Power distribution network scheduling optimization method and device, computer equipment, computer readable storage medium and computer program product

Abstract

The invention relates to a power distribution network dispatching optimization method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring power supply output information and load demand information of a plurality of polymers in a power distribution network; according to the power supply output information and the load demand information, determining reference interaction power among the plurality of polymers; the reference interaction power is optimized according to the parameter information of the power supply devices in the polymers through the polymers until a preset convergence condition is met, and target interaction power among the multiple polymers is obtained; according to the target interaction power and the parameter information of the power supply equipment in each aggregate, target output information of the power supply equipment in each aggregate is determined; and performing energy scheduling on the power supply equipment in each polymer according to the target output information. By adopting the method, the dispatching optimization efficiency and effect of the power distribution network can be improved.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for optimizing distribution network dispatch. Background Technology

[0002] With the integration of a large number of diverse distributed devices, the dispatching of smart distribution networks faces significant challenges. These distributed devices are numerous, varied in type and characteristics, and multiple uncertainties exist on both the source and load sides. Traditional optimization algorithms are weak in handling uncertainties and have high computational dimensionality, while purely data-driven methods suffer from high training costs and significant risks in online decision-making.

[0003] Therefore, a distribution network dispatch optimization scheme with higher processing efficiency and better dispatch effect is needed. Summary of the Invention

[0004] Therefore, it is necessary to provide a distribution network dispatch optimization method, device, computer equipment, computer-readable storage medium, and computer program product that can improve processing efficiency and dispatching effect in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for optimizing distribution network dispatch, including:

[0006] Obtain power output information and load demand information for multiple aggregates within the distribution network;

[0007] Based on the power output information and load demand information, the reference interaction power among the multiple aggregates is determined;

[0008] The reference interactive power is optimized by each of the aggregates based on the parameter information of the power supply devices within the aggregates until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates.

[0009] Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, the target output information of the power supply devices in each of the aggregates is determined respectively.

[0010] Energy scheduling is performed on the power supply equipment within each of the aggregates based on the target output information.

[0011] In one embodiment, the method further includes:

[0012] Using the interaction power between the multiple aggregates as coupling variables, a global scheduling optimization model corresponding to the distribution network is constructed based on the power output information and load demand information.

[0013] The reference interaction power is obtained by solving the global scheduling optimization model.

[0014] Based on the reference interactive power and the parameter information of the power supply devices in each aggregate, a local scheduling optimization model corresponding to each aggregate is constructed.

[0015] The local scheduling optimization model corresponding to each of the aggregates is solved to obtain the local optimal interaction power of each aggregate.

[0016] The model parameters of the global scheduling optimization model are updated based on the local optimal interaction power. The process then returns to the step of solving the global scheduling optimization model until the convergence condition is met. Finally, the optimal solution of the global scheduling optimization model is output as the target interaction power.

[0017] In one embodiment, the power output information includes the real-time power output capacity of the power supply equipment, the charging and discharging status of the energy storage system, and adjustable power parameters; the load demand information includes rigid load power, flexible load adjustment capability, and load forecast data; the method further includes:

[0018] Based on the real-time output capacity and adjustable power parameters of the power supply equipment, determine the local feasible output range of each aggregate;

[0019] Based on the rigid load power and flexible load adjustment capability, determine the net load requirement of each polymer;

[0020] Using the interaction power between the multiple aggregates as coupling variables, and constructing model constraints based on the local feasible output range and net load requirements, the global scheduling optimization model is obtained.

[0021] The process of determining the local scheduling optimization model includes:

[0022] For each of the aggregates, a local scheduling optimization model is constructed based on the reference interactive power, the charging and discharging state of the energy storage system of the aggregate, and the power supply equipment parameters.

[0023] In one embodiment, the parameter information of the power supply device includes rated capacity, upper and lower output limits, ramp rate, and operating cost coefficient; the method further includes:

[0024] A local scheduling optimization model for the aggregate is constructed with the objective functions of minimizing operating costs and penalizing interaction power deviation.

[0025] For the local scheduling optimization model, the optimal power allocation scheme that satisfies the equipment physical constraints and power balance constraints is solved to obtain the target power information.

[0026] In one embodiment, the method further includes:

[0027] The Pareto solution set of the global scheduling optimization model is generated based on a multi-objective optimization algorithm;

[0028] The Pareto solution set is dynamically selected based on the real-time state of the aggregate to obtain the target Pareto solution of the global scheduling optimization model;

[0029] Based on the constraints corresponding to the target Pareto solution, the global scheduling optimization model is reconstructed to obtain the reconstructed model;

[0030] The reconstructed model is solved to obtain the target interaction power.

[0031] In one embodiment, the method further includes:

[0032] Define the multi-objective function corresponding to the global scheduling optimization model;

[0033] Simulate the hunting behavior of eagle flocks to generate a set of multi-objective weight combinations corresponding to the multi-objective function;

[0034] Construct the action space based on the multi-objective weight combination set;

[0035] The state space is constructed based on the real-time state and the convergence state of the local scheduling optimization model;

[0036] Actions are selected in the action space based on a pre-defined policy network, and the results of action selection are evaluated based on a pre-defined evaluation network.

[0037] The evaluation result is selected based on the action output by the evaluation network, and the target Pareto solution is obtained from the multi-objective weight combination.

[0038] Secondly, this application also provides a power distribution network dispatch optimization device, comprising:

[0039] The acquisition module is used to acquire power output information and load demand information of multiple aggregates within the distribution network.

[0040] The first determining module is used to determine the reference interaction power between the plurality of aggregates based on the power output information and load demand information;

[0041] An optimization module is used to optimize the reference interaction power based on the parameter information of the power supply device within each aggregate until a preset convergence condition is met, thereby obtaining the target interaction power between the aggregates.

[0042] The second determining module is used to determine the target output information of the power supply device in each of the aggregates based on the target interactive power and the parameter information of the power supply device in each aggregate.

[0043] The scheduling module is used to schedule the energy of the power supply equipment in each of the aggregates according to the target output information.

[0044] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the power distribution network dispatch optimization method described in the first aspect.

[0045] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power distribution network scheduling optimization method described in the first aspect above.

[0046] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the power distribution network scheduling optimization method described in the first aspect.

