Source-load uncertainty-oriented photovoltaic storage direct-flexible distributed model prediction scheduling method and system

Abstract

The application discloses a light storage direct flexible distributed model prediction scheduling method and system for source load uncertainty, obtains the prediction data and the actual measurement data of each district control unit in each period in a historical natural cycle, obtains the current state of charge of each district energy storage unit, dynamically calibrates the adaptive node uncertainty set boundary of photovoltaic power generation output and load demand for each district and each period based on the bias mean, the bias standard deviation and the current state of charge, so that each district can independently set the uncertainty set boundary according to the historical prediction accuracy, the node with high prediction accuracy is no longer excessively constrained, and the node with low prediction accuracy can also obtain sufficient conservatism, and through the conservative degree additional adjustment of the energy storage state of charge coupling, the problem that there is no feasible solution in optimization due to the too large uncertainty set when the adjustment capacity is insufficient is avoided.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, and in particular to a distributed model predictive scheduling method and system for photovoltaic-storage-DC-flexible systems oriented towards source-load uncertainty. Background Technology

[0002] In rural power distribution networks with a high proportion of distributed photovoltaic and flexible loads, there are significant prediction uncertainties on both the source and load sides. Existing distributed model predictive control methods usually use a globally unified uncertainty set to describe the prediction error of all nodes, without considering the inconsistency of prediction error characteristics caused by differences in geographical location, photovoltaic orientation, and load type in different transformer areas. This results in nodes with high prediction accuracy being over-constrained and losing economic efficiency, while nodes with low prediction accuracy are under-constrained and have the risk of exceeding limits.

[0003] Furthermore, the current common practice is to use a fixed-period neighbor information exchange mechanism. Each round of optimization iteration requires the transmission of complete power trajectory data in the prediction time domain. Under the conditions of limited communication bandwidth, high latency, and high packet loss rate in rural power distribution networks, dense fixed-period communication is difficult to guarantee. Once communication is interrupted, the convergence of distributed optimization will be severely affected. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose a predictive scheduling method and system for a distributed optical-storage-direct-flexible model oriented to address source-load uncertainty.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by this invention is as follows: In a first aspect, the present invention provides a predictive scheduling method for a distributed optical-storage-direct-flexible model oriented to address source-load uncertainty, comprising, in some embodiments: The system acquires the predicted and actual measurement data of each control unit in each historical natural cycle, and calculates the mean and standard deviation of the prediction error independently for each area and each time period. At the same time, it acquires the current state of charge of each energy storage unit in the area. Based on the mean deviation, standard deviation deviation, and current state of charge, the adaptive nodal uncertainty set boundary of photovoltaic power generation output and load demand is dynamically calibrated for each transformer area and each time period. Each control unit of a transformer area aims to minimize the operating cost of its own area by establishing a local model predictive control optimization problem, which embeds power interaction coupling terms between adjacent transformer areas and uses adaptive node uncertainty sets as constraints. Each control unit in each transformer area solves the local optimization problem in parallel and checks the preset event triggering conditions after solving. The event triggering conditions include the expansion of the boundary of the adaptive node uncertainty set exceeding the first preset threshold, or the deviation between the optimal power exchange plan of the transformer area and the power exchange reference value provided by the neighboring transformer area exceeding the second preset threshold. When the event triggering condition is met, the local power exchange reference trajectory in the prediction time domain starting from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time are sent to the relevant neighboring transformer areas. After receiving information from its neighbors, each control unit in each area re-solves the local optimization problem, performs distributed iterations until convergence or the preset maximum number of iterations is reached, and then executes the first element of the final determined control sequence.

[0006] In some embodiments, based on the mean deviation, standard deviation deviation, and current state of charge, the adaptive nodal uncertainty set boundary of photovoltaic power generation output and load demand is dynamically calibrated for each transformer area and each time period, including: The historical natural cycle is divided into several time periods. For each photovoltaic power generation unit and various loads in each distribution area, the mean deviation and standard deviation of the historical prediction error for each time period are calculated. The sum of the mean deviation and the standard deviation of the preset multiple is used as the initial upper boundary of the uncertain set of photovoltaic power output in this area during this period, and the difference between the mean deviation and the standard deviation of the preset multiple is used as the corresponding initial lower boundary. Obtain the current state of charge of the energy storage unit, and determine the conservatism factor based on the preset range in which the current state of charge is located; When the current state of charge is higher than the first state of charge threshold, the conservatism coefficient corresponding to the initial upper boundary of the uncertain set of photovoltaic power output is increased to generate the first corrected upper boundary. When the current state of charge is lower than the second state of charge threshold, the conservatism factor corresponding to the initial upper boundary of the load demand uncertainty set is increased to generate a second corrected upper boundary. The first or second modified upper boundary, together with the corresponding initial lower boundary, is established as the final adaptive node uncertainty set boundary.

[0007] In some embodiments, each control unit in a transformer area establishes a local model predictive control optimization problem with the objective of minimizing the operating cost of its area, including: The objective function is constructed by minimizing the sum of the cost of purchasing electricity from the AC side, the cost of energy storage charging and discharging losses, and the cost of comfort loss due to flexible load regulation in the local area within the predicted time domain. The energy storage power constraint, along with the dynamic state of charge constraint, the flexible load adjustment range constraint, the DC bus power balance constraint, and the photovoltaic power output uncertainty constraint and load demand uncertainty constraint generated by the adaptive node uncertainty set boundary, are collectively established as the set of constraints for the objective function. The interaction cost or voltage deviation penalty term when adjacent transformer substations exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term into the objective function or constraint set. The power interaction coupling term uses the power exchange reference value provided by the previous round of optimization of the neighboring transformer substation as a known parameter, rather than simultaneously optimizing the decision variables of the neighboring transformer substation.

[0008] In some embodiments, the interaction cost or voltage deviation penalty term when adjacent transformer areas exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term into the objective function or set of constraints, including: Obtain the power exchange reference value sequence between the neighboring transformer area and the interconnection node at each time in the prediction time domain after the previous round of optimization, as a known external input parameter; The power exchange decision variable corresponding to the interconnection node in the current optimization variables of this substation area is compared with the reference value at the corresponding time in the power exchange reference value sequence to obtain the power exchange deviation at each time in the prediction time domain. A preset penalty coefficient is applied to the power exchange deviation to form an interaction cost penalty term. The value of the penalty coefficient is dynamically set according to the electrical distance between this transformer area and neighboring transformer areas or the capacity margin of the flexible interconnection device. By directly incorporating the interaction cost penalty term into the objective function, the local power exchange decision variable of this transformer area tends to approximate the reference value provided by the neighboring transformer area when optimizing, so as to achieve implicit coordination of power exchange plans between adjacent transformer areas, while maintaining the independent solution characteristics of the optimization problem of each transformer area.

[0009] In some embodiments, each control unit in a distribution area solves the local optimization problem in parallel and checks preset event triggering conditions after solving, including: Each control unit in each transformer area solves the local optimization problem in parallel to obtain the optimal power exchange sequence and optimal control sequence for its respective transformer area in the prediction time domain. After the solution is completed, calculate the expansion magnitude of the adaptive node uncertainty set boundary of the current optimization time of this station area at the start time of the prediction time relative to the corresponding boundary of the previous optimization time. The expansion range is compared with the first preset threshold. If the expansion range exceeds the first preset threshold, the triggering condition for the first type of event is determined to be met. Calculate the absolute value of the deviation between the first power exchange value at the interconnection node with the neighboring transformer area in the optimal power exchange sequence of this transformer area and the corresponding power exchange reference value provided by the neighboring transformer area in the previous round; The absolute value of the deviation is compared with the second preset threshold. If the absolute value of the deviation exceeds the second preset threshold, the triggering condition of the second type of event is determined to be met. The first type of event triggering condition and the second type of event triggering condition are combined by a logical OR relationship. When at least one condition is true, the event triggering condition is determined to be satisfied.

[0010] In some embodiments, when the event triggering condition is met, the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time are sent to the relevant neighboring transformer areas, including: When the triggering condition of the first type of event is met, the boundary parameters of the adaptive node uncertainty set corresponding to the current optimization time of this substation are packaged as the first type of communication data, and the sending object is determined to be the neighboring substation that is electrically closely coupled with the expansion of the boundary of the adaptive node uncertainty set. When the second type of event triggering condition is met, the power exchange sequence at the interconnection node of the neighboring substation in the predicted time domain obtained by solving the current optimization time of this substation is packaged as the second type of communication data, and the sending object is determined to be the specific neighboring substation corresponding to the absolute value of the deviation. When both the first type of event triggering condition and the second type of event triggering condition are met, the first type of communication data and the second type of communication data are merged and packaged, and the sending object is determined to be the union of the neighboring stations pointed to by the first type of communication data and the second type of communication data; The packaged communication data is sent to the designated recipient via the communication link. After the transmission is completed, the local control unit enters a waiting state, waiting to receive corresponding communication data from neighboring units for the next round of local optimization problem solving.

[0011] In some embodiments, when the first type of event triggering condition is met, the adaptive node uncertainty set boundary parameters corresponding to the current optimization time of this transformer area are packaged as the first type of communication data, and the sending target is determined to be a neighboring transformer area with tight electrical coupling associated with the expansion of the adaptive node uncertainty set boundary, including: Identify the expansion direction and magnitude of the boundary of the adaptive node uncertainty set at the current optimization moment of this transformer area compared to the boundary corresponding to the previous optimization moment. The expansion direction points to the expansion of the upper or lower boundary of the photovoltaic power generation output uncertainty set, or the expansion of the upper or lower boundary of the load demand uncertainty set. Based on the expansion direction, a list of neighboring transformer substations with strong coupling relationship to the power flow direction corresponding to the expansion direction is queried from the preset neighboring transformer substation association table. The neighboring transformer substation association table stores the electrical coupling tightness of each transformer substation under different power flow directions in advance. The top-ranked number of neighboring stations in the list of neighboring stations based on their coupling degree are identified as the recipients of the first type of communication data. The local area identifier, the timestamp of the current optimization time, and the expanded boundary value and corresponding time period identifier in the boundary parameters of the adaptive node uncertainty set are encapsulated together into a first type of communication data packet and sent to the determined sending object through the communication link.

