation scheme and converting them into state values ​​to obtain a set of state values; calculating the variance of the set of state values ​​and determining whether the variance is greater than a preset variance value; if the variance is greater than the preset variance value, then calculating the iteration value of each distributed resource; wherein the formula for calculating the iteration value of each distributed resource is as follows: ， This represents the value of the (i+1)th iteration. Let V be the variance of the set of state values ​​corresponding to the i-th energy allocation scheme. This represents the variance of the set of state values ​​corresponding to the (i-1)th energy allocation scheme. The preset value is represented; each predicted state value is combined with the corresponding iterative value and input into the preset neural network to regenerate the energy allocation scheme. The predicted state values ​​are repeatedly obtained and energy allocation is performed until the variance of the set of predicted state values ​​of each distributed resource is less than the preset variance value.

4. The multi-modal energy optimization scheduling method for microgrids under the zero-carbon objective as described in claim 1, characterized in that, The steps of acquiring real-time operating data of each of the distributed resources, inputting the real-time operating data into the optimized scheduling model, and solving the optimized scheduling model using a preset optimization algorithm to generate a global optimized scheduling strategy include: acquiring real-time operating data of each of the distributed resources and inputting the real-time operating data into the optimized scheduling model to obtain a temporary scheduling model; transforming the multi-objective optimization function into a single-objective function using a linear weighting method; and solving the transformed single-objective function using a mixed-integer linear programming algorithm to generate a global optimized scheduling strategy.

5. The microgrid multimodal energy optimization scheduling method under the zero-carbon objective as described in claim 1, characterized in that, After the step of optimizing the energy scheduling of the specified microgrid according to the global optimization scheduling strategy, the method further includes: sending a strategy confirmation instruction to each communicable distributed resource according to the global optimization scheduling strategy; receiving feedback information from each communicable distributed resource; determining whether there is any rejection feedback information in each feedback information; if there is rejection feedback information, obtaining the corresponding target distributed resource; deleting the target distributed resource from the optimization scheduling model, and regenerating the optimization scheduling strategy.

6. The multi-modal energy optimization scheduling method for microgrids under the zero-carbon objective as described in claim 1, characterized in that, After the step of optimizing the energy scheduling of the specified microgrid according to the global optimization scheduling strategy, the method further includes: obtaining the actual operating parameters of each of the distributed resources; calculating the deviation between the actual operating parameters and the theoretical parameters in the global optimization scheduling strategy; determining whether the deviation is greater than a preset deviation; if it is greater than the preset deviation, then re-acquiring real-time operating data to regenerate the global optimization scheduling strategy.

7. The microgrid multimodal energy optimization scheduling method under the zero-carbon objective according to claim 1, characterized in that, Before the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: establishing a carbon flow distribution model of the designated microgrid based on the topology diagram; obtaining multiple carbon flow distribution branches based on the carbon flow distribution model; calculating the predicted carbon emission intensity of each carbon flow distribution branch according to the global optimization scheduling strategy; determining whether the predicted carbon emission intensity is greater than a preset carbon emission intensity; if the predicted carbon emission intensity is greater than the preset carbon emission intensity, regenerating the global optimization scheduling strategy until the predicted carbon emission intensity of each carbon flow distribution branch in the regenerated global optimization scheduling strategy is less than or equal to the preset carbon emission intensity.

8. The multi-modal energy optimization scheduling method for microgrids under the zero-carbon objective as described in claim 1, characterized in that, The steps of obtaining the topology diagram of the specified microgrid and establishing an equivalent aggregation model of the specified microgrid based on the target model calculated from each of the schedulable potentials and the topology diagram include: obtaining and parsing the electrical connection relationships in the topology diagram to determine the access nodes of each distributed resource; obtaining the distance between each distributed resource; forming a target distributed resource group from distributed resources whose access node distance is less than a preset distance; and linearly superimposing the schedulable potential models of multiple distributed resources accessed to the same access node and the distributed resources in the target distributed resource group to form an equivalent aggregation model of the access node.

9. The multi-modal energy optimization scheduling method for microgrids under the zero-carbon objective according to claim 1, characterized in that, The steps of establishing a multi-objective optimization function with the goal of minimizing the operating cost and carbon emissions of the specified microgrid, and inputting it into the equivalent aggregation model to obtain the optimized scheduling model, include: constructing a total operating cost objective function that includes fuel cost, operation and maintenance cost, and interaction cost with the main grid, and constructing a total carbon emission objective function based on the actual output and carbon emission coefficient of each distributed energy source; integrating the total operating cost objective function and the total carbon emission objective function into a single multi-objective optimization function using a linear weighted sum method or an ε-constraint method; and inputting the multi-objective optimization function into the equivalent aggregation model to obtain the optimized scheduling model.

10. The multi-modal energy optimization scheduling method for microgrids under the zero-carbon objective according to claim 1, characterized in that, After the step of optimizing the energy scheduling of the designated microgrid according to the global optimization scheduling strategy, the method further includes: obtaining early warning information of the designated microgrid to identify impending extreme scenarios; invoking a reinforcement learning agent pre-matched to the extreme scenario; wherein the agent is obtained through offline training using historical operating data under extreme scenarios; and the reinforcement learning agent outputs a temporary charging and discharging strategy to cope with the extreme scenario, which overrides the execution of the global optimization scheduling strategy.

Images