[0047] The aforementioned power distribution network dispatch optimization method, apparatus, computer equipment, computer-readable storage medium, and computer program product acquire power output information and load demand information of multiple aggregates within the power distribution network. By aggregating a large number of dispersed and heterogeneous demand-side resources into aggregates with better regulation characteristics, the computational load can be effectively reduced. Based on the power output information and load demand information, a reference interaction power is determined among the multiple aggregates. Each aggregate optimizes the reference interaction power according to the parameter information of the power supply equipment within it until a preset convergence condition is met, obtaining the target interaction power among the multiple aggregates. By optimizing the reference interaction power according to the parameter information of the power supply equipment within it, global coordination is achieved through information exchange between nodes, thereby reducing communication costs, computational burden, and protecting data privacy. Based on the target interaction power and the parameter information of the power supply equipment within each aggregate, the target output information of the power supply equipment within each aggregate is determined. Energy dispatch is performed on the power supply equipment within each aggregate based on the target output information, realizing regional autonomy within each aggregate and global coordination, thus achieving more efficient and effective power distribution network energy dispatch. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is an application environment diagram of the power distribution network scheduling optimization method in one embodiment;

[0050] Figure 2 This is a flowchart illustrating a distribution network scheduling optimization method in one embodiment;

[0051] Figure 3 This is a schematic diagram illustrating the state of charge transition of an electric vehicle in one embodiment;

[0052] Figure 4 This is a schematic diagram of the photovoltaic output power versus light intensity curve in one embodiment;

[0053] Figure 5 This is a flowchart illustrating the relationship between wind turbine output power and wind speed in one embodiment.

[0054] Figure 6 This is a schematic diagram of the multi-objective optimization process in a distribution network scheduling optimization method in one embodiment;

[0055] Figure 7 This is a flowchart illustrating the distribution network scheduling optimization method in another embodiment;

[0056] Figure 8 This is a schematic diagram of the two-level optimization iteration process in the distribution network scheduling optimization method in one embodiment;

[0057] Figure 9 This is a structural block diagram of a power distribution network dispatch optimization device in one embodiment;

[0058] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0060] Before describing the embodiments of the present invention, the related technologies and their problems will be explained:

[0061] Smart distribution network optimization and scheduling faces multiple challenges. The integration of numerous diverse distributed devices (such as electric vehicles and photovoltaic power generation) presents significant challenges. Their varied types, large numbers, and characteristics make direct regulation or market trading difficult, leading to complex communication networks, a surge in variables and constraints, and computational difficulties. In particular, resources characterized by "small capacity, large volume, and diverse demands" are prone to the curse of dimensionality. Therefore, aggregation techniques are needed to combine these dispersed resources into aggregates with better regulatory characteristics to maximize their value.

[0062] Secondly, there are multiple uncertainties on both the source and load sides. Source-side photovoltaic power generation is greatly affected by weather, resulting in uncertain output and significant prediction errors, especially at the day-ahead level. The load side is influenced by factors such as weather, behavioral habits, and electricity prices. Combined with behavioral uncertainties arising from demand-side resources, this makes the adjustability of demand-side resources difficult to predict.

[0063] Furthermore, coordinating different stakeholders (such as microgrids, aggregators, and energy networks) presents another challenge. Centralized optimization frameworks are difficult to implement due to issues such as information security, conflicting interests, data privacy, and communication burdens. Therefore, it is necessary to study distributed coordination mechanisms adapted to various stakeholders to achieve regional autonomy and overall coordination.

[0064] Optimized scheduling can be divided into day-ahead and intraday scheduling. Day-ahead scheduling is prone to deviation from the optimal state due to prediction errors, so intraday scheduling needs to dynamically adjust strategies based on real-time conditions to adapt to high randomness. This is crucial for improving photovoltaic power consumption, operational efficiency, and system safety.

[0065] Real-time energy storage scheduling involves sequential decision-making with time-by-time optimization, requiring consideration of cross-time period impacts. Centralized scheduling faces challenges such as high communication and computation costs and poor privacy and reliability, necessitating research into distributed decision-making based on information exchange between neighboring nodes. Modeling various resource types is complex, and traditional centralized optimization struggles to handle high-dimensional nonlinear constraints. Furthermore, the weights of economic, safety, and environmental objectives are difficult to dynamically adapt to complex operating conditions. Traditional optimization algorithms are weak in handling uncertainty and have high computational dimensionality, while purely data-driven methods suffer from high training costs and significant online decision-making risks.

[0066] Related technologies struggle to effectively handle multiple uncertainties on both the source and load sides, resulting in large day-ahead scheduling deviations and insufficient intraday dynamic adjustments. There is a lack of mature and efficient distributed coordination mechanisms to balance the regional autonomy and global coordination needs of multiple stakeholders. In terms of optimization algorithms, traditional methods are unable to handle high-dimensional nonlinear constraints and uncertainties, while pure data-driven methods face training and online decision-making risks. Overall, they cannot effectively address the complex challenges of multi-type resource collaborative control, sequential energy storage decision-making, and dynamic optimization of multiple objectives such as safety, economy, and environmental protection.

[0067] Therefore, the power distribution network dispatch optimization schemes in related technologies have at least the following problems:

[0068] 1. Difficulty in handling massive heterogeneous resources: A large number of diverse devices (such as electric vehicles and photovoltaic power generation) are connected, with a wide variety of types, huge numbers, and different characteristics. Direct control leads to complex communication, a surge in variables and constraints, and is prone to the "curse of dimensionality", making it difficult to achieve refined management.

[0069] 2. Insufficient ability to cope with multiple uncertainties: The output of photovoltaic power generation is greatly affected by weather, resulting in significant forecasting errors; the load side is affected by a variety of factors, coupled with the uncertainty of resource behavior on the demand side, making forecasting and scheduling difficult.

[0070] 3. Centralized optimization faces bottlenecks: Centralized methods are inefficient when calculating high-dimensional nonlinear constraints, and suffer from problems such as high communication and computation costs, poor system privacy, and insufficient scalability, making them difficult to adapt to large-scale complex systems.

[0071] 4. Difficulty in coordinating multiple objectives: There are conflicts among economic, safety, and environmental objectives, and the weights are difficult to dynamically adjust according to complex working conditions, making it difficult to find the global optimal solution.

[0072] 5. Limitations of traditional optimization algorithms: They are weak in handling uncertainty, have high computational dimensionality, and are difficult to cope with real-time fluctuations.

[0073] 6. Data-driven methods are high-risk: high training costs, risks in online decision-making, and insufficient real-time performance.

[0074] 7. Lack of an effective distributed coordination mechanism: Existing technologies are unable to achieve distributed real-time scheduling decisions based on information exchange between adjacent nodes and adapt to various stakeholders, so as to solve the problems caused by centralized systems.

[0075] The power distribution network scheduling optimization method provided in this application can be applied to, for example, Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 can obtain power output information and load demand information for multiple aggregates within the power distribution network through terminal 102. Each aggregate includes at least one power supply device, corresponding to at least one terminal 102. Based on the power output information and load demand information, a reference interaction power is determined among the multiple aggregates. Each aggregate optimizes the reference interaction power based on the parameter information of the power supply devices within it until a preset convergence condition is met, obtaining the target interaction power among the multiple aggregates. Based on the target interaction power and the parameter information of the power supply devices within each aggregate, the target output information of the power supply devices within each aggregate is determined. Energy scheduling is performed on the power supply devices within each aggregate based on the target output information. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, and projection devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0076] In one exemplary embodiment, such as Figure 2 As shown, a distribution network scheduling optimization method is provided, which is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 202 to 206. Wherein:

[0077] Step 202: Obtain the power output information and load demand information of multiple aggregates in the distribution network.

