Cross-regional power load cooperative dispatching method and device, terminal and storage medium

Abstract

The application discloses a cross-regional power load cooperative scheduling method and device, a terminal and a storage medium, and relates to the field of power load scheduling. The method comprises the following steps: based on a cross-regional scheduling center and at least one power control unit, the power control unit reports schedulable capacity information and operation constraint quantitative data, the cross-regional scheduling center generates a cooperative scheduling instruction with the target of minimizing the global power consumption cost variation, the power control unit determines an executable adjustment amount and generates a verification result, and the cross-regional scheduling center corrects the cooperative scheduling instruction decision according to the verification result until a convergence condition is met to form an executable scheduling instruction and then issues the executable scheduling instruction. The application achieves the technical effects of reasonably performing cross-regional power load cooperative scheduling, reducing the global power consumption cost variation, and improving the power scheduling efficiency and rationality.

Description

Technical Field

[0001] This application relates to the field of power system operation and control technology, and in particular to a method, device, terminal and storage medium for cross-regional power load collaborative dispatching. Background Technology

[0002] With the rapid development of the energy internet and new power systems, cross-regional power collaborative dispatch has become a key technology for improving energy efficiency, ensuring grid security and stability, and reducing overall energy costs. Especially against the backdrop of high-proportion renewable energy integration, widening peak-valley load differences, and deepening marketization of electricity prices, both traditional centralized dispatching models and isolated local demand-side management face severe challenges.

[0003] Currently, different units under the same group, the same park operator, or the same virtual power plant aggregator have the following shortcomings: First, the existing cross-regional dispatch models are mostly rigid optimizations, treating network topology and security constraints as insurmountable boundaries. They cannot be dynamically adjusted according to electricity price differences and load coupling characteristics. When there are significant differences in electricity prices in time and space, the optimization often fails due to local constraint conflicts, thus failing to release the potential for economic mutual assistance.

[0004] Secondly, there is a serious disconnect between scheduling instructions and local execution. The centralized optimization model generates instructions based on predicted data, but it cannot perceive subtle changes in the real-time operating status of each region. This leads to security risks or execution deviations when instructions are executed locally, creating a "decision-execution" gap.

[0005] Third, the collaborative logic is simple, mostly a one-way collaboration of "global dominance - local obedience", lacking dynamic interaction between the global and local sides, which easily leads to scheduling failure problems of "global optimal but local infeasibility".

[0006] Therefore, there is an urgent need for a collaborative scheduling method that can deeply integrate global optimization and local autonomy, and intelligently handle uncertainties and conflicts, in order to solve the core pain points of "rigid model, disconnected execution, and lack of collaboration". Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a cross-regional power load collaborative scheduling method, device, terminal and storage medium that can achieve collaborative optimization of load resources and power resources across regions. Through the collaborative design of coupled load forecasting, dynamic constraint relaxation and other technologies, the potential of cross-regional power scheduling is maximized while ensuring safety.

[0008] Firstly, this application provides a cross-regional power load collaborative dispatching method, which adopts the following technical solution:

[0009] A cross-regional power load collaborative dispatching method, based on a cross-regional dispatching center and at least one power control unit, includes the following steps:

[0010] Each of the power control units reports dispatchable capacity information and quantitative data of operational constraints for the future scheduling period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints.

[0011] The cross-regional dispatch center aims to minimize the global electricity cost variation. Based on the dispatchable capacity information, electricity price information and the quantitative data of operational constraints of each region, it performs optimization calculations to generate collaborative dispatch instructions containing the expected power exchange amount.

[0012] Based on the received coordinated scheduling instructions, the power control unit determines the executable adjustment amount with the goal of maximizing the local resource input-output ratio, and generates a verification result that includes the deviation between the executable adjustment amount and the coordinated scheduling instructions.

[0013] The cross-regional dispatch center performs a correction decision on the collaborative dispatch instruction based on the verification result. The power control unit corrects the instruction based on the correction decision until the verification result fed back from the region meets the preset convergence condition. The cross-regional dispatch center then generates an executable dispatch instruction and issues it to each of the power control units.

[0014] By adopting the above technical solutions, the traditional one-way scheduling mode has been changed. By introducing local verification and feedback links, and using this feedback as the basis for dynamic adjustment of global instructions, the mode of shifting from one-way instruction issuance to two-way collaborative decision-making has been realized, which effectively improves the executability of scheduling instructions and the overall robustness of the system in complex power grid environments. Through the iterative correction mechanism, the system has the ability to learn from conflicts and achieve a globally feasible solution, realizing intelligent collaboration of cross-regional scheduling.

[0015] Preferably, each of the power control units reports dispatchable capacity information and quantitative data of operational constraints for future scheduling periods. The dispatchable capacity information is determined based on load forecast data and operational safety constraints, specifically including the following steps:

[0016] The cross-regional dispatch center provides each of the power control units with the corresponding load forecast values ​​for the future dispatch period. The load forecast values ​​are generated by coupling and analyzing the local data reported by each region through a preset time-series network.

[0017] The power control unit extracts and quantifies the operational constraint quantification data based on local operational safety constraints. The operational constraint quantification data includes the maximum power supply capacity of the region and the critical load guarantee power.

[0018] The power control unit calculates the dispatchable capacity value for each future scheduling period based on the load forecast value, the maximum power supply capacity of the area, and the critical load guarantee power. The dispatchable capacity value includes negative dispatchable capacity value and positive dispatchable capacity value.

[0019] The power control unit reports the dispatchable capacity value and the quantitative data of the operational constraints to the cross-regional dispatch center.

[0020] By adopting the above technical solution, a method was established in which the load forecast value after spatiotemporal coupling analysis is uniformly provided by the dispatch center as the local calculation benchmark. This ensures that the load data on which the reporting capacity of each region is based has global consistency and spatial correlation, and avoids the conflict of dispatchable capacity information between regions due to inconsistent forecast benchmarks from the source. This provides reliable and consistent data input for subsequent global collaborative optimization.

[0021] Preferably, the power control unit calculates the dispatchable capacity value for each future scheduling period based on the load forecast value, the maximum power supply capacity of the area, and the critical load guarantee power, specifically including the following steps:

[0022] Construct a local optimization model with the regional net interactive power adjustment amount in the future scheduling period as the decision variable;

[0023] The objective function of the local optimization model is to minimize the total local operating cost, which includes grid power purchase cost, equipment operation cost, and capacity constraint violation penalty.

[0024] The local optimization model is constrained by the operational safety constraints, which include load guarantee constraints with the critical load guarantee power as the lower limit threshold and power supply capacity flexibility constraints.

[0025] Solving the local optimization model yields the schedulable capacity value of the net interactive power adjustment in the region under the condition of satisfying the operational safety constraints.