[0012] In some embodiments, after receiving neighbor information, each control unit in each transformer area re-solves the local optimization problem, performs distributed iterations until convergence or a preset maximum number of iterations is reached, and executes the first element of the finally determined control sequence, including: Each control unit updates the power exchange reference trajectory or adaptive node-level uncertainty set parameters received from neighboring control units into the power interaction coupling terms or uncertainty constraints in the local model predictive control optimization problem, and re-solves the local optimization problem to obtain the local optimal power exchange sequence for this iteration. The local optimal power exchange sequence obtained in this iteration is compared with the local optimal power exchange sequence obtained in the previous iteration. The sum of the absolute differences between the two power values ​​at corresponding times in the prediction time domain is calculated as the iterative change of this station area. The iterative change is compared with a preset convergence threshold. When the iterative change of all transformer areas is less than the preset convergence threshold, the distributed iterative convergence is determined, and the control unit of each transformer area executes the first element of the local optimal control sequence obtained in this round of iteration. When the iterative change of any region is greater than or equal to the preset convergence threshold, the distributed iteration is determined to be non-converged, and it is checked whether the current iteration round has reached the preset maximum number of iterations. If the current iteration round has not reached the preset maximum number of iterations, each control unit will use the local optimal power exchange sequence obtained in this iteration as the new power exchange reference value, send it to the relevant neighboring control units through the communication link, and enter the next iteration round. If the current iteration round has reached the preset maximum number of iterations, then each control unit calculates the weighted average of the local optimal power exchange sequence in all iteration rounds as the consensus power exchange sequence, and executes the first element of the local optimal control sequence corresponding to the consensus power exchange sequence.

[0013] In some embodiments, the predicted data and actual measurement data of each control unit in a historical natural cycle are obtained, and the mean deviation and standard deviation of the prediction error are independently calculated for each control unit and each time period, including: The historical natural cycle is divided into several continuous time periods with a preset time granularity. For each photovoltaic power generation unit and various loads in each transformer area, the day-ahead predicted power data and real-time measured power data corresponding to each time period are extracted as raw data pairs. Calculate the difference between the predicted power data and the real-time measured power data for each raw data pair to obtain the historical prediction error value of the equipment for that period. All historical prediction error values ​​of the same transformer area, the same time period, and the same type of equipment are collected into a prediction error sample set for that transformer area, the same time period, and that type of equipment. The prediction error sample set is fitted with a normal distribution, and the mean deviation and standard deviation of the deviation are calculated. The mean deviation represents the degree of systematic deviation of the prediction of the equipment type in the same area and time period, and the standard deviation represents the degree of dispersion of the prediction of the equipment type in the same area and time period.

[0014] In a second aspect, the present invention also provides a distributed model prediction and scheduling system for optical-storage direct-flexible systems accommodating source-load uncertainty, used in the method described in the first aspect, the system comprising: The data acquisition module is used to acquire the predicted and actual measurement data of each control unit in the historical natural cycle, as well as the current state of charge of each energy storage unit in the distribution area. The adaptive uncertainty set construction module, connected to the data acquisition module, is used to independently calculate the mean and standard deviation of the prediction error for each transformer area and each time period based on the predicted data and the actual measurement data. Based on the mean and standard deviation of the deviation and the current state of charge, the adaptive node uncertainty set boundary of photovoltaic power generation output and load demand is dynamically calibrated. The local optimization solution module is set in the control unit of each transformer area. It is used to establish and solve the local model predictive control optimization problem with the goal of minimizing the operating cost of the local transformer area. The optimization problem embeds the power interaction coupling term between adjacent transformer areas and uses the adaptive node uncertainty set as a constraint condition. The event trigger judgment module is connected to the local optimization solution module. It is used to check whether the preset event trigger conditions are met after the local optimization solution module solves the problem. The event trigger conditions include the expansion of the boundary of the adaptive node uncertainty set exceeding the first preset threshold, or the deviation between the optimal power exchange plan of this area and the power exchange reference value provided by the neighboring area exceeding the second preset threshold. The communication interaction module connects to the event trigger judgment module and is used to send the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters of the current optimization time to the relevant neighboring transformer areas when the event trigger condition is met. The distributed iterative convergence module connects the communication interaction module and the local optimization solution module. After the communication interaction module receives neighbor information, it triggers the local optimization solution module to resolve the problem and perform distributed iterations until convergence or the preset maximum number of iterations is reached. Finally, it outputs the first element of the final determined control sequence.

[0015] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: Unlike existing technologies, the above technical solution acquires the predicted and actual measurement data of each control unit in each historical natural cycle, and calculates the mean and standard deviation of the prediction error independently for each distribution area and each time period. At the same time, it acquires the current state of charge of each energy storage unit in each distribution area. Based on the mean deviation, the standard deviation, and the current state of charge, it dynamically calibrates the adaptive node uncertainty set boundary for photovoltaic power generation output and load demand for each distribution area and each time period. This allows each distribution area to independently set the uncertainty set boundary according to its own historical prediction accuracy. Nodes with high prediction accuracy are no longer over-constrained, and nodes with low prediction accuracy can also obtain sufficient conservatism. At the same time, through the additional adjustment of conservatism coupled with the energy storage state of charge, it avoids the problem of no feasible solution due to excessive uncertainty set when the adjustment capability is insufficient. Based on this, each distribution area control unit solves the local model predictive control optimization problem in parallel, and checks preset event triggering conditions after solving. These event triggering conditions include the expansion of the adaptive node uncertainty set boundary exceeding a first preset threshold, or the deviation between the optimal power exchange plan of the local distribution area and the power exchange reference value provided by neighboring distribution areas exceeding a second preset threshold. Neighbor information exchange only occurs when the triggering conditions are met. Communication frequency is significantly reduced during periods of stable source and load, and automatically increased when uncertainty increases, achieving an adaptive balance between communication burden and coordination accuracy. The above technical solution, through the construction of a three-dimensional adaptive uncertainty set at the node level, time period level, and energy storage coupling, reduces the conservatism of uncertainty handling and improves operational economy. Through the event-triggered communication mechanism, the communication burden of distributed control is reduced by more than 70%, adapting to the weak communication conditions of rural distribution networks, while ensuring control coordination during periods of severe source and load fluctuations. 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 This is a schematic diagram of steps S101 to S106 as described in the specific implementation method; Figure 2 This is a schematic diagram of steps S102 to S206 as described in the specific implementation method; Figure 3 This is a schematic diagram of the scheduling system described in the specific implementation.

[0018] Figure label: 1. Optical-storage-direct-flexible distributed model predictive scheduling system; 11. Data acquisition module; 12. Adaptive Uncertain Set Construction Module; 13. Local optimization solution module; 14. Event Trigger Judgment Module; 15. Communication and interaction module; 16. Distributed iterative convergence module. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 In a first aspect, this embodiment provides a predictive scheduling method for a distributed optical-storage-direct-flexible model oriented towards source-load uncertainty. In some embodiments, it includes: S101. Obtain the predicted data and actual measurement data of each control unit in the historical natural cycle for each substation, and calculate the mean deviation and standard deviation of the prediction error independently for each substation and each time period. At the same time, obtain the current state of charge of each energy storage unit in the substation. S102. Based on the mean deviation, standard deviation deviation, and current state of charge, dynamically calibrate the adaptive node uncertainty set boundary for photovoltaic power generation output and load demand for each transformer area and each time period. S103. Each control unit of the transformer area aims to minimize the operating cost of its own transformer area and establishes a local model predictive control optimization problem, in which the power interaction coupling term between adjacent transformer areas is embedded and the adaptive node uncertainty set is used as a constraint condition. S104. Each control unit in each transformer area solves the local optimization problem in parallel and checks the preset event triggering conditions after solving. The event triggering conditions include the expansion of the boundary of the adaptive node uncertainty set exceeding the first preset threshold, or the deviation between the optimal power exchange plan of this transformer area and the power exchange reference value provided by the neighboring transformer area exceeding the second preset threshold. S105. When the event triggering condition is met, send the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time to the relevant neighboring transformer areas. S106. After receiving the neighbor information, each control unit in each area re-solves the local optimization problem, performs distributed iterations until convergence or reaches the preset maximum number of iterations, and executes the first element of the finally determined control sequence.

[0021] In step S101, the acquired prediction and measurement data are derived from the operation logs or historical databases accumulated by the control units of each distribution area during continuous operation over a historical natural cycle. The natural cycle can be selected as a past quarter or a calendar year to cover the changes in source and load characteristics under different seasons. The granularity of time period division is determined by the length of the scheduling cycle, for example, with 15 minutes as a time period, totaling 96 time periods per day. The mean deviation reflects the direction and magnitude of the systematic deviation in the prediction of photovoltaic power generation output or load demand in that distribution area during that time period, while the standard deviation reflects the dispersion of the predicted values. Together, they constitute a statistical characteristic description of the prediction error of that distribution area during that time period. The current state of charge of the energy storage unit is acquired in real time through the battery management system and used to adjust the conservatism of the uncertainty set in subsequent steps.

[0022] In step S102, the dynamic calibration process of the adaptive node uncertainty set boundary integrates historical statistical characteristics and real-time adjustment capabilities. The sum of the mean deviation and the standard deviation of the deviation of a preset multiple constitutes the initial upper boundary of the uncertainty set, and the difference between the mean deviation and the standard deviation of the deviation of a preset multiple constitutes the initial lower boundary. The preset multiple is set according to the control accuracy requirements. The current state of charge of the energy storage unit is used to determine the conservatism additional coefficient. When the state of charge is close to the upper limit, the conservatism additional coefficient of the upper boundary of the photovoltaic power generation output uncertainty set is increased to reserve upward adjustment space for possible surplus photovoltaic power; when the state of charge is close to the lower limit, the conservatism additional coefficient of the upper boundary of the load demand uncertainty set is increased to reserve downward adjustment space for possible load surges. This achieves personalized compensation for the differentiated prediction accuracy of different distribution areas and different time periods.