[0078] In this system, the distribution network can be a smart distribution network, and the aggregate is a dispatchable unit that combines dispersed distributed resources into a single entity using virtual power plant technology. An aggregate can include at least one distributed resource. Aggregators can be categorized as follows: power-side aggregates such as photovoltaic power plants and energy storage systems; load-side aggregates such as interruptible loads and electric vehicle charging stations; and hybrid aggregates such as microgrids and integrated energy systems. Power output information characterizes the power output capacity of the aggregate, specifically including the real-time operating status and regulation capabilities of distributed power sources (such as photovoltaic, wind power, and energy storage) within the aggregate. Load demand information characterizes the power demand on the consumer side of the aggregate and its adjustability potential. This load demand information can include the adjustment capabilities of uninterruptible or non-adjustable rigid loads and flexible loads, as well as load forecast data.

[0079] Step 204: Determine the reference interaction power between the multiple aggregates based on the power output information and load demand information.

[0080] The reference interaction power refers to the planned power exchange value between multiple aggregators. Power exchange between multiple aggregators within the distribution network can optimize the overall operating efficiency and cost of the distribution network. To make the power interaction between aggregators more rational and efficient, the control center corresponding to the distribution network can be used as the coordinating entity between multiple aggregators. This control center is used to coordinate energy interaction between aggregators and electricity trading with the distribution network, achieving power regulation, and simultaneously transmitting the total regulated power of each aggregator to the aggregator operator. Through interaction between the control center and each aggregator, and by comprehensively analyzing the power output information and load demand information of each aggregator, the reference interaction power between multiple aggregators is obtained. The comprehensive analysis process can involve establishing a global dispatch optimization model corresponding to the distribution network based on the power output information and load demand information, and solving this global dispatch optimization model to obtain the reference interaction power. The objective function of the global dispatch optimization model can be to minimize the total operating cost of the distribution network (which may include generation costs and interaction losses between aggregators). The coupling variables of the global dispatch optimization model include the interaction power between aggregators, and the constraints of the global dispatch optimization model may include power balance, etc.

[0081] Step 206: Each of the aggregates optimizes the reference interactive power based on the parameter information of the power supply devices within the aggregate until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates.

[0082] In this embodiment of the invention, the initial reference interaction probability can be iteratively optimized based on an improved distributed iterative algorithm. Through the solution process of the aforementioned global scheduling optimization model of the distribution network, a reasonable reference interaction power and dual variable are obtained, entering the local optimization stage. Each aggregate solves its local subproblem in parallel and uploads the optimal interaction power obtained from the local subproblem to the distribution network. The distribution network then performs global coordination based on the solutions to the subproblems uploaded by all aggregates and updates the dual variable based on the coordination result. The distributed iterative algorithm is used to decompose the original problem into multiple easier-to-solve subproblems, approximating the optimal solution by alternately updating the original and dual variables. Examples include the ADMM (Alternating Direction Method of Multipliers) algorithm.

[0083] The preset convergence conditions may include at least one of the following: a preset number of iterations, a change in the interaction power between adjacent iteration steps being less than a preset value, or a change in the dual variables of the boundary coupling constraints being less than a preset value. Specifically, the rate of change of the interaction power between adjacent iteration steps characterizes the convergence degree of the primal problem in the distributed iterative algorithm, and the change in the dual variables of the boundary coupling constraints, i.e., the dual residual, characterizes the convergence degree of the dual problem.

[0084] Step 208: Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, determine the target output information of the power supply devices in each of the aggregates.

[0085] The parameter information of the power equipment within the aggregate can include basic operating parameters, economic parameters, energy storage-related parameters, and regulation characteristic parameters. Basic operating parameters can include rated capacity, real-time output upper and lower limits, ramp rate, start-stop time constant (applicable to controllable units), and minimum continuous operating time (applicable to thermal power units). Economic parameters can include operating cost coefficients, fuel cost curves, and maintenance cost coefficients (yuan / kWh). Penalty cost coefficients can include the unit price for wind / solar curtailment penalties and the unit price for load reduction compensation. Energy storage-specific parameters can include current state of charge, charge / discharge efficiency, maximum charge / discharge power, and capacity decay coefficient. Regulation characteristic parameters can include response delay time, adjustable bandwidth (for equipment involved in frequency regulation), and regulation accuracy error range. Based on the parameter information of the power devices within the polymer, the output constraints of each power device within the polymer are determined, such as upper and lower limits of power output, ramp rate constraints, and energy storage SOC constraints. Based on the output constraints of each power device within the polymer and the target interactive power, the output power is allocated among these power devices, thus obtaining the optimal output scheme for each power device. The target output information includes the planned output value and output cost of each power device.

[0086] Step 210: Perform energy scheduling on the power supply equipment in each of the aggregates according to the target output information.

[0087] Specifically, the power output of each power supply device within each aggregate is controlled based on the target output information in order to achieve effective scheduling of various resources within the power distribution network.

[0088] In one embodiment, the process of determining the target output information includes:

[0089] Using the interaction power between the multiple aggregates as coupling variables, a global scheduling optimization model corresponding to the distribution network is constructed based on the power output information and load demand information.

[0090] The reference interaction power is obtained by solving the global scheduling optimization model.

[0091] Based on the reference interactive power and the parameter information of the power supply devices in each aggregate, a local scheduling optimization model corresponding to each aggregate is constructed.

[0092] The local scheduling optimization model corresponding to each of the aggregates is solved to obtain the local optimal interaction power of each aggregate.

[0093] The model parameters of the global scheduling optimization model are updated based on the local optimal interaction power. The process then returns to the step of solving the global scheduling optimization model until the convergence condition is met. Finally, the optimal solution of the global scheduling optimization model is output as the target interaction power.

[0094] Among them, the reference interaction power is a coupling variable, representing the planned power exchange between aggregates, which directly affects the supply and demand balance of each aggregate and the overall operating cost.

[0095] The objective function of a global dispatch optimization model can be aimed at minimizing the overall operating cost or maximizing the efficiency of the distribution network, and may include the following components: Generation cost: fuel costs of controllable power sources (such as gas turbines and diesel generators) in each aggregate. Interaction cost: transmission losses or transaction costs of power exchange between aggregates. External grid transaction cost: cost / revenue of purchasing or selling electricity from the main grid. Renewable energy penalty: wind and solar curtailment penalties, and demand response cost: compensation costs for load adjustment (if there is adjustable load).

[0096] The decision variables in the global scheduling optimization model can include power generation variables and inter-aggregate interaction power variables. Power generation variables represent the actual output of controllable power sources (such as gas turbines and diesel generators) for each aggregator during a specific time period. Supply and demand gaps within aggregators can be balanced by adjusting the output of controllable power sources. Inter-aggregate interaction power variables represent the planned power exchange between each aggregator and other aggregators during a specific time period.