[0026] By adopting the above technical solution, a specific algorithm for calculating dispatchable capacity is constructed by building a local optimization model with flexible constraint penalty terms. The physical security constraints of power supply capacity are transformed into penalty terms in the objective function, so that the calculated dispatchable capacity value is no longer a simple static margin, but a dynamic and quantitative technical indicator that embeds the local system's tolerance to exceeding the constraint limit, thereby improving the technical content and decision support capability of the reported information.

[0027] Preferably, the cross-regional dispatch center aims to minimize the variation in global electricity costs. Based on the dispatchable capacity information, electricity price information, and quantitative data of operational constraints in each region, it performs optimization calculations to generate collaborative dispatch instructions containing the desired power exchange amount. Specifically, this includes the following steps:

[0028] Any two regions are combined to form a potential region pool. Each ordered region pair in the potential region pool contains a source region and a target region. For each ordered region pair, a first direction and a second direction for the exchange are constructed. The collaborative potential value under the two directions is calculated respectively. The collaborative potential value integrates static matching ability and dynamic trend matching degree. For ordered region pairs that include the same two regions, the direction with the higher collaborative potential value is selected. The selected ordered region pairs are arranged in descending order according to the collaborative potential value. The ordered region pairs that meet the preset ranking ratio are selected as candidate region pairs. It is verified whether the candidate region pairs still meet the operation safety constraints after power exchange. The candidate region pairs that pass the verification are selected as valid region pairs.

[0029] For each effective region pair, a local optimization sub-model is constructed to solve for the optimal expected power exchange amount between the effective region pairs and the marginal cost signal corresponding to the optimal expected power exchange amount. The marginal cost signal is used as the local excitation component of the effective region pair in this exchange.

[0030] The optimal expected power exchange amount of all the effective region pairs is integrated to form a global expected power exchange instruction. The local excitation components of each region in all exchanges are accumulated to form a comprehensive excitation signal for each region. The expected power exchange instruction and the comprehensive excitation signal are combined to form the cooperative scheduling instruction.

[0031] By adopting the above technical solution, a candidate region pair generation method and a distributed optimization process based on collaborative potential value screening and ranking are proposed. By integrating static matching capability and dynamic load trend matching degree, objects with high collaborative value are efficiently identified from multiple region pairs, significantly reducing the solution scale of the global optimization problem. At the same time, by extracting marginal signals in the optimization process and synthesizing comprehensive signals, key technical state parameters are provided for the autonomous decision-making of local units.

[0032] Preferably, the construction method of the local optimization sub-model specifically includes the following steps:

[0033] The net change in grid electricity purchase cost is calculated based on the electricity price information reported by the source region and the target region.

[0034] The desired power exchange amount is used as the decision variable of the local optimization sub-model, and minimizing the change in the total electricity cost caused by power exchange is used as the objective function.

[0035] The change in total grid electricity costs is the sum of the net change in grid electricity purchase costs, the execution cost function of the source region, and the acceptance cost function of the target region.

[0036] The source region execution cost function is the additional cost calculated based on the device action cost parameters in the quantified operational constraint data reported by the source region and the impact of the expected power exchange on local operation.

[0037] The target area acceptance cost function is an additional cost calculated based on the equipment action cost parameters in the operational constraint quantification data reported by the target area, and the impact of the expected power exchange on local operation.

[0038] By adopting the above technical solution, the construction rules of the local optimization sub-model are refined, and the impact of power exchange on the local operation of the source and target regions is clearly included in the optimization objective. This makes the optimization process not only consider the power flow at the network level, but also deeply couple the additional impact of the exchange behavior on the local system operation status at both ends (such as the frequency of equipment operation and the adjustment pressure), so that the generated exchange commands are more in line with the actual operating conditions.

[0039] Preferably, the step of determining the executable adjustment amount with the objective of maximizing the local resource input-output ratio and generating a verification result containing the deviation between the executable adjustment amount and the cooperative scheduling instruction specifically includes the following steps:

[0040] Based on the maximum power supply capacity of the region, the critical load guarantee power, and the load forecast value in the operational constraint quantification data, calculate the real-time safe power adjustment range;

[0041] After the power control unit receives the corresponding coordinated dispatch instruction, it converts the comprehensive incentive signal issued in the coordinated dispatch instruction into an equivalent unit power adjustment signal for local autonomous adjustment decision based on the expected power exchange amount of the region.

[0042] Using the equivalent unit power adjustment signal as the decision weight coefficient, an execution decision model that maximizes the local resource input-output ratio is constructed. The optimal solution is obtained by solving the execution decision model within the real-time safe power adjustment range. The optimal solution is used as the executable adjustment amount, and it is determined whether there is a deviation between the executable adjustment amount and the expected power exchange amount.

[0043] If the deviation exists, check all the operational safety constraints. If any inequality of the operational safety constraint is equal when the executable adjustment is executed, it is determined that the corresponding operational safety constraint has reached the constraint edge. The constraint identifiers of the corresponding operational safety constraints are counted, and the constraint identifiers, the deviation, and the executable adjustment are summarized into a verification result and fed back to the cross-regional scheduling center.

[0044] By adopting the above technical solution, a standard process is defined for local units to make autonomous decisions and generate structured verification results within a strict safety range based on received instructions and signals. This process includes specific technical steps such as safety range calculation, signal conversion, constraint satisfaction judgment and identification. Its output verification results not only include execution deviations, but more importantly, they include the constraint identifications that caused the deviations, providing accurate technical input for precise diagnosis of the upper-level system.

[0045] Preferably, the step of making an instruction correction decision for the coordinated scheduling instruction specifically includes the following steps:

[0046] If any of the feedback verification results contain the deviation, the cross-regional dispatch center extracts the deviation and the associated constraint identifier from the verification result;

[0047] The regions where the deviation exists but the constraint identifier is not reported are marked as execution deviation regions;

[0048] If the number of occurrences of any constraint identifier exceeds a preset threshold, the constraint corresponding to the constraint identifier is determined to be a critical constraint, and the corresponding area is marked as a constraint conflict area.

[0049] For the execution deviation region, calculate the historical deviation rate of the execution deviation region to obtain a correction value, and correct all the electricity price information in the next iteration based on the correction value;

[0050] For the constraint conflict region, count the number of constraint conflict regions and calculate a damping coefficient, determine the effective region pair related to the key constraint, and add a damping cost term to the objective function of the corresponding local optimization sub-model;

[0051] Solve the modified local optimization sub-model to generate new executable scheduling instructions.

[0052] By adopting the above technical solution, an algorithm for automated root cause diagnosis and targeted model correction based on verification results is provided. This algorithm can distinguish between two different types of problems: execution bias and constraint conflict, and trigger two different model adjustment strategies for each. This diagnostic-based differentiated correction mechanism makes the system iteration process more targeted and can guide the system state to converge to a globally feasible solution more quickly.