[0023] In step S103, the objective function of the local model predictive control optimization problem consists of three terms: the cost of purchasing electricity from the AC side for the local distribution area, the cost of energy storage charging and discharging losses, and the cost of comfort loss in flexible load regulation. The cost of purchasing electricity reflects the economic cost of obtaining electricity from the upper-level distribution network for the local distribution area. The cost of energy storage losses characterizes the reduction in energy storage lifespan due to charging and discharging cycles. The cost of comfort loss in load regulation quantifies the user's dissatisfaction when flexible loads are regulated. The power interaction coupling term uses the power exchange reference value provided by the previous optimization of adjacent distribution areas as a known parameter. By penalizing the deviation of the local distribution area's power exchange decision variable from this reference value, implicit coordination of the power exchange plan between adjacent distribution areas is achieved without simultaneously optimizing the decision variables of neighboring distribution areas, thereby maintaining the independence of the optimization problem for each distribution area.

[0024] In step S104, after each control unit of a transformer area solves its local optimization problem in parallel, it calculates the determination results of two types of event triggering conditions. The first type of triggering condition is based on the expansion magnitude of the adaptive node uncertainty set boundary. It calculates the change in the boundary at the start of the prediction time domain relative to the corresponding boundary at the previous optimization time. When the change exceeds a first preset threshold, it indicates a significant change in the transformer area's perception of future uncertainty. The second type of triggering condition is based on the deviation of the power exchange plan. It calculates the absolute value of the deviation between the first power exchange value at the interconnection node with the neighboring transformer area in the optimal power exchange sequence of the transformer area and the corresponding reference value provided by the neighboring transformer area in the previous round. When this deviation exceeds a second preset threshold, it indicates a significant divergence in the coordination plan between the transformer area and the neighboring transformer area. The two types of triggering conditions are combined using a logical OR relationship; satisfying either condition triggers communication.

[0025] In step S105, when the event triggering condition is met, the content of the transmitted communication data is associated with the type of triggering condition. Communication triggered by the first type of triggering condition transmits the adaptive node uncertainty set parameters at the current optimization time, enabling neighboring substations to perceive changes in their uncertainty perception. Communication triggered by the second type of triggering condition transmits the local power exchange reference trajectory in the prediction time domain, enabling neighboring substations to update the coupling term parameters in their local optimization problem. When both types of triggering conditions are met simultaneously, the two types of data are transmitted together. The transmission target is precisely determined according to the type of triggering condition. Communication triggered by the first type of triggering condition is sent to neighboring substations electrically closely coupled to the direction of uncertainty set boundary expansion; communication triggered by the second type of triggering condition is sent to the specific neighboring substation corresponding to the deviation, avoiding broadcasting to all neighbors and wasting communication resources.

[0026] In step S106, after receiving neighbor information, each control unit in a transformer area updates the received power exchange reference trajectory or uncertainty set parameters into the local optimization problem and re-solves to obtain the local optimal power exchange sequence for the current iteration. The iterative change of the transformer area is obtained by calculating the sum of the absolute differences in power values ​​at corresponding times in the prediction time domain between the current and previous sequences. When the iterative changes of all transformer areas are less than a preset convergence threshold, the distributed iteration converges, and each transformer area executes the first element of the current optimal control sequence. If the iterative change of any transformer area fails to converge and the current iteration round has not reached the preset maximum number of iterations, the current optimal power exchange sequence is sent as a new reference value to neighboring transformer areas, and the next iteration round begins. If convergence has not yet occurred after reaching the preset maximum number of iterations, the weighted average of the local optimal power exchange sequences in all iteration rounds is calculated as the consensus value, and the first element of the corresponding control sequence is executed, thereby ensuring the continuity and smoothness of control when communication is limited or the iteration diverges.

[0027] This embodiment calculates the mean and standard deviation of the prediction error independently for each distribution area and each time period, and dynamically calibrates the adaptive node uncertainty set boundary by combining it with the current state of charge of the energy storage unit. This allows each distribution area to customize the conservatism of the uncertainty set according to its own historical prediction accuracy and real-time adjustment capability, thereby solving the problem of excessive constraint on nodes with high prediction accuracy and loss of economy caused by the use of a globally unified uncertainty set in existing technologies. By using the expansion of the adaptive node uncertainty set boundary and the deviation between the optimal power exchange plan of the distribution area and the power exchange reference value provided by the neighboring distribution area as the event triggering condition, distributed communication is only carried out when there is a significant change in the perception of uncertainty or a large divergence in the coordination plan, thereby significantly reducing the communication frequency during periods of stable source and load. When the source load fluctuates drastically, the communication coordination density is automatically increased, solving the problem that the existing fixed-period communication mechanism is difficult to guarantee under the weak communication conditions in rural areas. By embedding the power exchange reference value provided by the previous optimization of neighboring substations as a known parameter into the power interaction coupling term of the local optimization problem, each substation can achieve implicit coordination of the power exchange plan while maintaining the independence of the optimization problem, thereby avoiding the dependence of centralized optimization on global communication. By calculating the weighted average of the local optimal power exchange sequence in all iteration rounds as the consensus value and executing the first element of the corresponding control sequence when the distributed iteration has not converged, the continuity and smoothness of control are guaranteed when communication is limited or the iteration diverges, thereby improving the engineering applicability and robustness of the entire scheduling method in the actual operation environment of rural power distribution networks.

[0028] Please see Figure 2 In some embodiments, based on the mean deviation, standard deviation deviation, and current state of charge, the adaptive nodal uncertainty set boundary of photovoltaic power generation output and load demand is dynamically calibrated for each transformer area and each time period, including: S201. Divide the historical natural cycle into several time periods, and calculate the mean deviation and standard deviation of the historical prediction error for each time period for each photovoltaic power generation unit and various loads in each distribution area. S202. The sum of the mean deviation and the standard deviation of the preset multiple is used as the initial upper boundary of the uncertain set of photovoltaic power generation output of the area during the time period, and the difference between the mean deviation and the standard deviation of the preset multiple is used as the corresponding initial lower boundary. S203. Obtain the current state of charge of the energy storage unit, and determine the conservatism coefficient based on the preset range in which the current state of charge is located. S204. When the current state of charge is higher than the first state of charge threshold, increase the conservatism coefficient corresponding to the initial upper boundary of the uncertain set of photovoltaic power output to generate the first corrected upper boundary. S205. When the current state of charge is lower than the second state of charge threshold, increase the conservatism coefficient corresponding to the initial upper boundary of the load demand uncertainty set to generate a second corrected upper boundary. S206. The first or second modified upper boundary, together with the corresponding initial lower boundary, is established as the final adaptive node uncertainty set boundary.

[0029] In step S201, the granularity of the historical natural cycle division is consistent with the length of the scheduling cycle, for example, with fifteen minutes as a time period, thereby ensuring that the statistical characteristics of the prediction error within each time period can reflect the unique source-load variation patterns of that time period. The mean deviation and standard deviation of the deviation are calculated independently for photovoltaic power generation units and various types of loads. This is because the prediction error of photovoltaic output is mainly affected by changes in meteorological conditions, while the prediction error of load demand is affected by user behavior patterns and seasonal factors. The statistical characteristics of the two are fundamentally different, and separate calculation can provide a more accurate data basis for the subsequent differential calibration of the uncertainty set boundary.

[0030] In step S202, the method for determining the initial upper and lower boundaries reflects the quantitative utilization of the statistical characteristics of historical prediction errors. The value of the preset multiple determines the baseline conservatism of the uncertainty set. This multiple can be set differently according to the power supply reliability requirements of different distribution areas. For example, a larger preset multiple can be used for distribution areas with a high proportion of critical loads, while a smaller preset multiple can be used for distribution areas with a high proportion of general loads, thereby achieving a differentiated balance between robustness and economy among different distribution areas.

[0031] In step S203, the current state of charge (SOC) of the energy storage unit is collected in real time by the battery management system. The preset range is divided according to the design capacity and safe operating range of the energy storage unit. The rule for determining the conservatism factor is as follows: when the SOC is in the middle range, the conservatism factor takes the baseline value; when the SOC deviates from the middle range, the conservatism factor increases accordingly to compensate for the potential power imbalance risk when the energy storage regulation margin is insufficient.

[0032] In step S204, a current state of charge higher than the first state of charge threshold indicates that the energy storage unit is close to full charge, and the adjustment space for further charging is limited. Increasing the conservatism coefficient corresponding to the initial upper boundary of the photovoltaic power output uncertainty set at this time is equivalent to reserving a larger upward fluctuation space for photovoltaic output in the optimization problem. This allows the optimization results to pre-arrange more power transmission or load increase when the actual photovoltaic output is higher than the predicted value, rather than relying on the nearly fully charged energy storage unit for absorption.

[0033] In step S205, a current state of charge below the second state of charge threshold indicates that the energy storage unit is nearing depletion, with limited room for downward discharge adjustment. Increasing the conservatism coefficient corresponding to the initial upper boundary of the load demand uncertainty set at this point is equivalent to reserving greater upward fluctuation space for load demand in the optimization problem. This allows the optimization results to pre-plan for more power purchases or energy storage charging reserves when the actual load demand exceeds the predicted value, rather than relying on the nearly depleted energy storage unit for discharge support.

[0034] In step S206, the final adaptive node uncertainty set boundary is determined by the modified upper boundary and the initial lower boundary. The modification is only applied to the upper boundary because the upper boundary corresponds to the risk direction of excessive photovoltaic output or excessive load demand, which happens to coincide with the risk direction of insufficient adjustment margin of the energy storage unit; while the risk direction corresponding to the lower boundary can usually be compensated by the normal charging and discharging of the energy storage unit, without the need to add extra conservatism.

[0035] This embodiment independently calculates the prediction errors of photovoltaic power generation units and various types of loads, enabling the calibration of the uncertainty set boundary to accurately reflect the essential differences in prediction errors of different equipment types, thereby avoiding feature confusion caused by unified statistics. By determining the conservatism coefficient based on the preset range of the current state of charge of the energy storage unit, and increasing the upper boundary of the uncertainty set of photovoltaic output when the state of charge is close to the upper limit, and increasing the upper boundary of the uncertainty set of load demand when the state of charge is close to the lower limit, the conservatism of the uncertainty set is dynamically coupled with the real-time adjustment capability of the energy storage unit. This maintains economy when the energy storage adjustment margin is sufficient and automatically enhances robustness when the adjustment margin is insufficient, solving the problem in the prior art where the fixed conservatism cannot adapt to changes in the energy storage state, making it difficult to balance optimization feasibility and economy.