[0097] Correspondingly, the constraints of the global scheduling optimization model may include power balance constraints, which are used to ensure that the power output, interactive power, and load of each aggregate meet the real-time balance; interactive power coupling constraints, which are used to ensure that the interactive power between all aggregates meets the net balance (total interactive power is zero); power generation capacity constraints, which are used to constrain the upper and lower limits of the output of controllable power sources; interactive power limits, which are used to constrain the capacity limits of transmission lines between aggregates; main grid interaction constraints, which are used to constrain the power exchange limits between the distribution network and the main grid; and renewable energy consumption constraints, which are used to introduce wind and solar curtailment penalties or mandatory consumption.

[0098] For each aggregate, the construction process of the corresponding local scheduling optimization model may include: analyzing the power devices within the aggregate to obtain detailed parameters of the distributed power sources (output limits, ramp-up capabilities, etc.) and the real-time status of the energy storage system (SOC (State of Charge), charge / discharge efficiency); based on the analysis results and the reference interactive power requirements, using the output plan of each distributed power source as the decision variable; and considering the physical limits of the equipment and operational safety constraints as constraints, with the optimization objectives being tracking the reference interactive power and minimizing the operating cost of the aggregate. The solution results of the local scheduling optimization model corresponding to the aggregate may include the output plans of each power device, the local optimal interactive power value, and actual operating cost data.

[0099] The model parameters of the global scheduling optimization model are updated based on the local optimal interactive power values ​​solved by all aggregates. The process then returns to the step of solving the global scheduling optimization model. This process involves iterative local solving by aggregates, feeding the solution results back to the distribution network, global coordination and updating of the reference interactive power by the distribution network, and redistributing the updated reference interactive power to the aggregates so that they can solve locally again until convergence. Finally, the target interactive power between multiple aggregates is obtained.

[0100] In one embodiment, the power output information includes the real-time output capacity of the power supply equipment, the charging and discharging status of the energy storage system, and adjustable power parameters; the load demand information includes rigid load power, flexible load adjustment capability, and load forecast data. The real-time output capacity characterizes the actual power generation of the power supply at the current moment (MW / MVar), which may be affected by factors such as weather and equipment status. The charging and discharging status of the energy storage system may include SOC, charging and discharging power limits (such as maximum charging / discharging rate), and remaining available energy (MWh). Adjustable power parameters may include adjustable parameters of the power supply, such as ramp rate (power change capability per unit time, such as MW / min), minimum / maximum output limits, and reactive power regulation capability (such as the power factor control range of the photovoltaic inverter). The determination process of the global scheduling optimization model includes:

[0101] Based on the real-time output capacity and adjustable power parameters of the power supply equipment, determine the local feasible output range of each aggregate;

[0102] Based on the rigid load power and flexible load adjustment capability, determine the net load requirement of each polymer;

[0103] Using the interaction power between the multiple aggregates as coupling variables, and constructing model constraints based on the local feasible output range and net load requirements, the global scheduling optimization model is obtained.

[0104] The process of determining the local scheduling optimization model includes:

[0105] For each of the aggregates, a local scheduling optimization model is constructed based on the reference interactive power, the charging and discharging state of the energy storage system of the aggregate, and the power supply equipment parameters.

[0106] The real-time output capacity of the power equipment includes technical parameters such as the rated maximum output and minimum stable output of controllable power sources (e.g., gas turbines, diesel generators). Adjustable power parameters include dynamic adjustment characteristics such as the unit's ramp rate (the output value that can be increased or decreased per minute) and start-up / shutdown time. Based on the real-time output capacity and adjustable power parameters of the power equipment, the allowable output range of the controllable power sources for each aggregate during the scheduling period is calculated. For example, the real-time output of a gas turbine must be between its minimum and maximum technical output, and the output variation between adjacent time periods must not exceed the ramp rate limit. This ensures that the scheduling instructions for the controllable power sources conform to the actual operating capacity of the equipment, avoiding equipment damage or efficiency reduction due to exceeding technical limits.

[0107] Rigid load power characterizes fixed load demands that need to be met in real time (such as basic residential electricity consumption). Flexible load adjustment capability characterizes loads that can be reduced, delayed, or transferred (such as interruptible industrial loads) and their adjustment range. In this embodiment of the invention, rigid loads are directly included in total demand, and net load demand is dynamically adjusted according to the adjustment capability of flexible loads (such as maximum reduction ratio and delay duration), thereby obtaining the net load demand of each aggregate. For example, an aggregate can reduce its net demand by reducing flexible loads by 20% during peak hours. By participating in optimization through flexible loads, the flexibility of supply and demand balance is enhanced, and dependence on high-priced external electricity purchases is reduced.

[0108] The process of constructing constraints for the global scheduling optimization model can include: establishing the following constraints based on interactive power: Power balance constraint: The power output, interactive power, and renewable energy of each aggregate must meet its net load demand. Interactive power balance constraint: The sum of the interactive power of all aggregates is zero, ensuring energy conservation. Technical constraints: These include capacity limitations of interactive power transmission lines and power output range. Through interactive power, energy mutual assistance between aggregates is achieved, optimizing overall operating costs while meeting local supply and demand.

[0109] The process of solving the global scheduling optimization model and generating reference interactive power can include: the control center, with the objective of minimizing the total operating cost of the distribution network, solves the global scheduling optimization model to obtain the reference interactive power between each aggregate. Costs include the generation cost of controllable power sources, the transmission loss cost of interactive power, and the electricity trading cost with the main grid. Thus, the reference interactive power provides planned values ​​for energy exchange between aggregates, avoiding efficiency losses caused by blind scheduling.

[0110] Correspondingly, the charge / discharge status of an energy storage system includes the current state of charge (SOC), charge / discharge efficiency, and power limits. Power equipment parameters include, for example, the predicted output of renewable energy and the real-time regulation capability of controllable units.

[0111] The construction process of the local scheduling optimization model includes: each aggregate, with the goal of tracking the reference interaction power, constructs a local scheduling optimization model by combining local energy storage and power parameters, and optimizes the output of controllable power sources and the charging and discharging plans of energy storage. For example, if the reference interaction power requires the delivery of electricity during a certain period, then surplus renewable energy or energy storage discharge is prioritized to meet this demand, thereby achieving fine-grained scheduling of local resources within the global framework and ensuring the feasibility of the reference interaction power.

[0112] In one embodiment, the process of determining the target output information includes:

[0113] A local scheduling optimization model for the aggregate is constructed with the objective functions of minimizing operating costs and penalizing interaction power deviation.

[0114] For the local scheduling optimization model, the optimal power allocation scheme that satisfies the equipment physical constraints and power balance constraints is solved to obtain the target power information.