[0053] Preferably, for each pair of ordered regions, the calculation process of the cooperative potential value in the first direction specifically includes the following steps:

[0054] The static matching capability value is obtained based on the minimum value among the negative schedulable capability values ​​and the positive schedulable capability values ​​of the two regions in the ordered region pair;

[0055] Obtain load forecast value sequences for the source region and the target region in multiple consecutive future scheduling periods, calculate the load change in adjacent future scheduling periods in each load forecast value sequence to obtain a load trend sequence, and perform decentralization on each load trend sequence;

[0056] Calculate the negative correlation cosine similarity between the two decentralized load trend sequences, obtain the dynamic trend matching degree based on the negative correlation cosine similarity, and then perform weighted fusion on the static matching ability value and the dynamic trend matching degree after normalization to obtain the collaborative potential value in the first direction.

[0057] By adopting the above technical solution, the specific calculation method of dynamic trend matching degree in the collaborative potential value is revealed. That is, the collaborative effectiveness is predicted by analyzing the negative correlation of load change trends between regions. The trend of time series forecast data is introduced into the early screening stage, which improves the technical ability to identify stable and efficient collaborative objects in a dynamic environment and enhances the foresight of the scheduling strategy.

[0058] Secondly, this application provides a cross-regional power load collaborative dispatching device, which adopts the following technical solution:

[0059] A cross-regional power load collaborative dispatching device includes a cross-regional dispatching center and at least one power control unit, and further includes the following modules:

[0060] The data sensing and reporting module is configured to report dispatchable capacity information and operational constraint quantification data for each of the power control units during the future scheduling period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints.

[0061] The centralized optimization and instruction generation module is configured to perform optimization calculations based on the dispatchable capacity information, electricity price information and the quantitative data of the operation constraints of each region, with the goal of minimizing the global electricity cost variation in the cross-regional dispatch center, and generate collaborative dispatch instructions containing the expected power exchange amount.

[0062] The local execution and verification feedback module is configured so that the power control unit determines the executable adjustment amount based on the received coordinated scheduling instruction with the goal of maximizing the local resource input-output ratio, and generates a verification result containing the deviation between the executable adjustment amount and the coordinated scheduling instruction;

[0063] The iterative correction and collaborative decision-making module is configured such that the cross-regional dispatch center executes correction decisions on the collaborative dispatch instructions based on the verification results, and the power control unit corrects the instructions based on the correction decisions until the verification results fed back from the region meet the preset convergence conditions, thereby forming an executable dispatch instruction and sending it to each of the power control units.

[0064] Thirdly, this application provides a smart terminal, which adopts the following technical solution:

[0065] A smart terminal includes a memory and a processor. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the cross-regional power load collaborative scheduling method as described above.

[0066] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0067] A computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the cross-regional power load collaborative scheduling method as described above.

[0068] In summary, this application includes at least the following beneficial effects:

[0069] (1) This application solves the problem of the difficulty in balancing global instructions and local security in traditional scheduling by real-time interaction of central instructions, local verification and feedback correction, and achieves the unity of instruction security and global feasibility;

[0070] (2) This application integrates the static capacity matching degree and the correlation of dynamic load trend to generate a collaborative potential value as a quantitative indicator. Based on this, the candidate objects with the highest collaborative value are efficiently identified and ranked from a large number of regional combinations. This effectively reduces the optimization search space, reduces the computational complexity, and locks in high-value targets for subsequent optimization, thereby improving the generation efficiency and quality of the overall scheduling strategy.

[0071] (3) The cross-regional dispatch center and the power control unit of this application interact multiple times to make the verification results meet the convergence conditions and form executable dispatch instructions, thereby realizing dynamic interaction between the global and local systems and avoiding the dispatch failure problem of global optimal but local infeasibility. Attached Figure Description

[0072] Figure 1 This is a flowchart of an embodiment of a cross-regional power load collaborative scheduling method;

[0073] Figure 2 This is an architectural diagram of the cross-regional power load collaborative dispatching device in an embodiment. Detailed Implementation

[0074] This application provides a method, apparatus, terminal, and storage medium for cross-regional power load collaborative scheduling. To make the objectives, technical solutions, and advantages of this application clearer, the implementation methods of this application will be further described in detail below.

[0075] The following describes in further detail an embodiment of a cross-regional power load collaborative dispatching method of this application, with reference to the accompanying drawings.

[0076] This embodiment presents a cross-regional power load collaborative dispatching method, the process of which is as follows: Figure 1 As shown, the specific steps include the following:

[0077] The method is based on a cross-regional dispatch center and at least one power control unit. The cross-regional dispatch center has global information and is responsible for formulating macro-level goals for cross-regional power transfer. The power control units are deployed in each region and are responsible for the fine-grained management of their respective regions. The method is specifically executed by a centralized decision engine deployed in the cross-regional dispatch center, and includes the following steps:

[0078] S1. Each power control unit reports its dispatchable capacity information and quantitative data on operational constraints for the future dispatch period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints, and specifically includes the following steps:

[0079] S11. The cross-regional dispatch center receives local historical data and real-time operation data reported by each power control unit. The cross-regional dispatch center uses a preset time-series network to perform coupled analysis on the aggregated local data from multiple regions to generate load forecast values ​​for each region during the future dispatch period.

[0080] This embodiment uses an interactive spatiotemporal coupling prediction network for prediction. The process of generating load prediction values ​​is as follows: the interactive spatiotemporal coupling prediction network is called. The input data of the interactive spatiotemporal coupling prediction network includes at least: hourly historical load data of each region over the past 72 hours, real-time load data at the current moment, hourly temperature sequences that are strictly aligned with the hourly historical load data over the past 72 hours, binarized production plan time series data that are strictly aligned with the hourly historical load data over the past 72 hours, and a preset inter-regional business association mapping table. The production plan time series data includes the production load allocation scheme for each time period.

[0081] The network consists of four parts: an input processing layer, a region feature extraction layer, a cross-region interactive fusion layer, and a prediction output layer.

[0082] A. First, the original local data is converted into a standardized tensor through the input processing layer. Then, the channels are grouped and spliced ​​according to the inter-regional business association mapping table to construct a unique input sample tensor for each region.

[0083] The stitching process involves reading local data from any region, including [historical load (72 steps), real-time load (1 step), temperature (72 steps), and production plan (72 steps)]. Based on the business association mapping table, the associated regions for the current region are located, and the [historical load (72 steps) and production plan (72 steps)] of the associated regions are read.

[0084] The local data of the current region and its associated regions are concatenated along the feature channel dimension to form the final input tensor of the current region, which has the dimensions [73, 1, 6], that is, 73 time steps, 1 spatial dimension, and 6 feature channels.

[0085] This step encodes the associated information in a fixed channel order, providing structured prior knowledge for subsequent networks.

[0086] B. The region feature extraction layer is composed of multiple parallel and independent dilated causal convolutional networks. Each network processes the final input tensor of a region and learns the temporal pattern of that region.

[0087] The region feature extraction layer uses four convolutional layers with dilation coefficients of [1, 2, 4, 8] to explicitly capture dependencies at different time scales, such as daily cycles (24 steps) and half-day cycles (12 steps), with a convolutional kernel size of 3×3.