[0036] In some embodiments, each control unit in a transformer area establishes a local model predictive control optimization problem with the objective of minimizing the operating cost of its area, including: The objective function is constructed by minimizing the sum of the cost of purchasing electricity from the AC side, the cost of energy storage charging and discharging losses, and the cost of comfort loss due to flexible load regulation in the local area within the predicted time domain. The energy storage power constraint, along with the dynamic state of charge constraint, the flexible load adjustment range constraint, the DC bus power balance constraint, and the photovoltaic power output uncertainty constraint and load demand uncertainty constraint generated by the adaptive node uncertainty set boundary, are collectively established as the set of constraints for the objective function. The interaction cost or voltage deviation penalty term when adjacent transformer substations exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term into the objective function or constraint set. The power interaction coupling term uses the power exchange reference value provided by the previous round of optimization of the neighboring transformer substation as a known parameter, rather than simultaneously optimizing the decision variables of the neighboring transformer substation.

[0037] In this embodiment, when constructing the objective function, the cost of purchasing electricity from the AC side is calculated based on the electricity purchase agreement price between the local distribution area and the upper-level distribution network, and the purchased power at each moment in the predicted time domain. A positive purchased power indicates power absorption from the upper-level grid, while a negative purchased power indicates power fed back to the upper-level grid. The electricity price at the time of feeding back is usually lower than the purchased electricity price to reflect the difference between the grid connection price and the purchased electricity price. The energy storage charging and discharging loss cost is calculated through the mapping relationship between charging and discharging power and the energy storage cycle life decay model. Frequent high-power charging and discharging will accelerate energy storage aging. Therefore, this cost item will guide the energy storage unit to prioritize the shallow charging and discharging operation mode during optimization. The comfort loss cost of flexible load adjustment is modeled according to the adjustable characteristics of various flexible loads. For example, the comfort loss of time-shiftable loads is related to the load transfer time length, the comfort loss of interruptible loads is related to the number of interruptions and the interruption duration, and the comfort loss of adjustable loads is related to the power adjustment range.

[0038] In this embodiment, when establishing the set of constraints, the energy storage power constraint limits the charging and discharging power of the energy storage unit to not exceed its rated power upper limit. The dynamic state of charge constraint describes the recursive relationship between the state of charge of the energy storage unit at each moment and the state of charge and charging and discharging power at the previous moment. The flexible load adjustment range constraint stipulates that the power adjustment amount of various flexible loads shall not exceed their preset adjustable upper and lower limits. The DC bus power balance constraint ensures that the sum of photovoltaic power generation output, energy storage charging and discharging power, flexible load power consumption, and power exchanged with the AC side through flexible interconnection devices is zero. This constraint is the core guarantee for real-time power balance within the distribution area. The photovoltaic power generation output uncertainty constraint and the load demand uncertainty constraint respectively limit the photovoltaic power generation and load demand within the allowable range determined by the boundary of the adaptive node uncertainty set, so that the optimization result can still satisfy power balance under the worst case of uncertainty.

[0039] In this embodiment, when embedding the power interaction coupling term, the interaction cost penalty term is obtained by calculating the deviation between the power exchange decision variable of the local transformer area and the power exchange reference value provided by the neighboring transformer area, and applying a preset penalty coefficient to this deviation. The value of the penalty coefficient is dynamically set according to the electrical distance between the local transformer area and the neighboring transformer area or the capacity margin of the flexible interconnection device. The closer the electrical distance or the smaller the capacity margin, the larger the penalty coefficient, indicating that the local transformer area should more strictly follow the reference value of the neighboring transformer area to avoid tie line overload or voltage exceedance. The voltage deviation penalty term is calculated based on the voltage deviation of the interconnection node between the local transformer area and the neighboring transformer area. When the voltage of the interconnection node deviates from the rated value, the penalty term increases, guiding the local transformer area to adjust its power exchange plan to support the voltage recovery of the interconnection node. By using the power exchange reference value of the neighboring transformer area as a known parameter rather than a decision variable, the optimization problem of each transformer area maintains the independent solution characteristic, eliminating the need to build a global optimization model, thereby reducing computational complexity and communication dependence.

[0040] This embodiment incorporates the costs of electricity purchase, energy storage loss, and the comfort loss from flexible load adjustment into the objective function, enabling the optimization result to comprehensively balance the three dimensions of economy, equipment lifespan, and user comfort, avoiding the unbalanced results caused by single-objective optimization. By transforming the boundary of the adaptive node uncertainty set into uncertain constraints on photovoltaic power generation output and load demand, the local optimization problem can perceive and adapt to the prediction uncertainty characteristics unique to the local distribution area, thereby maximizing economy while ensuring robustness. By embedding the power exchange reference values ​​provided by the previous optimization of neighboring distribution areas as known parameters into the power interaction coupling term, each distribution area can achieve implicit coordination of the power exchange plan while maintaining the independence of the optimization problem. This solves the technical problems of existing centralized optimization relying on global communication and distributed optimization requiring frequent exchange of neighbor decision variables, reducing communication burden and computational complexity.

[0041] In some embodiments, the interaction cost or voltage deviation penalty term when adjacent transformer areas exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term into the objective function or set of constraints, including: Obtain the power exchange reference value sequence between the neighboring transformer area and the interconnection node at each time in the prediction time domain after the previous round of optimization, as a known external input parameter; The power exchange decision variable corresponding to the interconnection node in the current optimization variables of this substation area is compared with the reference value at the corresponding time in the power exchange reference value sequence to obtain the power exchange deviation at each time in the prediction time domain. A preset penalty coefficient is applied to the power exchange deviation to form an interaction cost penalty term. The value of the penalty coefficient is dynamically set according to the electrical distance between this transformer area and neighboring transformer areas or the capacity margin of the flexible interconnection device. By directly incorporating the interaction cost penalty term into the objective function, the local power exchange decision variable of this transformer area tends to approximate the reference value provided by the neighboring transformer area when optimizing, so as to achieve implicit coordination of power exchange plans between adjacent transformer areas, while maintaining the independent solution characteristics of the optimization problem of each transformer area.

[0042] In this embodiment, the power exchange reference value sequence sent by neighboring transformers after the previous round of optimization, covering each time point in the prediction time domain with the interconnected node, is obtained as a known external input parameter. This sequence covers every scheduling time point in the prediction time domain, with each time point corresponding to a power exchange reference value. The length of the sequence is determined by the number of scheduling cycles contained in the prediction time domain. Neighboring transformers send the reference value sequence to this transformer via event-triggered communication or fixed-period communication. This transformer stores the reference value sequence and uses it as a known external input parameter for the current round of optimization. This parameter remains unchanged during the optimization process and is not updated with changes in the transformer's decision variables.

[0043] In this embodiment, the power exchange decision variable corresponding to the interconnection node in the current optimization variables of this transformer area is compared with the reference value at the corresponding time in the power exchange reference value sequence to obtain the power exchange deviation at each time in the prediction time domain. The positive and negative directions of this deviation sequence reflect the deviation direction of the power exchange plan of this transformer area relative to the reference value of neighboring transformer areas, and the absolute value of the deviation reflects the degree of deviation.

[0044] In this embodiment, a preset penalty coefficient is applied to the power exchange deviation, forming an interaction cost penalty term. The value of the penalty coefficient is dynamically set based on the electrical distance between the local transformer area and neighboring transformer areas or the capacity margin of the flexible interconnection device: the closer the electrical distance between adjacent transformer areas, the more significant the impact of power exchange on each other's voltage and power flow, so the penalty coefficient should be larger; the smaller the capacity margin of the flexible interconnection device, the easier it is for the deviation of the power exchange plan to lead to device overload, so the penalty coefficient should also be larger; conversely, the penalty coefficient can be appropriately reduced to give the local transformer area greater freedom in power exchange decisions.

[0045] In this embodiment, the interaction cost penalty term is directly added to the objective function, causing the local power exchange decision variable to approximate the reference value provided by the neighboring substation during optimization, unless the scheme would significantly increase other cost terms. This cost-driven rather than mandatory constraint approach achieves implicit coordination of power exchange plans between adjacent substations. Each substation does not need to simultaneously optimize the decision variables of its neighbors, nor does it need to construct a global optimization model, thus maintaining the independent solution characteristics of each substation's optimization problem.

[0046] This embodiment uses the power exchange reference value sequence sent by neighboring substations after the previous round of optimization as a known external input parameter, and guides the decision variables of the current substation to approximate it in the form of a penalty term. This enables implicit coordination between adjacent substations' power exchange plans without exchanging decision variables or building a global model, thus solving the problem of heavy communication burden and high solution complexity caused by frequent exchange of neighbor decision variables or construction of coupling constraints in existing distributed optimization. By dynamically setting the penalty coefficient according to the electrical distance or capacity margin of flexible interconnection devices, the coupling strength can be adaptively adjusted. The coordination between substations with tight electrical coupling or tight capacity margin is more stringent, while the decision freedom between substations with loose electrical coupling or sufficient capacity margin is higher. This maximizes the operational economy of each substation while ensuring the safety of the tie line.

[0047] In some embodiments, each control unit in a distribution area solves the local optimization problem in parallel and checks preset event triggering conditions after solving, including: Each control unit in each transformer area solves the local optimization problem in parallel to obtain the optimal power exchange sequence and optimal control sequence for its respective transformer area in the prediction time domain. After the solution is completed, calculate the expansion magnitude of the adaptive node uncertainty set boundary of the current optimization time of this station area at the start time of the prediction time relative to the corresponding boundary of the previous optimization time. The expansion range is compared with the first preset threshold. If the expansion range exceeds the first preset threshold, the triggering condition for the first type of event is determined to be met. Calculate the absolute value of the deviation between the first power exchange value at the interconnection node with the neighboring transformer area in the optimal power exchange sequence of this transformer area and the corresponding power exchange reference value provided by the neighboring transformer area in the previous round; The absolute value of the deviation is compared with the second preset threshold. If the absolute value of the deviation exceeds the second preset threshold, the triggering condition of the second type of event is determined to be met. The first type of event triggering condition and the second type of event triggering condition are combined by a logical OR relationship. When at least one condition is true, the event triggering condition is determined to be satisfied.