[0115] The construction process of the local scheduling optimization model includes:

[0116] The core objectives of the local scheduling optimization model can include minimizing operating costs and penalizing interaction power deviations. The interaction power deviation penalty applies a penalty term to deviations in actual interaction power if the aggregate needs to track the reference interaction power issued by the global scheduler, ensuring consistency with the global plan. Constraints of the local scheduling optimization model can include: equipment physical constraints and power balance constraints.

[0117] The model solution method for the local scheduling optimization model can employ algorithms such as linear programming or quadratic programming. Under the constraints mentioned above, the output values ​​of each power supply device that minimize the objective function are found. In this embodiment of the invention, coordination between the global distribution network and the local aggregation is achieved, and an interactive power deviation penalty term is set to ensure the consistency between local scheduling and the global plan, avoiding overall efficiency loss caused by local optimization. Furthermore, physical feasibility guarantees are introduced through power supply device parameters, and strict device constraints prevent over-limit scheduling, thereby improving system operational reliability.

[0118] In one embodiment, the solution process of the global scheduling optimization model includes:

[0119] The Pareto solution set of the global scheduling optimization model is generated based on a multi-objective optimization algorithm;

[0120] The Pareto solution set is dynamically selected based on the real-time state of the aggregate to obtain the target Pareto solution of the global scheduling optimization model;

[0121] Based on the constraints corresponding to the target Pareto solution, the global scheduling optimization model is reconstructed to obtain the reconstructed model;

[0122] The reconstructed model is solved to obtain the target interaction power.

[0123] Among these considerations, it is possible that there are multiple optimization objectives in distribution network dispatch optimization, such as minimizing operating costs.

[0124] Maximizing renewable energy utilization, minimizing network losses, and minimizing voltage deviation are multiple optimization objectives that may conflict. Therefore, a multi-objective optimization algorithm can be used to simultaneously optimize multiple conflicting objectives, generating a set of Pareto optimal solutions. Each solution represents a trade-off; for example, solution A: lowest cost but more renewable energy is discarded; solution B: full utilization of renewable energy but higher cost. This provides diverse scheduling strategies, laying the foundation for dynamically selecting solutions adapted to different scenarios.

[0125] Based on a preset deep learning algorithm, the Pareto solution set can be dynamically selected according to the real-time state of the aggregate to obtain the target Pareto solution of the global scheduling optimization model. For example, a preset decision-evaluation network can be used to perform selection actions on the Pareto solution set based on its state, and then the selection results can be evaluated. The action selection strategy can then be updated based on the evaluation results.

[0126] The reconstruction of the global scheduling optimization model can be achieved by adding or adjusting model constraints based on the constraints corresponding to the objective Pareto solution (such as selecting a solution that requires carbon emissions to not exceed a certain threshold). In this embodiment of the invention, the optimal solution is switched according to the real-time state to improve the robustness of the scheduling scheme. Furthermore, the solution space is first reduced through a multi-objective optimization algorithm, and then the reconstructed model is used to achieve rapid convergence, thus balancing computational efficiency and accuracy.

[0127] In one embodiment, the process of determining the target Pareto solution includes:

[0128] Define the multi-objective function corresponding to the global scheduling optimization model;

[0129] Simulate the hunting behavior of eagle flocks to generate a set of multi-objective weight combinations corresponding to the multi-objective function;

[0130] Construct the action space based on the multi-objective weight combination set;

[0131] The state space is constructed based on the real-time state and the convergence state of the local scheduling optimization model;

[0132] Actions are selected in the action space based on a pre-defined policy network, and the results of action selection are evaluated based on a pre-defined evaluation network.

[0133] The evaluation result is selected based on the action output by the evaluation network, and the target Pareto solution is obtained from the multi-objective weight combination.

[0134] First, a set of multi-objective functions corresponding to the global scheduling optimization model is defined, which may include: minimizing operating costs, maximizing renewable energy consumption, minimizing network losses, and minimizing voltage deviation. Each objective function corresponds to an optimization dimension, and there may be conflicting relationships between the dimensions. By establishing a complete optimization objective system, a foundation is provided for subsequent multi-objective optimization.

[0135] The process of generating a multi-objective weight combination set may include: initializing individual eagles, each representing a weight combination; simulating the three stages of hunting: search, encirclement, and attack; search stage: random exploration of weight combinations; encirclement stage: moving towards the current optimal solution; and attack stage: local fine-grained search, thereby obtaining a uniformly distributed weight combination through a biomimetic algorithm.

[0136] The process of constructing the action space includes: mapping the generated set of weight combinations to a discrete action space, assigning each action to a specific set of weight combinations, and ensuring that the dimension of the action space is the same as the number of weight combinations, thereby providing an operable decision space for reinforcement learning.

[0137] Real-time status includes: renewable energy output deviation, load fluctuation, network topology changes, etc. Local scheduling convergence status can include optimization residuals, number of iterations, degree of constraint violation, etc.

[0138] The state space construction process may include: normalizing the above real-time states and then constructing a state vector.

[0139] In the policy network and evaluation network, the policy network has the following inputs: state vector and output: action selection probability distribution; the evaluation network has the following inputs: state-action pairs and output: value assessment. The policy network and evaluation network can be trained using the Actor-Critic framework.

[0140] The process of selecting the target Pareto solution includes: 1. Inputting the current state into the policy network; 2. Obtaining the selection probability of each action; 3. Selecting the optimal action according to the probability distribution; 4. Obtaining the corresponding weight combination, which is the target Pareto solution.

[0141] In this embodiment of the invention, diverse weight combinations are obtained through biomimetic algorithms, and intelligent decision-making is achieved using deep reinforcement learning, which dynamically adapts to changes in system state, thereby improving the robustness of the scheduling scheme.

[0142] In one embodiment, taking ADMM-reinforcement learning-based collaborative optimization as an example, the scheme for distribution network scheduling optimization is illustrated:

[0143] First, a unified mathematical model is established for the various types of resources in the distribution network:

[0144] The process of constructing the electric vehicle model: For electric vehicles, it is assumed that they adopt square wave charging. The initial charging time of the electric vehicle is assumed to be... When connecting to a charging station, the user's initial battery state of charge is The user's expected disconnection time is Furthermore, it requires that the battery state of charge must meet certain conditions at the time of off-grid connection. So, what is the energy demand of electric vehicles? for:

[0145]

[0146] In the formula, Rated capacity of electric vehicles; The charging efficiency of electric vehicles. Assuming the electric vehicle charges within a certain time frame... With average power To charge,

[0147]

[0148] EVs can deviate from this expected charge trajectory to provide flexibility, such as Figure 3 As shown, two parallel boundary curves are set above and below the average charging curve.

[0149]

[0150]

[0151] In the formula: This represents the distance from the upper boundary curve to the average charging curve. This represents the distance from the lower boundary curve to the average charging curve.