[0088] The individual time steps representing the real-time load in the final input tensor are directly mapped and added to the features output by the last convolutional layer along the channel dimension, outputting a depth feature vector for each region. This ensures that the current state of the real-time load is used as a bias term to correct historical trends and prevent prediction starting point drift.

[0089] C. The cross-regional interactive fusion layer adopts a multi-head scaling dot product attention mechanism, using the depth feature vectors of all regions. As input, an interpretable influence weight matrix is ​​dynamically generated by calculating the correlation between query (Q), key (K), and value (V) vectors, and the deep feature vectors are weighted and fused accordingly to output the fused feature vector for each region.

[0090] Among them, each region Through three different linear layers , , Mapping to obtain the query vector key vector value vector .

[0091] Output fused feature vector Calculate as follows: ;

[0092] Among them, the weighting coefficient ;

[0093] This is the scaled relevance score. , It is a predefined scaling constant, usually taken as the dimension value of the key vector.

[0094] D. The prediction output layer is based on fused feature vectors. Conditional decoding is performed, and the decoder for each region is a two-layer feedforward network.

[0095] The first floor will The first layer maps the high-dimensional features to a high-dimensional space and activates them; the second layer is the output layer, which maps the high-dimensional features into a vector of length 24. The output layer uses linear activation, and the model learns through training how to stack these on top of each other. External disturbances learned from it.

[0096] The final output of this layer's output network is a sequence of hourly load forecasts for each region over the next 24 hours. This forecasting method is only one of the optional implementations.

[0097] S12. Based on local operational safety constraints, extract and quantify operational constraint data, including the maximum power supply capacity of the region, the critical load guarantee power, and the equipment operation loss cost.

[0098] In one specific implementation, the maximum power supply capacity of the area where the constraint quantization data is run is taken as 85% to 90% of the rated capacity of the corresponding equipment.

[0099] S13. The power control unit calculates the dispatchable capacity value for each future dispatch period based on the load forecast value, the maximum power supply capacity of the area, and the critical load guarantee power.

[0100] Specifically, the power control unit uses the obtained quantitative data of operating constraints as boundary conditions and aims to minimize the total local operating cost to solve a local optimization model. The limit values ​​representing the range of load increase and load decrease in the solution set of this model are used as the negative dispatchable capacity value and the positive dispatchable capacity value, respectively. The process includes the following steps:

[0101] S131. Construct a local optimization model with the regional net interactive power adjustment amount as the decision variable during the future scheduling period;

[0102] S132. The objective function of the local optimization model is to minimize the total local operating cost, which includes the grid purchase cost, equipment operation cost, and capacity constraint violation penalty.

[0103] S133. Constrain the local optimization model by operating safety constraints. The operating constraints are constructed based on the dispatchable capacity information and quantitative data of the operating constraints reported by the region. These constraints include load guarantee constraints with critical load guarantee power as the lower limit threshold, power supply capacity flexibility constraints, and dispatchable capacity constraints that do not exceed the reported dispatchable capacity value.

[0104] The load guarantee constraint is that the total predicted load of the region after power exchange in each region is not less than the critical load guarantee power; the power supply capacity flexible constraint will treat the constraint on the maximum power supply capacity of the region as a flexible constraint that can be violated at a dynamic cost. The power supply capacity flexible constraint condition is that the predicted load after power exchange in each region does not exceed the sum of the maximum power supply capacity of the corresponding region and the predefined non-negative relaxation variable.

[0105] Penalty for capacity constraint violation ;

[0106] in, This is the penalty coefficient (ranging from 1.2 to 1.5). For cross-regional electricity price differences, For power supply capacity constraints, there are non-negative relaxation variables.

[0107] The objective function is,

[0108] ;

[0109] in, For the cost of purchasing electricity from the power grid, For equipment operating costs, Penalties for violating capacity constraints.

[0110] The grid purchase cost includes a weighted power sum and a peak power penalty. The weighted power sum is calculated based on the difference between the load forecast and the regional net interactive power adjustment, determining the actual power required to be obtained from the upstream grid. This power value is then multiplied by a weighting coefficient reflecting the grid load level at different times to obtain the weighted power for that time period. Finally, the weighted power for all future dispatch periods is summed together. The peak power penalty requires determining the greater of the historically recorded maximum grid power withdrawal and the maximum grid power withdrawal calculated for the future dispatch cycle, as the peak power reference value for this month's billing. This peak reference value is then multiplied by a preset penalty coefficient used to suppress power peaks.

[0111] The cost of purchasing electricity from the grid is used to guide the optimized electricity consumption curve, which uses more power during periods of low grid load and minimizes peak power consumption, thereby reducing the overall impact on the grid system.

[0112] Equipment operation cost quantifies the cumulative impact of equipment state changes. It requires pre-setting a state change loss coefficient for each controllable device, reflecting the equivalent loss incurred per action or unit power adjustment. Then, iterate through all controlled devices and all scheduling periods, checking if the state of each device in the current period (e.g., on / off state or power setpoint) is the same as in the previous period. If the state has changed, calculate the absolute value of the change and multiply it by the corresponding loss coefficient for that device, adding it to the total loss. Finally, sum up the loss values ​​from all devices and all periods to obtain the final equipment operation cost.

[0113] Equipment operating costs are used to guide the optimization model to generate the smoothest and most coherent sequence of control commands, minimizing unnecessary equipment switching frequency and drastic power fluctuations, thereby reducing overall equipment operating losses and extending its service life.

[0114] S134. Solve the local optimization model to obtain the schedulable capacity value of the regional net interactive power adjustment under the condition of satisfying the operational safety constraints.

[0115] S14. The power control unit reports the dispatchable capacity value and the quantitative data of operating constraints to the cross-regional dispatch center.

[0116] S2. The cross-regional dispatch center, aiming to minimize the variation in global electricity costs, performs optimization calculations based on dispatchable capacity information, electricity price information, and quantitative data of operational constraints in each region, generating collaborative dispatch instructions that include the desired power exchange volume. Specifically, this includes the following steps:

[0117] S21. Combine any two regions to form a potential region pool. Each ordered region pair in the potential region pool contains a source region i and a target region j. For each ordered region pair, construct the first direction and the second direction in which the exchange occurs, and calculate the cooperative potential value under the two directions respectively.

[0118] In a specific implementation, assuming any two regions are A and B, the first direction is from A to B, where A is the source region and B is the target region. First, the synergistic potential value from A to B is calculated. The second direction is from B to A, where B is the source region and A is the target region. Then, the synergistic potential value from B to A is calculated.

[0119] The synergistic potential value combines static matching ability with dynamic trend matching degree.

[0120] The calculation process for the cooperative potential value in any direction for each ordered region pair includes the following steps:

[0121] The static matching capability value is obtained by taking the minimum of the negative and positive schedulable capability values ​​of the two regions in the ordered region pair. ;

[0122] This value represents the maximum power exchange from region i to region j during the current scheduling period t (determined by the minimum schedulable capacity), and is a hard capacity limit based on the current state.