[0048] In this embodiment, after each transformer area control unit solves the local optimization problem in parallel, it obtains the optimal power exchange sequence and the optimal control sequence for that transformer area in the prediction time domain. The optimal power exchange sequence includes the planned power exchange values ​​at the interconnection nodes between the transformer area and neighboring transformer areas at each time in the prediction time domain, and the optimal control sequence includes the control command values ​​for the energy storage charging and discharging power and the flexible load adjustment at each time in the prediction time domain. Together, they constitute the complete operation plan of the transformer area at the current optimization time.

[0049] In this embodiment, after the solution is completed, the expansion magnitude of the adaptive node uncertainty set boundary of the current optimization time of the local transformer area relative to the corresponding boundary of the previous optimization time at the start time of the prediction time domain is calculated. This expansion magnitude reflects the degree of change in the local transformer area's perception of future uncertainties. The larger the expansion magnitude, the more significant the change in the local transformer area's perception of the prediction uncertainty of photovoltaic output or load demand, which may be caused by sudden changes in meteorological conditions or load mode switching. This change information needs to be transmitted to neighboring transformer areas to update the uncertainty constraints in their local optimization problems.

[0050] In this embodiment, the expansion magnitude is compared with a first preset threshold. If the expansion magnitude exceeds the first preset threshold, the triggering condition for the first type of event is determined to be met. The setting of the first preset threshold determines the sensitivity of this distribution area to changes in the perception of uncertainty. The smaller the threshold, the more frequent the communication and the higher the coordination accuracy, but the heavier the communication burden; the larger the threshold, the sparser the communication and the lighter the communication burden, but the coordination accuracy may decrease. This threshold can be set differently according to the intensity of source-load fluctuations in the area where the distribution area is located. For example, a smaller first preset threshold is used for distribution areas with intense source-load fluctuations, and a larger first preset threshold is used for distribution areas with stable source-load fluctuations.

[0051] In this embodiment, the absolute value of the deviation between the first power exchange value at the interconnection node with the neighboring substation in the optimal power exchange sequence of this substation and the corresponding power exchange reference value provided by the neighboring substation in the previous round is calculated. The first power exchange value corresponds to the power exchange plan at the first scheduling moment to be executed in the prediction time domain. The power exchange deviation at this moment will directly affect the power balance of the current control cycle. Therefore, it is used as the basis for judging the triggering condition of the second type of event, which can promptly capture power exchange discrepancies that need to be coordinated immediately.

[0052] In this embodiment, the absolute value of the deviation is compared with a second preset threshold. If the absolute value of the deviation exceeds the second preset threshold, the triggering condition for the second type of event is determined to be met. The setting of the second preset threshold determines the tolerance of this substation for disagreements in the power switching plan. The smaller the threshold, the more frequent the communication and the higher the coordination accuracy of the power switching plan; the larger the threshold, the sparser the communication, but the accumulation of disagreements may lead to power fluctuations or voltage exceeding limits on the tie line.

[0053] In this embodiment, the first type of event triggering condition and the second type of event triggering condition are combined using a logical OR relationship. When at least one condition is met, the event triggering condition is determined to be satisfied. This combination method ensures that communication can be triggered regardless of whether it is due to changes in uncertainty perception or disagreements in power exchange plans. This allows neighboring transformer areas to obtain the latest status of their local transformer area in a timely manner, thereby updating the coupling term parameters or uncertainty constraints in their local optimization problem in the next round of optimization, achieving timeliness and accuracy in distributed coordination.

[0054] This embodiment uses the expansion of the uncertainty set boundary of the adaptive node as the first type of event triggering condition, enabling distributed communication to respond to changes in the local area's perception of future uncertainties. This allows for automatic increases in communication frequency to enhance coordination accuracy when source-load prediction uncertainty increases significantly, and automatic decreases in communication frequency to reduce communication burden when uncertainty is stable. The second type of event triggering condition uses the absolute value of the deviation between the first power exchange value at the interconnected node in the local area's optimal power exchange sequence and the corresponding reference value provided by the neighboring area in the previous round. This enables distributed communication to respond to upcoming power exchange plan disagreements, triggering coordination in a timely manner before the accumulation of disagreements may lead to tie-line power fluctuations or voltage over-limits. By combining the two types of triggering conditions with a logical OR relationship, the communication triggering mechanism can simultaneously cover both scenarios of changes in uncertainty perception and power exchange plan disagreements. This solves the problems of communication redundancy during stable source-load periods and insufficient communication during severe source-load fluctuations in existing fixed-period communication mechanisms, achieving an adaptive balance between communication burden and coordination accuracy.

[0055] In some embodiments, when the event triggering condition is met, the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time are sent to the relevant neighboring transformer areas, including: When the triggering condition of the first type of event is met, the boundary parameters of the adaptive node uncertainty set corresponding to the current optimization time of this substation are packaged as the first type of communication data, and the sending object is determined to be the neighboring substation that is electrically closely coupled with the expansion of the boundary of the adaptive node uncertainty set. When the second type of event triggering condition is met, the power exchange sequence at the interconnection node of the neighboring substation in the predicted time domain obtained by solving the current optimization time of this substation is packaged as the second type of communication data, and the sending object is determined to be the specific neighboring substation corresponding to the absolute value of the deviation. When both the first type of event triggering condition and the second type of event triggering condition are met, the first type of communication data and the second type of communication data are merged and packaged, and the sending object is determined to be the union of the neighboring stations pointed to by the first type of communication data and the second type of communication data; The packaged communication data is sent to the designated recipient via the communication link. After the transmission is completed, the local control unit enters a waiting state, waiting to receive corresponding communication data from neighboring units for the next round of local optimization problem solving.

[0056] In this embodiment, when the first type of event triggering condition is met, the adaptive node uncertainty set boundary parameters corresponding to the current optimization time of this transformer area are packaged as the first type of communication data. These boundary parameters include the upper and lower boundary values ​​of the photovoltaic power output uncertainty set and the load demand uncertainty set at the current optimization time, as well as the corresponding time period identifier, enabling the receiver to accurately identify the time period range to which the boundary parameters apply. The sending target is determined based on neighboring transformer areas with close electrical coupling associated with the expansion of the adaptive node uncertainty set boundary, i.e., the neighboring transformer area with the highest degree of electrical coupling to this transformer area in the power flow direction corresponding to the expansion direction of the boundary, avoiding broadcasting to all neighbors and wasting communication resources.

[0057] In this embodiment, when the second type of event triggering condition is met, the power exchange sequence at the interconnection nodes of neighboring substations within the predicted time domain, obtained from the current optimization time of this substation, is packaged as the second type of communication data. This power exchange sequence contains the power exchange plan values ​​at the interconnection nodes of this substation and specific neighboring substations at each time in the predicted time domain, enabling the receiver to obtain the power exchange intention of this substation within a future period. The sending target is determined based on the specific neighboring substation corresponding to the absolute value of the deviation, that is, the neighboring substation that disagrees with the power exchange plan of this substation, rather than all neighboring substations, thereby achieving precise positioning of the communication target.

[0058] In this embodiment, when both the first and second type of event triggering conditions are met simultaneously, the first and second type of communication data are merged and packaged. This allows the receiver to simultaneously obtain the uncertainty set boundary parameters and power exchange sequence of its local substation, facilitating the receiver's simultaneous updating of the uncertainty constraints and power interaction coupling term parameters in its local optimization problem during the next round of optimization. The sending targets are determined by the union of the neighboring substations pointed to by the first and second type of communication data. That is, data is simultaneously sent to neighboring substations that need communication due to changes in uncertainty perception and neighboring substations that need communication due to disagreements in the power exchange plan, ensuring that all neighboring substations requiring updated information receive the data.

[0059] In this embodiment, the packaged communication data is sent to the designated recipient via a communication link. After transmission, the local control unit enters a waiting state, waiting to receive corresponding communication data from neighboring units. The duration of the waiting state is determined by a preset timeout. If communication data from a neighboring unit is received within the timeout period, it is used to resolve the local optimization problem in the next round. If no data is received within the timeout period, the reference value provided by the neighboring unit in the previous round is used as a substitute to ensure the continuity of the optimization process.

[0060] This embodiment uses the boundary parameters of the adaptive node uncertain set as the first type of communication data and the power exchange sequence as the second type of communication data. It determines the content and target of the data to be sent based on the type of triggering condition, ensuring a precise match between the content and target of the communication data and the triggering cause. This solves the problem of communication redundancy caused by sending all data in every communication in existing distributed control systems. By merging and packaging the data when both the first and second type of event triggering conditions are met and sending it to the union object, the receiver can obtain complete information at once, reducing the number of communication rounds. Furthermore, by entering a waiting state after transmission and setting a timeout substitution mechanism, the control unit in this area can continue to operate even with communication delays or packet loss, improving its applicability in rural areas with weak communication conditions.

[0061] In some embodiments, when the first type of event triggering condition is met, the adaptive node uncertainty set boundary parameters corresponding to the current optimization time of this transformer area are packaged as the first type of communication data, and the sending target is determined to be a neighboring transformer area with tight electrical coupling associated with the expansion of the adaptive node uncertainty set boundary, including: Identify the expansion direction and magnitude of the boundary of the adaptive node uncertainty set at the current optimization moment of this transformer area compared to the boundary corresponding to the previous optimization moment. The expansion direction points to the expansion of the upper or lower boundary of the photovoltaic power generation output uncertainty set, or the expansion of the upper or lower boundary of the load demand uncertainty set. Based on the expansion direction, a list of neighboring transformer substations with strong coupling relationship to the power flow direction corresponding to the expansion direction is queried from the preset neighboring transformer substation association table. The neighboring transformer substation association table stores the electrical coupling tightness of each transformer substation under different power flow directions in advance. The top-ranked number of neighboring stations in the list of neighboring stations based on their coupling degree are identified as the recipients of the first type of communication data. The local area identifier, the timestamp of the current optimization time, and the expanded boundary value and corresponding time period identifier in the boundary parameters of the adaptive node uncertainty set are encapsulated together into a first type of communication data packet and sent to the determined sending object through the communication link.