[0152] Electric vehicles from time At rated charging power Charging begins. Because the slope of the rated charging curve is greater than that of the average charging curve, the electric vehicle stops charging when it intersects the upper boundary curve, and restarts charging when it intersects the lower boundary curve. The formula is:

[0153]

[0154]

[0155] In the formula, This represents the charging status of the electric vehicle; 1 indicates charging, and 0 indicates not charging. When the rated charging curve of the electric vehicle intersects with the upper boundary... When the rated charging curve changes from 1 to 0 and intersects the lower boundary, The change from 0 to 1. To describe the charging process, the controllable energy state margin of an electric vehicle is defined. .

[0156]

[0157] In the formula, The charge capacity for charging electric vehicles at average power; The charge capacity for electric vehicles charged using square wave charging; This is the size of the power dead zone, used to limit the error between the power output when disconnected from the grid and the user's target power. From the above formula, it can be seen that... The value ranges from -1 to 1. When it approaches 0, it means that the adjustable margin is larger, the electric vehicle deviates less from the charge trajectory, and the user satisfaction is higher. When it approaches -1 or 1, it means that the electric vehicle's charge capacity dead zone is at its upper and lower limits.

[0158] Therefore, the charging model for electric vehicles can be expressed as:

[0159]

[0160] It can be simplified to:

[0161]

[0162] In the formula: , , All are model coefficients.

[0163] The process of constructing an energy storage device model: The energy storage device represents its electrical state using its state of charge. The state model for the charging and discharging process is as follows:

[0164]

[0165]

[0166]

[0167] In the formula: and They represent Time and State of charge (SOC) that stores energy at all times; The efficiency coefficient of energy storage; and These are the charging and discharging efficiencies, respectively. For energy storage devices Operating power at any given time; The rated battery capacity of the energy storage device; This represents the operational status of distributed energy storage. A value of 1 indicates that the energy storage is in a charging state. A value of -1 indicates that the energy storage is in a discharge state. When the value is 0, it is in an idle state; This refers to the rated power of the energy storage device.

[0168] The DoCE (charge-discharge efficiency) of distributed energy storage is defined as:

[0169]

[0170] Therefore, the dynamic model of distributed energy storage can be represented as:

[0171]

[0172] In power systems, the uncertainty of renewable energy output is the main source of uncertainty, with wind power and solar power being the two most stochastic renewable energy sources. Solar power output is primarily determined by solar irradiance, and is also influenced by factors such as cloud cover and temperature. Compared to wind power, its fluctuations are weaker on the 15-minute to hourly scale, but stronger within 5 minutes. The main influencing factor for wind power generation is wind speed, and wind turbine output exhibits a significant anti-peak-shaving characteristic, easily leading to wind curtailment and load shedding. Therefore, stochastic modeling of renewable energy to characterize the uncertainty of wind and solar power output is of significant research value.

[0173] The process of constructing a photovoltaic power generation model:

[0174] External factors such as light intensity and temperature affect photovoltaic (PV) output. Among these, light intensity has the strongest correlation with PV output. Statistically, the distribution pattern of light intensity over a certain time interval can be described by a Beta distribution, and its probability density function can be expressed as:

[0175]

[0176] In the formula: This represents the maximum output power of the solar cell array.

[0177] The relationship between photovoltaic power output and solar irradiance can be expressed as:

[0178]

[0179] In the formula: , Photovoltaic output power and rated power, in kW; This is the rated light intensity.

[0180] The functional relationship between photovoltaic output power and light intensity is as follows: Figure 4 As shown.

[0181] The process of constructing a wind power generation model may include:

[0182] Because wind speed is greatly affected by season and region, it exhibits strong asymmetry and peak characteristics. Research shows that the probability density function of wind speed follows a two-parameter Weibull distribution, which can be expressed as:

[0183]

[0184] Where k is the shape coefficient, which describes the curve shape of the probability density function; c is the scale coefficient, which reflects the average wind speed over a certain period; and v is the actual wind speed, in m / s.

[0185] Considering the characteristics of wind speed and wind turbine output function, the wind power output model can be expressed as follows:

[0186]

[0187] In the formula: , The output power and rated power of the wind turbine are in kW. , , This refers to the cut-in wind speed, cut-out wind speed, and rated wind speed of the wind turbine. The functional relationship between the wind turbine's output power and wind speed is as follows: Figure 5 As shown.

[0188] Minimize the cost of purchasing electricity from the upper-level power system. Considering the influence of uncertainties, the objective is to maximize the expected value under stochastic conditions, and its objective function can be expressed as follows;

[0189]

[0190]

[0191] In the formula: E represents the expectation operator.

[0192] Secondly, considering that traditional methods using fixed weight coefficients are difficult to adapt to dynamic operating scenarios in multi-objective optimization problems of distribution networks, this invention proposes a dynamic weight adaptive mechanism based on reinforcement learning. This mechanism autonomously generates the optimal combination of objective weights by sensing the system state in real time, thus resolving the time-varying conflict problem among multiple objectives. The multi-objective optimization process based on MOHHO-DDPG can be referenced. Figure 6 .

[0193] The multi-objective Harris Eagle algorithm can be used: The Harris Eagle algorithm (HHO) mimics the group hunting behavior and ambush strategy of eagle flocks, performing global exploration and local exploitation, and has advantages such as fewer parameters to be tuned and ease of implementation. Global exploration is defined as follows:

[0194]

[0195] In the formula: —Harris Eagle in the The position and random position at the next iteration; —The position of the optimal individual; —Population center location; Random numbers within an interval; —The upper and lower limits of the search space.

[0196] The HHO algorithm is based on the energy of prey escaping. Shift from global exploration to local development (At that time), and thus change different development strategies:

[0197]

[0198] Implement the following encirclement strategy:

[0199]

[0200] In the formula: A random number between [a certain number]. When At that time, execute the following rapid dive strategy:

[0201]

[0202] If the fitness level does not improve when compared to the current position, indicating a failed rapid dive, the flock will then conduct a levy flight random walk.

[0203]

[0204] In the formula: —Location update; —Dimensions A random vector. Define the following multi-objective optimization:

[0205]

[0206] Its solution is a set of Pareto optimal solutions. Given a vector and ,say Pareto dominance If and only if for

[0207]

[0208] untie It is the Pareto optimal solution, if The set of Pareto optimal solutions is denoted as:

[0209]

[0210] Furthermore, This is known as the Pareto front.

[0211] The process of selecting target weight combinations based on deep learning can include:

[0212] DDPG (Deep Deterministic Policy Gradient) is a type of reinforcement learning algorithm. While classic Q-learning methods are suitable for decision-making problems in discrete spaces, DDPG demonstrates more powerful decision-making capabilities for problems with continuous state-action spaces. It incorporates different types of neural networks: a policy network generates the actions the agent will take, and a value network evaluates the output of the policy network.

[0213] In this embodiment of the invention, the following definition is given for the distribution network scheduling optimization problem: State It consists of continuous historical power and load characteristic data. ,in This is the input length of the model. Action Describe the selection scheme for the Pareto optimal solution set. ,in This represents the total number of actions the agent can take, i.e., the number of Pareto optimal solutions. When choosing the... When there is a solution, there are .