[0123] Obtain the load forecast sequence for source region i and target region j over multiple consecutive future scheduling periods.

[0124] ,

[0125] ;

[0126] at this time, and It is a sequence of length N-1. Positive values ​​indicate increased load (increased demand), and negative values ​​indicate decreased load (decreased demand).

[0127] Decentralization is performed on each load trend sequence to generate a centralized trend sequence:

[0128] ;

[0129] ;

[0130] in, for The mean; for The mean.

[0131] Calculate the negative correlation cosine similarity between two decentralized load trend sequences.

[0132] Cosine similarity , which is between [-1, 1].

[0133] That is, cosine similarity is the ratio of the dot product of two centered trend sequences to the product of the moduli of the two centered trend sequences;

[0134] Based on the negative correlation cosine similarity, it is mapped to a dynamic trend matching degree between 0 and 1:

[0135] ;

[0136] When the cosine similarity is -1, the load trends of the source region and the target region are completely negatively correlated, indicating the highest trend matching degree. At this time, the dynamic trend matching degree is 1 (the highest score).

[0137] When the cosine similarity is +1, the load trends of the source region and the target region are completely positively correlated, that is, they increase or decrease together, and the trend matching degree is the lowest. At this time, the dynamic trend matching degree is 0 (the lowest score).

[0138] After normalizing the static matching capability value, it is weighted and fused with the dynamic trend matching degree to obtain the collaborative potential value in the current direction.

[0139] The sum of the weighting coefficients is 1. For example, if more emphasis is placed on immediate capabilities, the coefficient for static matching capability can be 0.7, and the coefficient for dynamic trend matching degree can be 0.3; if more emphasis is placed on future trends, the coefficient for static matching capability can be 0.3, and the coefficient for dynamic trend matching degree can be 0.7.

[0140] The calculation method for the synergistic potential value in the other direction is the same as above, with the only difference being that the source region and the target region are opposite.

[0141] For ordered pairs of regions that include the same two regions, directions with higher collaborative potential values ​​are selected.

[0142] If the collaborative potential value from A to B is greater than the collaborative potential value from B to A, then the ordered region pair (A, B) is retained; if the collaborative potential value from B to A is greater than the collaborative potential value from A to B, then the ordered region pair (B, A) is retained; if the two are equal or lower than a preset threshold, then the entire pair is discarded.

[0143] The selected ordered region pairs are sorted in descending order according to their collaborative potential value, and the ordered region pairs that meet the preset ranking ratio (such as the top 20%) are selected as candidate region pairs.

[0144] Verify whether the candidate region pairs still meet the operational safety constraints after power exchange, and take the candidate region pairs that pass the verification as valid region pairs.

[0145] S22. For each candidate region pair, based on the reported quantitative data of operational constraints and schedulable capability information, verify whether the operational safety constraints of each region can still be met after power exchange. Determine the candidate region pairs that pass the verification as valid region pairs and output a list of valid region pairs. Valid region pairs consist of a source region and a target region.

[0146] Operational safety constraints include,

[0147] The scheduling capacity constraints are as follows: the power that the source region can transmit to the outside does not exceed the reported positive scheduling capacity value; the power that the target region can receive from the outside does not exceed the reported negative scheduling capacity value; and the optimal expected power exchange amount does not exceed the positive scheduling capacity value and the negative scheduling capacity value and is not negative.

[0148] The critical load guarantee constraint is that the total regional load forecast after power exchange between regions shall not be lower than the critical load guarantee power.

[0149] The power supply capacity is subject to flexible constraints, meaning that after power exchange in each region, the predicted total load of the region does not exceed the sum of the region's maximum power supply capacity and predefined slack variables.

[0150] S23. The cross-regional dispatch center constructs a local optimization sub-model for each effective region pair (source region i, target region j);

[0151] With the goal of minimizing the global electricity cost variation during future scheduling periods, a local optimization sub-model is solved based on the electricity price information of the current effective regional pairs and the equipment action cost parameters in the quantitative data of operational constraints, to obtain the optimal expected power exchange between effective regional pairs.

[0152] The construction of the local optimization sub-model specifically includes the following steps:

[0153] S231. Calculate the net change in power grid purchase cost based on the electricity price information reported by the source region and the target region;

[0154] S232, Exchange desired power amount As the decision variable of the local optimization sub-model, minimizing the change in the total electricity cost of the entire network caused by power exchange is taken as the objective function;

[0155] Change in total grid electricity costs = Net change in grid electricity purchase costs + Source region execution cost function + Target region acceptance cost function;

[0156] That is, the expression for the objective function is,

[0157] ;

[0158] in and These are the real-time electricity prices for target region j and source region i, respectively, during the current scheduling period. This represents the net change in the cost of electricity purchased from the power grid. Execute the cost function for the source region. The cost function for accepting the target region. and It is a function constructed based on the quantitative data of the operational constraints reported by the source region i and the target region j, which characterizes the execution cost and constraint violation risk.

[0159] S233. The source region execution cost function is the additional cost calculated based on the equipment action cost parameters in the quantified operational constraint data reported by the source region and the impact of the expected power exchange on local operation.

[0160] The target area acceptance cost function is the additional cost calculated based on the equipment action cost parameters in the operational constraint quantification data reported by the target area, and the impact of the expected power exchange on local operation (such as the occupation of reserve capacity).

[0161] Specifically, and The function corresponding to the item Each component consists of three core components, corresponding to three types of constraints and risks, namely, the equipment movement cost component. Risk cost component of critical load protection and the cost component of the risk of exceeding power supply capacity limits. ;

[0162] ;

[0163] , , These are the weighting coefficients for each component. and It can be linked to system stress levels; for example, when the entire network's backup capacity is insufficient, the capacity penalty weight can be automatically increased. This makes the system more conservative.

[0164] for Extract equipment action cost parameters from the quantified operational constraint data reported by source region i. , It is used to simulate the wear and tear, aging, and equivalent costs of equipment caused by its operation;

[0165] for Obtain critical load backup power from reported data and load forecast , When performing the exchange After that, the total regional load If the load falls below the critical load guarantee, the function value will be greater than zero, and the penalty increases quadratically as the gap widens, strongly preventing this from happening.

[0166] for Obtain the maximum power supply capacity of the area from the reported data. , If the total load of the area after the exchange exceeds the power supply capacity, a high penalty is imposed, which further realizes the idea of ​​flexible constraints, allowing slight overruns, but at a high cost.

[0167] , The default penalty coefficient is greater than zero. The value of is negatively correlated with the predicted reserve capacity of the entire network in the current period; the lower the total reserve capacity of the entire network, the lower the value of . The higher the value, the better.

[0168] S24. Obtain the marginal cost signal corresponding to the optimal expected power exchange amount, and use the marginal cost signal as the local excitation component of the effective region in this exchange.