[0062] In this embodiment, the expansion direction and magnitude of the uncertainty set boundary of the adaptive node at the current optimization moment of this distribution area compared to the corresponding boundary at the previous optimization moment are identified. Specifically, the expansion direction points to the expansion of the upper or lower boundary of the uncertain set of photovoltaic power generation output, or the expansion of the upper or lower boundary of the uncertain set of load demand. An expansion of the upper boundary of the uncertain set of photovoltaic power generation output means that the predicted photovoltaic output of this distribution area may be higher than the previous expectation, while an expansion of the lower boundary means it may be lower; similarly, an expansion of the upper boundary of the uncertain set of load demand means that the load demand may be higher than the previous expectation, while an expansion of the lower boundary means it may be lower. The expansion direction determines the potential trend of power surplus or deficit and is the core basis for identifying the communication target.

[0063] In this embodiment, a list of neighboring transformer substations with strong coupling relationships to the power flow corresponding to the expansion direction is retrieved from a preset neighboring transformer substation association table. The neighboring transformer substation association table pre-stores a ranking of the electrical coupling tightness of each substation under different power flow directions. This ranking is calculated based on factors such as electrical distance between substations, tie-line impedance, and historical power exchange data. For example, when the expansion direction is the upper boundary of the uncertain set of photovoltaic power generation output, it means that the current substation may experience a power surplus in the future. In this case, communication should be prioritized with the neighboring substation that historically absorbed the most power when the current substation had a power surplus. Conversely, when the expansion direction is the upper boundary of the uncertain set of load demand, it means that the current substation may experience a power deficit in the future. In this case, communication should be prioritized with the neighboring substation that historically provided the most power when the current substation had a power deficit.

[0064] In this embodiment, a predetermined number of neighboring stations ranked by coupling degree in the neighboring station list are identified as the recipients of the first type of communication data. The predetermined number is set based on a balance between communication bandwidth and coordination accuracy requirements. A larger number results in wider communication coverage and higher coordination accuracy, but also a heavier communication burden; a smaller number results in a lighter communication burden, but may miss neighboring stations that need coordination. This number can be set differently based on the complexity of the network structure in the area where the station is located.

[0065] In this embodiment, the local area identifier, the timestamp of the current optimization time, and the expansion boundary value and corresponding time period identifier in the adaptive node uncertainty set boundary parameters are encapsulated together into a first type of communication data packet, and sent to the determined sending target through the communication link. The local area identifier is used by the receiver to identify the data source, the timestamp is used by the receiver to determine the timeliness of the data, the expansion boundary value contains quantitative information on the expansion direction and expansion magnitude, and the time period identifier indicates the time period range to which the boundary parameter applies.

[0066] This embodiment identifies the expansion direction of the uncertainty set boundary of the adaptive node and queries a list of neighboring substations with strong coupling to the corresponding power flow direction from a preset neighboring substation association table. This allows the determination of the communication object to accurately match the directional characteristics of the uncertainty perception change, thus solving the problem of communication redundancy caused by broadcasting to all neighbors in existing event-triggered communication. By determining the top preset number of neighboring substations with the highest coupling degree in the neighboring substation list as the sending object, the communication coverage can be adaptively adjusted according to the complexity of the network structure, minimizing the communication burden while ensuring coordination accuracy. By encapsulating the local substation identifier, timestamp, expansion boundary value, and time period identifier into a first type of communication data packet, the receiver can accurately identify the data source, judge the timeliness of the data, and obtain quantitative information on the uncertainty change. This allows for accurate updating of relevant parameters in its local optimization problem in the next round of optimization, achieving efficient and accurate distributed coordination communication based on the direction of uncertainty perception change.

[0067] In some embodiments, after receiving neighbor information, each control unit in each transformer area re-solves the local optimization problem, performs distributed iterations until convergence or a preset maximum number of iterations is reached, and executes the first element of the finally determined control sequence, including: Each control unit updates the power exchange reference trajectory or adaptive node-level uncertainty set parameters received from neighboring control units into the power interaction coupling terms or uncertainty constraints in the local model predictive control optimization problem, and re-solves the local optimization problem to obtain the local optimal power exchange sequence for this iteration. The local optimal power exchange sequence obtained in this iteration is compared with the local optimal power exchange sequence obtained in the previous iteration. The sum of the absolute differences between the two power values ​​at corresponding times in the prediction time domain is calculated as the iterative change of this station area. The iterative change is compared with a preset convergence threshold. When the iterative change of all transformer areas is less than the preset convergence threshold, the distributed iterative convergence is determined, and the control unit of each transformer area executes the first element of the local optimal control sequence obtained in this round of iteration. When the iterative change of any region is greater than or equal to the preset convergence threshold, the distributed iteration is determined to be non-converged, and it is checked whether the current iteration round has reached the preset maximum number of iterations. If the current iteration round has not reached the preset maximum number of iterations, each control unit will use the local optimal power exchange sequence obtained in this iteration as the new power exchange reference value, send it to the relevant neighboring control units through the communication link, and enter the next iteration round. If the current iteration round has reached the preset maximum number of iterations, then each control unit calculates the weighted average of the local optimal power exchange sequence in all iteration rounds as the consensus power exchange sequence, and executes the first element of the local optimal control sequence corresponding to the consensus power exchange sequence.

[0068] In this embodiment, each transformer control unit updates the power exchange reference trajectory or adaptive node-level uncertainty set parameters received from neighboring transformers into the power interaction coupling term or uncertainty constraint conditions in the local model predictive control optimization problem. If the received parameter is a power exchange reference trajectory, it is used as a new known parameter in the power interaction coupling term, replacing the reference value used in the previous round. If the received parameter is an adaptive node-level uncertainty set parameter, it is updated into the boundary conditions of the photovoltaic power generation output uncertainty constraint and the load demand uncertainty constraint. After the update, the local optimization problem is solved again to obtain the local optimal power exchange sequence for this iteration.

[0069] In this embodiment, the local optimal power exchange sequence obtained in the current iteration is compared with the local optimal power exchange sequence obtained in the previous iteration. The sum of the absolute differences in power values ​​at corresponding times in the prediction time domain is calculated as the iterative change of the local area. This iterative change quantifies the degree of change in the power exchange plan of the local area between two iterations. The larger the change, the greater the influence of neighbor information on the local area, and the more likely the distributed iteration has not yet reached a stable state.

[0070] In this embodiment, the iterative change is compared with a preset convergence threshold. When the iterative change of all transformer areas is less than the preset convergence threshold, distributed iterative convergence is determined, indicating that the power exchange plans of each transformer area have become consistent and no further iteration is needed. The control unit of each transformer area executes the first element of the local optimal control sequence obtained in this round of iteration. This first element corresponds to the control command at the current scheduling moment, including the energy storage charging and discharging power, the flexible load adjustment amount, and the power exchange value to the AC side.

[0071] In this embodiment, if the iterative change of any transformer area is greater than or equal to a preset convergence threshold, the distributed iteration is determined to have failed to converge. At this time, it is checked whether the current iteration round has reached the preset maximum number of iterations. If it has not reached the maximum number of iterations, each transformer area control unit uses the local optimal power exchange sequence obtained in this iteration round as the new power exchange reference value, sends it to the relevant neighboring transformer areas through the communication link, and enters the next iteration round to continue coordinating the power exchange plans of each transformer area.

[0072] In this embodiment, if the current iteration round has reached the preset maximum number of iterations, each control unit calculates the weighted average of the local optimal power exchange sequences across all iteration rounds as the consensus power exchange sequence. The weights of the weighted average can be set according to the convergence trend of each iteration round; for example, the weights closer to the latest iteration round are larger to reflect the reliability of the latest information. Each control unit executes the first element of its local optimal control sequence corresponding to the consensus power exchange sequence, thereby ensuring the continuity and smoothness of control when communication is limited or iteration diverges.

[0073] This embodiment updates the power interaction coupling terms or uncertainty constraints in the local optimization problem with the received neighbor information and resolves the problem, enabling each transformer area's power exchange plan to gradually respond to the state changes of neighboring transformer areas, thereby achieving iterative advancement of distributed coordination. By calculating the sum of the absolute differences between the current and previous rounds of the local optimal power exchange sequence at corresponding moments in the prediction time domain as the iterative change amount, and setting a convergence criterion that all transformer areas' iterative change amounts are less than a preset convergence threshold, the distributed iteration can terminate in a timely manner when the power exchange plans of each transformer area tend to be consistent, thus avoiding unnecessary communication and computational overhead. By calculating the weighted average of the local optimal power exchange sequences in all iteration rounds as a consensus value when the iteration has not converged and the preset maximum number of iterations has been reached, each transformer area can obtain a compromise coordination scheme under conditions of limited communication or iteration divergence. This solves the problem that existing distributed optimization methods directly interrupt control or use the results of the previous round when the iteration has not converged, leading to control abrupt changes, and improves the robustness and engineering applicability of the scheduling method under weak communication conditions in rural areas.

[0074] In some embodiments, the predicted data and actual measurement data of each control unit in a historical natural cycle are obtained, and the mean deviation and standard deviation of the prediction error are independently calculated for each control unit and each time period, including: The historical natural cycle is divided into several continuous time periods with a preset time granularity. For each photovoltaic power generation unit and various loads in each transformer area, the day-ahead predicted power data and real-time measured power data corresponding to each time period are extracted as raw data pairs. Calculate the difference between the predicted power data and the real-time measured power data for each raw data pair to obtain the historical prediction error value of the equipment for that period. All historical prediction error values ​​of the same transformer area, the same time period, and the same type of equipment are collected into a prediction error sample set for that transformer area, the same time period, and that type of equipment. The prediction error sample set is fitted with a normal distribution, and the mean deviation and standard deviation of the deviation are calculated. The mean deviation represents the degree of systematic deviation of the prediction of the equipment type in the same area and time period, and the standard deviation represents the degree of dispersion of the prediction of the equipment type in the same area and time period.

[0075] In this embodiment, the historical natural cycle is divided into several consecutive time periods with a preset time granularity. The preset time granularity is consistent with the length of the scheduling cycle, for example, fifteen minutes per time period, thereby ensuring that the statistical characteristics of the prediction error within each time period can reflect the unique source-load variation patterns of that time period. For each photovoltaic power generation unit and various loads in each distribution area, the day-ahead predicted power data and real-time measured power data corresponding to each time period are extracted as raw data pairs. The day-ahead predicted power data comes from the predicted values ​​in the day-ahead scheduling plan, and the real-time measured power data comes from the actual operating data collected by the on-site measurement equipment. The two constitute a set of basic data pairs used to calculate the prediction error.