[0214]

[0215] In the formula; — Fixed value.

[0216] Specifically, the process of optimizing distribution network scheduling may include the following steps: distributed control of multiple resource types in the distribution network based on ADMM-reinforcement learning collaborative optimization, such as... Figure 7 As shown, the process is as follows:

[0217] (1) Load the data, initialize the parameters, and set the number of iterations k=1.

[0218] (2) Start the aggregate system layer optimization scheduling.

[0219] (3) Communication is performed between adjacent aggregates, and the interaction power of their respective tie lines is calculated. .

[0220] (4) All aggregates simultaneously solve for decision variables. .

[0221] (5) Calculate the dual variables for each aggregate separately. .

[0222] (6) Global update dual variable and auxiliary variables .

[0223] (7) Update the decision variables of each aggregate. .

[0224] (8) Using MOHHO (Multi-Objective Harris Hawks Optimization), solve the multi-objective combinatorial optimization problem to obtain the Pareto optimal solution set of the combinatorial weight vector. Define the Pareto optimal solution set as the agent's action vector, and use the DDPG algorithm to solve for the dynamic optimal weights of the combinatorial prediction. Obtain the final weight result.

[0225] (9) Calculate the original residual and dual residual for each aggregate and determine whether it converges.

[0226] (10) If the convergence accuracy is satisfied, the iteration stops, the optimization result is obtained, and step (11) is continued. If convergence is not satisfied, the process returns to step (3).

[0227] (11) Optimize the system layer results Substitute the local layer optimization, local load value, and renewable energy prediction value to solve for the output of each distributed power source.

[0228] (12) Update the distributed power source operating cost function in the system layer optimization.

[0229] (13) Two-layer optimization stops.

[0230] Each aggregate only needs to exchange boundary decision variables in each iteration. No decision variables required It enables global interaction without involving the internal system operation data of the aggregate. Throughout the computation process, it reduces the communication costs of the cluster system and protects the privacy information of each aggregate operator.

[0231] ADMM (Alternating Direction Method of Multipliers) is an iterative algorithm for solving decomposable optimization problems, particularly suitable for distributed computing scenarios. Its core idea is to break down a large optimization problem into multiple sub-problems, gradually approaching the global optimum through alternating optimization and coordination. Considering that in traditional centralized optimization, all data needs to be aggregated to a central node (such as a dispatch center) for unified calculation, resulting in a heavy computational and communication burden, ADMM addresses this by decomposing the original problem into multiple sub-problems, each relying only on local data (such as the power generation and load information of a specific aggregate). In the distribution network dispatch optimization problem of this invention embodiment: each aggregate (microgrid / park) only needs to calculate its local optimum based on its own power parameters and load demand, without needing to know the detailed data of other aggregates. Global consistency is coordinated through interactive power (coupling variables).

[0232] The ADMM iterative process consists of two steps: local optimization, where each aggregate solves its own sub-problems in parallel (such as optimizing local power output), and global coordination, where information is exchanged through interactive power or price signals (Lagrange multipliers) to adjust their respective strategies. Example: Aggregate A calculates its optimal scheduling of photovoltaic + energy storage, while aggregate B calculates its optimal scheme for wind power + load. Through ADMM iterative coordination and interactive power exchange, they ultimately achieve global optimum.

[0233] In this embodiment of the invention, the aggregates do not need to share sensitive data (such as user electricity usage habits or cost functions), but only need to exchange boundary variables (such as interactive power), which can effectively protect the privacy of devices within the aggregates. In addition, since only a small amount of coordination information (such as Lagrange multipliers) needs to be transmitted in each iteration, it is suitable for scenarios with limited communication resources (such as distributed energy networks).

[0234] In embodiments of the present invention, such as Figure 8 As shown, the distributed characteristics of ADMM are highly compatible with the physical structure of the distribution network. The main problem (centralized layer) is to coordinate the interactive power between various aggregates to ensure global power balance. The sub-problems (distributed layer) are: each aggregate optimizes its own output based on local constraints (such as power ramp rate and energy storage capacity). Through iterative convergence, local decision-making and global coordination variables are updated alternately to finally obtain a distributed and consistent optimal solution.

[0235] Example scenario: Coordinated dispatch of three microgrids (A / B / C) in three industrial parks; Initial state: A (excess photovoltaic power): 60kW output available; B (peak load): 40kW deficit available; C (balanced state): 20kW available for reception. The iterative process may include:

[0236] 1. Initial solution at the system level: A outputs 40kW to B, and A outputs 20kW to C;

[0237] 2. Local optimization in section A revealed: Line limitation: A output to B is a maximum of 30kW; Feedback correction: A outputs 30kW to B, and A outputs 20kW to C;

[0238] 3. System-level recalculation: The 10kW shortfall in B will be replaced by C (C will provide 10kW of power to B).

[0239] 4. The final convergence target output scheme: 30kW from A to B, 10kW from A to C, and 10kW from C to B.

[0240] In this embodiment of the invention, the MOHHO algorithm reveals all possible optimization directions (i.e., the Pareto front). RL selects the most robust optimization direction based on real-time data. ADMM is used for distributed execution, protecting privacy and ensuring scalability. The system-level aggregator collects power output and load demand information from the local-level aggregator, using the interaction power between aggregators as a coupling variable to achieve centralized and distributed transformation. The proposed ADMM is used for distributed iterative solution. After the solution is completed, the system-level control center adjusts the energy interaction of aggregators and the power trading with the distribution network to achieve power regulation, while simultaneously transmitting the total regulated power of each aggregator to the local-level aggregator operator. The local-level optimization problem is solved using a centralized method. After the solution is completed, parameters are updated based on the regulating capacity power of each unit, operating costs, penalty costs, etc., and fed back to the system-level control center. Alternating iterations between the system-level and local-level layers are implemented to solve for the optimal trading volume of the aggregator, the optimal trading volume between the operator and the distribution network, and the optimal power output of each aggregator. Simultaneously, the EMC (Energy Management Center) transmits the total power generation of each aggregator as a known quantity to the aggregator operator. The local layer optimization problem is solved centrally by each control center, and the optimization results serve as the basis for scheduling the output of each distributed power source in the aggregate. After the solution is completed, the relevant system parameters of the aggregate are updated based on the output and operating costs of each unit, and fed back to the aggregate system EMC at the system layer. The system layer and the local layer iterate alternately to solve for the optimal interaction between aggregates, the optimal power transaction between the aggregate and the distribution network, and the optimal output of each distributed power source, until the decision variables of the aggregate system EMC no longer change.

[0241] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0242] Based on the same inventive concept, this application also provides a distribution network dispatch optimization device for implementing the distribution network dispatch optimization method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more distribution network dispatch optimization device embodiments provided below can be found in the limitations of the distribution network dispatch optimization method described above, and will not be repeated here.