[0169] The marginal cost signal is obtained by using standard mathematical programming solvers such as the Interior Point Method or the Active Set Method to solve the local optimization sub-model. The solver outputs the optimal solution and the optimal Lagrange multiplier for each constraint at the same time.

[0170] Therefore, to obtain the optimal desired power exchange quantity At the same time, acquire and Corresponding to the schedulable capacity constraints The associated Lagrange multiplier, whose value represents the marginal tension caused by this exchange to the region, is used as the marginal cost signal of this exchange, and its functional form constitutes the local incentive component.

[0171] For example, in a specific implementation, suppose that solving the model for the region pair (A, B) yields the optimal solution. Corresponding constraints The optimal Lagrange multiplier is L, if the accepting capacity of region B is... An increase of 1 kWh can reduce the overall cost by L. Therefore, the center can use this as an incentive signal to send to region B to encourage it to tap its potential; at the same time, it can also send a related signal to region A to encourage it to output.

[0172] S25. Integrate the set of optimal expected power exchange quantities and their directions (i to j) of all effective region pairs to form a global expected power exchange command;

[0173] The local excitation components of each region across all exchanges are summed to form the composite excitation signal for each region; that is, assuming the composite excitation signal for region i is... = equals the sum of the local stimulus components of all exchanges with region i as the source region, minus the sum of the local stimulus components of all exchanges with region i as the destination region.

[0174] The desired power exchange command and the integrated excitation signal are combined to form the cooperative scheduling command. That is, assuming that the cooperative scheduling command contains a subset of commands for each region, the subset includes at least: all the desired power exchange quantities to be performed in region i. The summary information, and the comprehensive excitation signal assigned to it. .

[0175] The cumulative action means that the dispatch center is making overall allocations of incentive budgets. If a region participates in both power transmission and power reception at the same time, its incentives may be partially offset, reflecting refined dispatching.

[0176] S3. Based on the received coordinated dispatch instructions, the power control unit determines the executable adjustment amount with the goal of maximizing the local resource input-output ratio, and generates a verification result containing the deviation between the executable adjustment amount and the coordinated dispatch instructions. Specifically, this includes the following steps:

[0177] S31. Based on the regional maximum power supply capacity, critical load guarantee power, and load forecast value in the operational constraint quantification data, calculate the real-time safe power adjustment range, including the maximum power increase limit and the maximum power decrease limit.

[0178] The maximum power increase limit U is obtained by subtracting the load forecast value from the maximum power supply capacity of the area, which represents the remaining space for increasing the load.

[0179] The maximum reducible power limit D is obtained by subtracting the critical load guarantee power from the load forecast value, which is the maximum load that can be reduced.

[0180] S32. After receiving the corresponding coordinated dispatch instruction, the power control unit converts the comprehensive excitation signal issued in the coordinated dispatch instruction into an equivalent unit power adjustment signal for local autonomous adjustment decision-making, based on the expected power exchange volume of the region. ,Right now,

[0181] ;

[0182] in, To provide comprehensive incentive signals, This represents the desired power exchange.

[0183] S33. Using the equivalent unit power adjustment signal as the decision weight coefficient, construct an execution decision model that maximizes the local resource input-output ratio. Solve the execution decision model within the real-time safe power adjustment range to obtain the optimal solution, and use the obtained optimal solution as the executable adjustment amount.

[0184] The execution decision model is as follows: ;

[0185] The local decision variable is the actual power adjustment. ;

[0186] Net efficiency function Command response performance is the difference between command response performance and baseline cost. Command response performance measures the actual power adjustment of local actions. Contribution to responding to and achieving the goals of central instructions. , where a and b are the linear and quadratic cost coefficients of local actions, respectively.

[0187] Security penalty function Constructed as when A convex function whose value increases sharply when local operating security constraints are violated;

[0188] Construct a safety penalty function Where U is the maximum power increase limit based on real-time safety constraints, D is the maximum power decrease limit, and K is a preset maximum penalty coefficient, such as 10. 6 .

[0189] maximize , It is a continuously differentiable, strictly concave function. The equation can be solved by taking the first derivative and setting it to zero. Obtain the unique optimal solution ,

[0190] This equation can be solved quickly using numerical iteration methods (such as Newton's method), where x is the power adjustment amount that maximizes the local resource input-output ratio. Due to the existence of the penalty P(x), x will automatically fall within the safe interval [-D, U], thus finding the optimal solution. As an executable adjustment variable.

[0191] S34. Calculate the difference between the executable adjustment and the desired power exchange to determine if there is a deviation between them.

[0192] If there is no deviation, directly submit a verification report containing the executable adjustment amount and the expected power exchange amount.

[0193] If a deviation exists, check all operational safety constraints. If any operational safety constraint inequality holds true when executing the executable adjustment, then the corresponding operational safety constraint is considered to have reached the constraint boundary. This indicates that the constraint is the cause of the deviation in the executable adjustment. Therefore, if the solution to the executable adjustment is restricted by any operational safety constraint boundary, the identifier of that constraint must be reported.

[0194] The constraint identifiers corresponding to the operational safety constraints that reach the constraint edge are statistically analyzed, and the constraint identifiers, deviations, and executable adjustments are summarized into verification results and fed back to the cross-regional scheduling center.

[0195] S4. The cross-regional dispatch center makes corrective decisions on the coordinated dispatch instructions based on the verification results. The power control units make corrections based on the corrective decisions until the verification results fed back from the region meet the preset convergence conditions. The cross-regional dispatch center then generates executable dispatch instructions and issues them to each power control unit. The specific steps include the following:

[0196] S41. The cross-regional dispatch center receives and analyzes the verification results. If there is a deviation in any of the feedback verification results, the cross-regional dispatch center extracts the deviation in the verification results and the associated constraint identifiers.

[0197] S42. Mark the areas where deviations exist and constraint identifiers are not reported as execution deviation areas;

[0198] S43. Count all reported constraint identifiers. If the number of occurrences of any constraint identifier exceeds a preset threshold (e.g., 30%), the constraint corresponding to the constraint identifier is determined to be a critical constraint, and the corresponding area is marked as a constraint conflict area.

[0199] S44. For the execution deviation region, retrieve the historical verification results of the region for the most recent N scheduling cycles, calculate the average execution deviation rate of the execution deviation region, and use it as the historical deviation rate.

[0200] The correction value is obtained by multiplying the preset excitation intensity coefficient and the historical deviation rate.

[0201] Based on the correction value, all electricity price information in the next iteration is corrected. That is, in the next iteration, all sub-models containing valid region pairs of the region are found, and the electricity price information related to the region in their objective function is replaced with the difference between the current electricity price information and the correction value.

[0202] When the center reruns the optimization model, it lowers the value of the electricity price parameter for a certain region in this iteration. The purpose is to make the model more inclined to dispatch power to that region, thereby generating a new instruction that provides a stronger incentive to that region.