[0076] In this embodiment, the difference between the predicted power data and the real-time measured power data for each raw data pair is calculated to obtain the historical prediction error value of the device during that period. The positive or negative direction of this difference reflects the direction of the prediction deviation; a positive difference indicates that the predicted value is higher than the actual value, and a negative difference indicates that the predicted value is lower than the actual value. The absolute value of the difference reflects the degree of prediction deviation. All historical prediction error values ​​of the same transformer area, the same time period, and the same type of device are collected into a prediction error sample set for that transformer area, the same time period, and that type of device. This sample set covers historical data for all the same time periods within the historical natural cycle, such as all photovoltaic prediction error data for the period from 10:00 AM to 10:15 AM within the past quarter.

[0077] In this embodiment, a normal distribution is fitted to the prediction error sample set, and the mean and standard deviation of the deviation are calculated. The mean deviation characterizes the degree of systematic deviation in the prediction of this equipment type for this substation area during this time period. For example, if the photovoltaic prediction value of a certain substation area is systematically higher during the midday period, the mean deviation is positive. The standard deviation characterizes the degree of dispersion in the prediction of this equipment type for this substation area during this time period. The larger the standard deviation, the wider the fluctuation range of the predicted value around the actual value, and the lower the prediction accuracy. By fitting a normal distribution, the statistical characteristics of historical prediction errors can be quantified into two parameters, providing a data basis for the dynamic calibration of the boundary of the subsequent adaptive node uncertainty set.

[0078] This embodiment divides the historical natural cycle into several continuous time periods with a preset time granularity. For each photovoltaic power generation unit and various loads within a distribution area, it extracts day-ahead predicted power data and real-time measured power data as raw data pairs. This allows the statistical analysis of prediction errors to be accurate to the granularity of each distribution area, each time period, and each type of equipment. This solves the problem in existing technologies where globally unified statistics lead to the averaging of prediction error characteristics across different distribution areas, time periods, and equipment, resulting in the loss of personalized information. Historical prediction error values ​​are obtained by calculating the difference between the raw data pairs. All historical prediction error values ​​for the same distribution area, time period, and type of equipment are aggregated into a prediction error sample set. This sample set has clear spatiotemporal and equipment dimension identifiers, providing a structured data foundation for subsequent refined statistics. By fitting the prediction error sample set to a normal distribution, the mean deviation and standard deviation deviation are obtained. This allows the statistical characteristics of historical prediction errors to be quantified into two parameters with clear physical meaning: the mean deviation reflects the direction and degree of systematic offset, and the standard deviation deviation reflects the degree of prediction dispersion. This provides accurate and interpretable input parameters for the dynamic calibration of the uncertainty set boundary of subsequent adaptive nodes.

[0079] Please see Figure 3 In a second aspect, this embodiment also provides a distributed model prediction and scheduling system 1 for optical-storage direct-flexible-distributed systems with source-load uncertainty, used in the method described in the first aspect. The system includes: The data acquisition module 11 is used to acquire the predicted data and actual measurement data of each control unit in the historical natural cycle, as well as the current state of charge of each energy storage unit in the distribution area. The adaptive uncertainty set construction module 12 is connected to the data acquisition module 11. It is used to independently calculate the mean and standard deviation of the prediction error for each transformer area and each time period based on the prediction data and the actual measurement data. Based on the mean deviation, standard deviation and current state of charge, it dynamically calibrates the adaptive node uncertainty set boundary of photovoltaic power generation output and load demand. The local optimization solution module 13 is set in each transformer control unit. It is used to establish and solve the local model predictive control optimization problem with the goal of minimizing the operating cost of the transformer area. The optimization problem embeds the power interaction coupling term between adjacent transformer areas and uses the adaptive node uncertainty set as a constraint condition. The event trigger judgment module 14 is connected to the local optimization solution module 13. It is used to check whether the preset event trigger conditions are met after the local optimization solution module 13 solves the problem. The event trigger conditions include the expansion of the boundary of the adaptive node uncertainty set exceeding the first preset threshold, or the deviation between the optimal power exchange plan of this area and the power exchange reference value provided by the neighboring area exceeding the second preset threshold. The communication interaction module 15 is connected to the event trigger judgment module 14, and is used to send the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters of the current optimization time to the relevant neighboring transformer areas when the event trigger condition is met. The distributed iterative convergence module 16 connects the communication interaction module 15 and the local optimization solution module 13. After the communication interaction module 15 receives the neighbor information, it triggers the local optimization solution module 13 to solve again and perform distributed iteration until convergence or the preset maximum number of iterations is reached, and outputs the first element of the final determined control sequence.

[0080] The system provided in this embodiment is a product embodiment corresponding to the method embodiment of the first aspect. The data acquisition module 11 corresponds to the operation of acquiring predicted data, measurement data, and state of charge in the method steps; the adaptive uncertainty set construction module 12 corresponds to the step of dynamically calibrating the boundary of the adaptive node uncertainty set based on the mean deviation, standard deviation of the deviation, and state of charge; the local optimization solution module 13 corresponds to the step of establishing and solving a local optimization problem with embedded power interaction coupling terms and adaptive node uncertainty sets as constraints; the event trigger judgment module 14 corresponds to the step of checking the two types of event trigger conditions after solving: the expansion magnitude of the uncertainty set boundary and the deviation of the power exchange plan; the communication interaction module 15 corresponds to the step of sending the power exchange reference trajectory and uncertainty set parameters to neighboring substations when the trigger conditions are met; and the distributed iterative convergence module 16 corresponds to the step of receiving neighbor information, resolving, and iterating until convergence or reaching the maximum number of iterations, and outputting the first element of the control sequence. The connection relationships and data flow between modules strictly follow the execution order of the method steps, and the module names and functional descriptions are consistent with the terminology in the method, ensuring that the system receives full support from the method.

[0081] Unlike existing technologies, the above technical solution acquires the predicted and actual measurement data of each control unit in each historical natural cycle, and calculates the mean and standard deviation of the prediction error independently for each distribution area and each time period. At the same time, it acquires the current state of charge of each energy storage unit in each distribution area. Based on the mean deviation, the standard deviation, and the current state of charge, it dynamically calibrates the adaptive node uncertainty set boundary for photovoltaic power generation output and load demand for each distribution area and each time period. This allows each distribution area to independently set the uncertainty set boundary according to its own historical prediction accuracy. Nodes with high prediction accuracy are no longer over-constrained, and nodes with low prediction accuracy can also obtain sufficient conservatism. At the same time, through the additional adjustment of conservatism coupled with the energy storage state of charge, it avoids the problem of no feasible solution due to excessive uncertainty set when the adjustment capability is insufficient. Based on this, each distribution area control unit solves the local model predictive control optimization problem in parallel, and checks preset event triggering conditions after solving. These event triggering conditions include the expansion of the adaptive node uncertainty set boundary exceeding a first preset threshold, or the deviation between the optimal power exchange plan of the local distribution area and the power exchange reference value provided by neighboring distribution areas exceeding a second preset threshold. Neighbor information exchange only occurs when the triggering conditions are met. Communication frequency is significantly reduced during periods of stable source and load, and automatically increased when uncertainty increases, achieving an adaptive balance between communication burden and coordination accuracy. The above technical solution, through the construction of a three-dimensional adaptive uncertainty set at the node level, time period level, and energy storage coupling, reduces the conservatism of uncertainty handling and improves operational economy. Through the event-triggered communication mechanism, the communication burden of distributed control is reduced by more than 70%, adapting to the weak communication conditions of rural distribution networks, while ensuring control coordination during periods of severe source and load fluctuations.

[0082] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0083] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A predictive scheduling method for a distributed optical-storage-direct-flexible model oriented towards source-load uncertainty, characterized in that, include: The system acquires the predicted and actual measurement data of each control unit in each historical natural cycle, and calculates the mean and standard deviation of the prediction error independently for each area and each time period. At the same time, it acquires the current state of charge of each energy storage unit in the area. Based on the mean deviation, the standard deviation of the deviation, and the current state of charge, the adaptive node uncertainty set boundary of photovoltaic power generation output and load demand is dynamically calibrated for each transformer area and each time period. Each control unit of a transformer area aims to minimize the operating cost of its own area by establishing a local model predictive control optimization problem, which embeds a power interaction coupling term between adjacent transformer areas and uses the adaptive node uncertainty set as a constraint condition. Each control unit in each transformer area solves the local optimization problem in parallel and checks the preset event triggering conditions after solving the problem. The event triggering conditions include the expansion of the boundary of the uncertainty set of the adaptive node exceeding a first preset threshold, or the deviation between the optimal power exchange plan of the transformer area and the power exchange reference value provided by the neighboring transformer area exceeding a second preset threshold. When the event triggering condition is met, the local power exchange reference trajectory in the prediction time domain starting from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time are sent to the relevant neighboring transformer areas. After receiving neighbor information, each control unit in each area re-solves the local optimization problem, performs distributed iterations until convergence or reaches the preset maximum number of iterations, and executes the first element of the finally determined control sequence.

2. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, Based on the mean deviation, the standard deviation of the deviation, and the current state of charge, the adaptive nodal uncertainty set boundary for photovoltaic power generation output and load demand is dynamically calibrated for each distribution area and each time period, including: The historical natural cycle is divided into several time periods. For each photovoltaic power generation unit and various loads in each distribution area, the mean deviation and standard deviation of the historical prediction error for each time period are calculated. The sum of the mean deviation and the standard deviation of the deviation by a preset multiple is used as the initial upper boundary of the uncertain set of photovoltaic power output of the area during the time period, and the difference between the mean deviation and the standard deviation of the deviation by a preset multiple is used as the corresponding initial lower boundary. Obtain the current state of charge of the energy storage unit, and determine the conservatism coefficient based on the preset range in which the current state of charge is located; When the current state of charge is higher than the first state of charge threshold, the conservatism addition coefficient corresponding to the initial upper boundary of the photovoltaic power output uncertainty set is increased to generate the first corrected upper boundary. When the current state of charge is lower than the second state of charge threshold, the conservatism coefficient corresponding to the initial upper boundary of the load demand uncertainty set is increased to generate a second corrected upper boundary. The first or second corrected upper boundary, together with the corresponding initial lower boundary, is established as the final adaptive node uncertainty set boundary.

3. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, Each control unit in a transformer area aims to minimize the operating cost of its area by establishing a local model predictive control optimization problem, including: The objective function is constructed by minimizing the sum of the cost of purchasing electricity from the AC side, the cost of energy storage charging and discharging losses, and the cost of comfort loss due to flexible load regulation in the local area within the predicted time domain. The energy storage power constraint, along with the dynamic state of charge constraint, the flexible load adjustment range constraint, the DC bus power balance constraint, and the photovoltaic power output uncertainty constraint and load demand uncertainty constraint generated by the boundary of the adaptive node uncertainty set, are collectively established as the set of constraints for the objective function. The interaction cost or voltage deviation penalty term when adjacent transformer areas exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term in the objective function or the set of constraints. The power interaction coupling term uses the power exchange reference value provided by the previous round of optimization of the neighboring transformer area as a known parameter, rather than simultaneously optimizing the decision variables of the neighboring transformer area.

4. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 3, characterized in that, The interaction cost or voltage deviation penalty term when adjacent transformer areas exchange power through flexible interconnection devices or AC tie lines is embedded as a power interaction coupling term into the objective function or the set of constraints, including: Obtain the power exchange reference value sequence between the neighboring transformer area and the interconnection node at each time in the prediction time domain after the previous round of optimization, as a known external input parameter; The power exchange decision variable corresponding to the interconnection node in the current optimization variables of this substation area is compared with the reference value at the corresponding time in the power exchange reference value sequence to obtain the power exchange deviation at each time in the prediction time domain. A preset penalty coefficient is applied to the power exchange deviation to form an interaction cost penalty term. The value of the penalty coefficient is dynamically set according to the electrical distance between the local transformer area and the neighboring transformer area or the capacity margin of the flexible interconnection device. By directly incorporating the interaction cost penalty term into the objective function, the local power exchange decision variable of this transformer area tends to approximate the reference value provided by the neighboring transformer area when optimizing, so as to achieve implicit coordination of power exchange plans between adjacent transformer areas, while maintaining the independent solution characteristics of the optimization problem of each transformer area.

5. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, Each control unit in each area solves the local optimization problem in parallel, and checks preset event triggering conditions after solving, including: Each control unit in each transformer area solves the local optimization problem in parallel to obtain the optimal power exchange sequence and the optimal control sequence for that transformer area in the prediction time domain. After the solution is completed, calculate the expansion magnitude of the adaptive node uncertainty set boundary at the current optimization time of this station area relative to the boundary at the start time of the prediction time domain. The expansion range is compared with a first preset threshold. If the expansion range exceeds the first preset threshold, the triggering condition for the first type of event is determined to be met. Calculate the absolute value of the deviation between the first power exchange value at the interconnection node with the neighboring substation in the optimal power exchange sequence of this substation and the corresponding power exchange reference value provided by the neighboring substation in the previous round; The absolute value of the deviation is compared with a second preset threshold. If the absolute value of the deviation exceeds the second preset threshold, the triggering condition for the second type of event is determined to be met. The first type of event triggering condition and the second type of event triggering condition are combined by a logical OR relationship. When at least one condition is true, it is determined that the event triggering condition is satisfied.

6. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, When the event triggering condition is met, the local power exchange reference trajectory in the prediction time domain starting from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time are sent to the relevant neighboring transformer areas, including: When the first type of event triggering condition is met, the adaptive node uncertainty set boundary parameters corresponding to the current optimization time of this substation are packaged as the first type of communication data, and the sending object is determined to be a neighboring substation that is electrically closely coupled with the expansion of the adaptive node uncertainty set boundary. When the second type of event triggering condition is met, the power exchange sequence at the interconnection node of the neighboring substation in the predicted time domain obtained by solving the current optimization time of this substation is packaged as the second type of communication data, and the sending object is determined to be the specific neighboring substation corresponding to the absolute value of the deviation. When the first type of event triggering condition and the second type of event triggering condition are met simultaneously, the first type of communication data and the second type of communication data are merged and packaged, and the sending object is determined to be the union of the neighboring stations pointed to by the first type of communication data and the second type of communication data; The packaged communication data is sent to the designated recipient via the communication link. After the transmission is completed, the local control unit enters a waiting state, waiting to receive corresponding communication data from neighboring units for the next round of local optimization problem solving.

7. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 6, characterized in that, When the first type of event triggering condition is met, the adaptive node uncertainty set boundary parameters corresponding to the current optimization time of this substation are packaged as the first type of communication data, and the sending target is determined to be a neighboring substation with close electrical coupling associated with the expansion of the adaptive node uncertainty set boundary, including: Identify the expansion direction and magnitude of the boundary of the adaptive node uncertainty set at the current optimization time of this transformer area compared to the boundary corresponding to the previous optimization time. The expansion direction points to the expansion of the upper or lower boundary of the photovoltaic power output uncertainty set, or the expansion of the upper or lower boundary of the load demand uncertainty set. According to the expansion direction, a list of neighboring transformer substations that have a strong coupling relationship with the power flow direction corresponding to the expansion direction is queried from a preset neighboring transformer substation association table. The neighboring transformer substation association table stores the electrical coupling tightness of each transformer substation under different power flow directions in advance. The top-ranked number of neighboring stations in the list of neighboring stations, based on their coupling degree, are identified as the recipients of the first type of communication data. The local area identifier, the timestamp of the current optimization time, and the expanded boundary value and corresponding time period identifier in the boundary parameters of the adaptive node uncertainty set are encapsulated together into a first type of communication data packet and sent to the determined sending object through the communication link.

8. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, After receiving neighbor information, each control unit in each transformer area re-solves the local optimization problem, performing distributed iterations until convergence or reaching a preset maximum number of iterations, and then executes the first element of the finally determined control sequence, including: Each control unit updates the power exchange reference trajectory or adaptive node-level uncertainty set parameters received from neighboring control units into the power interaction coupling terms or uncertainty constraints in the local model predictive control optimization problem, and re-solves the local optimization problem to obtain the local optimal power exchange sequence for this iteration. The local optimal power exchange sequence obtained in this iteration is compared with the local optimal power exchange sequence obtained in the previous iteration. The sum of the absolute differences between the two power values ​​at corresponding times in the prediction time domain is calculated as the iterative change of this station area. The iterative change is compared with a preset convergence threshold. When the iterative change of all transformer areas is less than the preset convergence threshold, the distributed iterative convergence is determined, and the control unit of each transformer area executes the first element of the local optimal control sequence obtained in this round of iteration. When the iterative change of any region is greater than or equal to the preset convergence threshold, it is determined that the distributed iteration has not converged, and it is checked whether the current iteration round has reached the preset maximum number of iterations. If the current iteration round does not reach the preset maximum number of iterations, each control unit will use the local optimal power exchange sequence obtained in this iteration as a new power exchange reference value, send it to the relevant neighboring control units through the communication link, and enter the next iteration round. If the current iteration round has reached the preset maximum number of iterations, then each control unit calculates the weighted average of the local optimal power exchange sequence in all iteration rounds as the consensus power exchange sequence, and executes the first element of the local optimal control sequence corresponding to the consensus power exchange sequence.

9. The optical-storage-direct-flexible distributed model prediction and scheduling method for source-load uncertainty as described in claim 1, characterized in that, Acquire the predicted and actual measurement data for each control unit within the historical natural cycle, and independently calculate the mean and standard deviation of the prediction error for each control unit and each time period, including: The historical natural cycle is divided into several continuous time periods with a preset time granularity. For each photovoltaic power generation unit and various loads in each transformer area, the day-ahead predicted power data and real-time measured power data corresponding to each time period are extracted as raw data pairs. Calculate the difference between the day-ahead predicted power data and the real-time measured power data in each raw data pair to obtain the historical prediction error value of the device for that period. All historical prediction error values ​​of the same transformer area, the same time period, and the same type of equipment are collected into a prediction error sample set for that transformer area, the same time period, and that type of equipment. The prediction error sample set is fitted with a normal distribution, and the mean deviation and standard deviation of the deviation are calculated. The mean deviation represents the degree of systematic deviation in the prediction of the equipment type in the substation area during the time period, and the standard deviation represents the degree of dispersion in the prediction of the equipment type in the substation area during the time period.

10. A distributed model prediction and scheduling system for optical-storage direct-drive flexible architecture accommodating source-load uncertainty, characterized in that, The system for performing the method according to any one of claims 1 to 9, the system comprising: The data acquisition module is used to acquire the predicted and actual measurement data of each control unit in the historical natural cycle, as well as the current state of charge of each energy storage unit in the distribution area. An adaptive uncertainty set construction module, connected to the data acquisition module, is used to independently calculate the mean and standard deviation of the prediction error for each transformer area and each time period based on the predicted data and the actual measurement data, and dynamically calibrate the adaptive nodal uncertainty set boundary of photovoltaic power generation output and load demand based on the mean deviation, the standard deviation, and the current state of charge. The local optimization solution module is set in the control unit of each transformer area. It is used to establish and solve the local model predictive control optimization problem with the goal of minimizing the operating cost of the transformer area. The optimization problem embeds the power interaction coupling term between adjacent transformer areas and uses the adaptive node uncertainty set as a constraint condition. The event trigger judgment module is connected to the local optimization solution module and is used to check whether the preset event trigger conditions are met after the local optimization solution module solves the problem. The event trigger conditions include the expansion of the boundary of the adaptive node uncertainty set exceeding a first preset threshold, or the deviation between the optimal power exchange plan of the local area and the power exchange reference value provided by the neighboring area exceeding a second preset threshold. The communication interaction module, connected to the event triggering judgment module, is used to send the local power exchange reference trajectory in the prediction time domain from the current optimization time and the adaptive node uncertainty set parameters at the current optimization time to the relevant neighboring transformer areas when the event triggering condition is met. The distributed iterative convergence module connects the communication interaction module and the local optimization solution module. After the communication interaction module receives neighbor information, it triggers the local optimization solution module to resolve the problem and perform distributed iteration until convergence or the preset maximum number of iterations is reached, and outputs the first element of the final determined control sequence.

Images