[0243] In one exemplary embodiment, such as Figure 9 As shown, a power distribution network dispatch optimization device is provided, wherein:

[0244] The acquisition module is used to acquire power output information and load demand information of multiple aggregates within the distribution network.

[0245] The first determining module is used to determine the reference interaction power between the plurality of aggregates based on the power output information and load demand information;

[0246] An optimization module is used to optimize the reference interaction power based on the parameter information of the power supply device within each aggregate until a preset convergence condition is met, thereby obtaining the target interaction power between the aggregates.

[0247] The second determining module is used to determine the target output information of the power supply device in each of the aggregates based on the target interactive power and the parameter information of the power supply device in each aggregate.

[0248] The scheduling module is used to schedule the energy of the power supply equipment in each of the aggregates according to the target output information.

[0249] Each module in the aforementioned power distribution network dispatch optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0250] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10The computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a power distribution network scheduling optimization method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0251] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0252] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0253] Obtain power output information and load demand information for multiple aggregates within the distribution network;

[0254] Based on the power output information and load demand information, the reference interaction power among the multiple aggregates is determined;

[0255] The reference interactive power is optimized by each of the aggregates based on the parameter information of the power supply devices within the aggregates until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates.

[0256] Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, the target output information of the power supply devices in each of the aggregates is determined respectively.

[0257] Energy scheduling is performed on the power supply equipment within each of the aggregates based on the target output information.

[0258] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0259] Obtain power output information and load demand information for multiple aggregates within the distribution network;

[0260] Based on the power output information and load demand information, the reference interaction power among the multiple aggregates is determined;

[0261] The reference interactive power is optimized by each of the aggregates based on the parameter information of the power supply devices within the aggregates until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates.

[0262] Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, the target output information of the power supply devices in each of the aggregates is determined respectively.

[0263] Energy scheduling is performed on the power supply equipment within each of the aggregates based on the target output information.

[0264] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0265] Obtain power output information and load demand information for multiple aggregates within the distribution network;

[0266] Based on the power output information and load demand information, the reference interaction power among the multiple aggregates is determined;

[0267] The reference interactive power is optimized by each of the aggregates based on the parameter information of the power supply devices within the aggregates until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates.

[0268] Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, the target output information of the power supply devices in each of the aggregates is determined respectively.

[0269] Energy scheduling is performed on the power supply equipment within each of the aggregates based on the target output information.

[0270] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0271] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0272] 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 application.

[0273] 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 application should be determined by the appended claims.

Claims

1. A method for optimizing distribution network dispatch, characterized in that, The method includes: Obtain power output information and load demand information for multiple aggregates within the distribution network; Based on the power output information and load demand information, the reference interaction power among the multiple aggregates is determined; The reference interactive power is optimized by each of the aggregates based on the parameter information of the power supply devices within the aggregates until a preset convergence condition is met, thereby obtaining the target interactive power between the aggregates. Based on the target interactive power and the parameter information of the power supply devices in each of the aggregates, the target output information of the power supply devices in each of the aggregates is determined respectively. Energy scheduling is performed on the power supply equipment within each of the aggregates based on the target output information.

2. The method according to claim 1, characterized in that, The process of determining the target output information includes: Using the interaction power between the multiple aggregates as coupling variables, a global scheduling optimization model corresponding to the distribution network is constructed based on the power output information and load demand information. The reference interaction power is obtained by solving the global scheduling optimization model. Based on the reference interactive power and the parameter information of the power supply devices in each aggregate, a local scheduling optimization model corresponding to each aggregate is constructed. The local scheduling optimization model corresponding to each of the aggregates is solved to obtain the local optimal interaction power of each aggregate. The model parameters of the global scheduling optimization model are updated based on the local optimal interaction power. The process then returns to the step of solving the global scheduling optimization model until the convergence condition is met. Finally, the optimal solution of the global scheduling optimization model is output as the target interaction power.

3. The method according to claim 2, characterized in that, The power output information includes the real-time output capacity of the power equipment, the charging and discharging status of the energy storage system, and adjustable power parameters; the load demand information includes rigid load power, flexible load adjustment capability, and load forecast data. The process of determining the global scheduling optimization model includes: Based on the real-time output capacity and adjustable power parameters of the power supply equipment, determine the local feasible output range of each aggregate; Based on the rigid load power and flexible load adjustment capability, determine the net load requirement of each polymer; Using the interaction power between the multiple aggregates as coupling variables, and constructing model constraints based on the local feasible output range and net load requirements, the global scheduling optimization model is obtained. The process of determining the local scheduling optimization model includes: For each of the aggregates, a local scheduling optimization model is constructed based on the reference interactive power, the charging and discharging state of the energy storage system of the aggregate, and the power supply equipment parameters.

4. The method according to claim 1, characterized in that, The power supply equipment's parameter information includes rated capacity, upper and lower output limits, ramp rate, and operating cost coefficient. The process of determining the target output information includes: A local scheduling optimization model for the aggregate is constructed with the objective functions of minimizing operating costs and penalizing interaction power deviation. For the local scheduling optimization model, the optimal power allocation scheme that satisfies the equipment physical constraints and power balance constraints is solved to obtain the target power information.

5. The method according to claim 2, characterized in that, The solution process of the global scheduling optimization model includes: The Pareto solution set of the global scheduling optimization model is generated based on a multi-objective optimization algorithm; The Pareto solution set is dynamically selected based on the real-time state of the aggregate to obtain the target Pareto solution of the global scheduling optimization model; Based on the constraints corresponding to the target Pareto solution, the global scheduling optimization model is reconstructed to obtain the reconstructed model; The reconstructed model is solved to obtain the target interaction power.

6. The method according to claim 5, characterized in that, The process of determining the target Pareto solution includes: Define the multi-objective function corresponding to the global scheduling optimization model; Simulate the hunting behavior of eagle flocks to generate a set of multi-objective weight combinations corresponding to the multi-objective function; Construct the action space based on the multi-objective weight combination set; The state space is constructed based on the real-time state and the convergence state of the local scheduling optimization model; Actions are selected in the action space based on a pre-defined policy network, and the results of action selection are evaluated based on a pre-defined evaluation network. The evaluation result is selected based on the action output by the evaluation network, and the target Pareto solution is obtained from the multi-objective weight combination.

7. A power distribution network dispatch optimization device, characterized in that, The device includes: The acquisition module is used to acquire power output information and load demand information of multiple aggregates within the distribution network. The first determining module is used to determine the reference interaction power between the plurality of aggregates based on the power output information and load demand information; An optimization module is used to optimize the reference interaction power based on the parameter information of the power supply devices within each aggregate until a preset convergence condition is met, thereby obtaining the target interaction power between the multiple aggregates. The second determining module is used to determine the target output information of the power supply device in each of the aggregates based on the target interactive power and the parameter information of the power supply device in each aggregate. The scheduling module is used to schedule the energy of the power supply equipment in each of the aggregates according to the target output information.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images