[0203] S45. For constraint conflict regions, count the number M of constraint conflict regions and calculate a damping coefficient to obtain the urgency of the constraint.

[0204] θ = β × M;

[0205] Where θ is the damping coefficient, β is the preset unit damping coefficient, and β is positively correlated with the frequency at which the constraint has been reported as a conflict constraint in the past several scheduling periods.

[0206] In the model, we find the set of effective regions that all power exchanges need to pass through the key constraint, and add a damping cost term proportional to the absolute value of the desired exchange power to the objective function of the local optimization sub-model for all effective regions in this set. .

[0207] The aim is to impose additional penalties between pairs of effective regions involved in the critical constraints, thereby reducing the attractiveness of power exchange between them.

[0208] S46. Solve the corrected local optimization sub-model, generate new executable scheduling instructions and issue them to each power control unit. Repeat steps S3 and S4. If, in two consecutive iterations, the sum of the absolute values ​​of the adjustment deviations in all verification results no longer decreases, or the sum of the absolute values ​​of the adjustment deviations is less than a preset threshold, terminate the iteration and issue the current instruction as the final executable scheduling instruction.

[0209] Based on the same inventive concept described above, this application also discloses a cross-regional power load collaborative dispatching device, the structure of which is as follows: Figure 2 As shown, the device includes the following modules:

[0210] A cross-regional power load collaborative dispatching device includes a cross-regional dispatching center and at least one power control unit, and further includes the following modules:

[0211] The data sensing and reporting module is configured to report dispatchable capacity information and quantitative data of operational constraints for each power control unit during future scheduling periods. The dispatchable capacity information is determined based on load forecast data and operational safety constraints.

[0212] The centralized optimization and instruction generation module is configured as a cross-regional dispatch center with the goal of minimizing the global electricity cost variation. Based on the dispatchable capacity information, electricity price information and quantitative data of operational constraints of each region, it performs optimization calculations and generates collaborative dispatch instructions containing the expected power exchange amount.

[0213] The local execution and verification feedback module is configured to enable the power control unit to determine the executable adjustment amount based on the received coordinated scheduling instructions, with the goal of maximizing the local resource input-output ratio, and generate verification results that include the executable adjustment amount and the deviation of the coordinated scheduling instructions.

[0214] The iterative correction and collaborative decision-making module is configured so that the cross-regional dispatch center can make correction decisions on the collaborative dispatch instructions based on the verification results, and the power control unit can make corrections based on the correction decisions until the verification results fed back from the region meet the preset convergence conditions, forming an executable dispatch instruction and sending it to each power control unit.

[0215] Based on the same inventive concept described above, this application also discloses a smart terminal, including a memory and a processor. The memory stores at least one instruction, at least one program, code set, or instruction set. The processor loads and executes the at least one instruction, at least one program, code set, or instruction set to implement the cross-regional power load collaborative scheduling method described above.

[0216] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the cross-regional power load collaborative scheduling method described above.

[0217] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0218] Those skilled in the art will understand that the step numbers of the above methods or processes are only used to distinguish different steps and do not constitute an absolute restriction on the execution order. Some steps may be executed simultaneously or in a different order than the numbers.

[0219] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for cross-regional coordinated power load dispatching, characterized in that, The method is based on a cross-regional dispatch center and at least one power control unit, and includes the following steps: Each of the power control units reports dispatchable capacity information and quantitative data of operational constraints for the future scheduling period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints. The cross-regional dispatch center aims to minimize the fluctuation of global electricity costs. Based on the dispatchable capacity information, electricity price information, and quantitative data of operational constraints in each region, it performs optimization calculations to generate collaborative dispatch instructions containing the desired power exchange amount. Specifically, this includes the following steps. Any two regions are combined to form a potential region pool. Each ordered region pair in the potential region pool contains a source region and a target region. For each ordered region pair, a first direction and a second direction for the exchange are constructed. The collaborative potential value under the two directions is calculated respectively. The collaborative potential value integrates static matching ability and dynamic trend matching degree. For ordered region pairs that include the same two regions, the direction with the higher collaborative potential value is selected. The selected ordered region pairs are arranged in descending order according to the collaborative potential value. The ordered region pairs that meet the preset ranking ratio are selected as candidate region pairs. It is verified whether the candidate region pairs still meet the operation safety constraints after power exchange. The candidate region pairs that pass the verification are selected as valid region pairs. For each effective region pair, a local optimization sub-model is constructed to solve for the optimal expected power exchange amount between the effective region pairs and the marginal cost signal corresponding to the optimal expected power exchange amount. The marginal cost signal is used as the local excitation component of the effective region pair in this exchange. The optimal expected power exchange amount of all the effective region pairs is integrated to form a global expected power exchange instruction. The local excitation components of each region in all exchanges are accumulated to form a comprehensive excitation signal for each region. The expected power exchange instruction and the comprehensive excitation signal are combined to form the cooperative scheduling instruction. Based on the received coordinated scheduling instructions, the power control unit determines the executable adjustment amount with the goal of maximizing the local resource input-output ratio, and generates a verification result that includes the deviation between the executable adjustment amount and the coordinated scheduling instructions. The cross-regional dispatch center performs a correction decision on the collaborative dispatch instruction based on the verification result. The power control unit corrects the instruction based on the correction decision until the verification result fed back from the region meets the preset convergence condition. The cross-regional dispatch center then generates an executable dispatch instruction and issues it to each of the power control units.

2. The cross-regional power load collaborative dispatching method according to claim 1, characterized in that, Each of the power control units reports dispatchable capacity information and quantitative data of operational constraints for the future scheduling period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints, and specifically includes the following steps: The cross-regional dispatch center provides each of the power control units with the corresponding load forecast values ​​for the future dispatch period. The load forecast values ​​are generated by coupling and analyzing the local data reported by each region through a preset time-series network. The power control unit extracts and quantifies the operational constraint quantification data based on local operational safety constraints. The operational constraint quantification data includes the maximum power supply capacity of the region and the critical load guarantee power. The power control unit calculates the dispatchable capacity value for each future scheduling period based on the load forecast value, the maximum power supply capacity of the area, and the critical load guarantee power. The dispatchable capacity value includes negative dispatchable capacity value and positive dispatchable capacity value. The power control unit reports the dispatchable capacity value and the quantitative data of the operational constraints to the cross-regional dispatch center.

3. The cross-regional power load coordinated dispatching method according to claim 2, characterized in that, The power control unit calculates the dispatchable capacity value for each future scheduling period based on the load forecast value, the maximum power supply capacity of the area, and the critical load guarantee power, specifically including the following steps: Construct a local optimization model with the regional net interactive power adjustment amount in the future scheduling period as the decision variable; The objective function of the local optimization model is to minimize the total local operating cost, which includes grid power purchase cost, equipment operation cost, and capacity constraint violation penalty. The local optimization model is constrained by the operational safety constraints, which include load guarantee constraints with the critical load guarantee power as the lower limit threshold and power supply capacity flexibility constraints. Solving the local optimization model yields the schedulable capacity value of the net interactive power adjustment in the region under the condition of satisfying the operational safety constraints.

4. The cross-regional power load coordinated dispatching method according to claim 2, characterized in that, The construction method of the local optimization sub-model specifically includes the following steps: The net change in grid electricity purchase cost is calculated based on the electricity price information reported by the source region and the target region. The desired power exchange amount is used as the decision variable of the local optimization sub-model, and minimizing the change in the total electricity cost caused by power exchange is used as the objective function. The total change in electricity costs across the entire network is the sum of the net change in the power grid purchase cost, the execution cost function of the source region, and the acceptance cost function of the target region. The source region execution cost function is an additional cost calculated based on the device action cost parameters in the operational constraint quantification data reported by the source region and the impact of the expected power exchange on local operation. The target area acceptance cost function is an additional cost calculated based on the equipment action cost parameters in the operational constraint quantification data reported by the target area, and the impact of the expected power exchange on local operation.

5. The cross-regional power load collaborative dispatching method according to claim 4, characterized in that, The step of determining the executable adjustment amount with the objective of maximizing the local resource input-output ratio and generating a verification result containing the deviation between the executable adjustment amount and the cooperative scheduling instruction specifically includes the following steps: Based on the maximum power supply capacity of the region, the critical load guarantee power, and the load forecast value in the operational constraint quantification data, calculate the real-time safe power adjustment range; After the power control unit receives the corresponding coordinated dispatch instruction, it converts the comprehensive incentive signal issued in the coordinated dispatch instruction into an equivalent unit power adjustment signal for local autonomous adjustment decision based on the expected power exchange amount of the region. Using the equivalent unit power adjustment signal as the decision weight coefficient, an execution decision model that maximizes the local resource input-output ratio is constructed. The optimal solution is obtained by solving the execution decision model within the real-time safe power adjustment range. The optimal solution is used as the executable adjustment amount, and it is determined whether there is a deviation between the executable adjustment amount and the expected power exchange amount. If the deviation exists, check all the operational safety constraints. If any inequality of the operational safety constraint is equal when the executable adjustment is executed, it is determined that the corresponding operational safety constraint has reached the constraint edge. The constraint identifiers of the corresponding operational safety constraints are counted, and the constraint identifiers, the deviation, and the executable adjustment are summarized into a verification result and fed back to the cross-regional scheduling center.

6. The cross-regional power load coordinated dispatching method according to claim 5, characterized in that, The step of making instruction correction decisions for the coordinated scheduling instructions specifically includes the following steps: If any of the feedback verification results contain the deviation, the cross-regional dispatch center extracts the deviation and the associated constraint identifier from the verification result; The regions where the deviation exists but the constraint identifier is not reported are marked as execution deviation regions; If the number of occurrences of any constraint identifier exceeds a preset threshold, the constraint corresponding to the constraint identifier is determined to be a critical constraint, and the corresponding area is marked as a constraint conflict area. For the execution deviation region, calculate the historical deviation rate of the execution deviation region to obtain a correction value, and correct all the electricity price information in the next iteration based on the correction value; For the constraint conflict region, count the number of constraint conflict regions and calculate a damping coefficient, determine the effective region pair related to the key constraint, and add a damping cost term to the objective function of the corresponding local optimization sub-model; Solve the modified local optimization sub-model to generate new executable scheduling instructions.

7. The cross-regional power load coordinated dispatching method according to claim 2, characterized in that, For each ordered region pair, the calculation process for the cooperative potential value in the first direction specifically includes the following steps: The static matching capability value is obtained based on the minimum value among the negative schedulable capability values ​​and the positive schedulable capability values ​​of the two regions in the ordered region pair; Obtain load forecast value sequences for the source region and the target region in multiple consecutive future scheduling periods, calculate the load change in adjacent future scheduling periods in each load forecast value sequence to obtain a load trend sequence, and perform decentralization on each load trend sequence; Calculate the negative correlation cosine similarity between the two decentralized load trend sequences, obtain the dynamic trend matching degree based on the negative correlation cosine similarity, and then perform weighted fusion on the static matching ability value and the dynamic trend matching degree after normalization to obtain the collaborative potential value in the first direction.

8. A cross-regional power load collaborative dispatching device, comprising a cross-regional dispatching center and at least one power control unit, characterized in that, Includes the following modules: The data sensing and reporting module is configured to report dispatchable capacity information and operational constraint quantification data for each of the power control units during the future scheduling period. The dispatchable capacity information is determined based on load forecast data and operational safety constraints. The centralized optimization and instruction generation module is configured by the cross-regional dispatch center to minimize the global electricity cost fluctuation. Based on the dispatchable capacity information, electricity price information, and operational constraint quantification data of each region, it performs optimization calculations to generate collaborative dispatch instructions containing the desired power exchange volume. Specifically, any two regions are combined to form a potential region pool. Each ordered region pair in the potential region pool includes a source region and a target region. For each ordered region pair, a first direction and a second direction for the exchange are constructed, and collaborative potential values ​​are calculated for both directions. These collaborative potential values ​​integrate static matching capability and dynamic trend matching degree. For ordered region pairs including the same two regions, the direction with the higher collaborative potential value is selected, and the selected ordered region pairs are sorted in descending order according to their collaborative potential values. The system selects ordered region pairs that meet a preset ranking ratio as candidate region pairs, verifies whether the operational safety constraints are still met after power exchange in the candidate region pairs, and designates the candidate region pairs that pass the verification as valid region pairs. For each valid region pair, a local optimization sub-model is constructed to solve for the optimal expected power exchange amount between the valid region pairs and the marginal cost signal corresponding to the optimal expected power exchange amount. The marginal cost signal is used as the local incentive component of the valid region pair in this exchange. The optimal expected power exchange amounts of all valid region pairs are integrated to form a global expected power exchange instruction. The local incentive components of each region in all exchanges are accumulated to form a comprehensive incentive signal for each region. The expected power exchange instruction and the comprehensive incentive signal together constitute the cooperative scheduling instruction. The local execution and verification feedback module is configured so that the power control unit determines the executable adjustment amount based on the received coordinated scheduling instruction with the goal of maximizing the local resource input-output ratio, and generates a verification result containing the deviation between the executable adjustment amount and the coordinated scheduling instruction; The iterative correction and collaborative decision-making module is configured such that the cross-regional dispatch center executes correction decisions on the collaborative dispatch instructions based on the verification results, and the power control unit corrects the instructions based on the correction decisions until the verification results fed back from the region meet the preset convergence conditions, thereby forming an executable dispatch instruction and sending it to each of the power control units.

9. A smart terminal, characterized in that, The method includes a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the cross-regional power load collaborative scheduling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by a processor to implement the cross-regional power load collaborative scheduling method as described in any one of claims 1 to 7.

Images