Unit combination determination method and device, computer device and storage medium

Abstract

This application relates to a method, apparatus, computer equipment, and storage medium for determining unit combination. By inputting predicted electricity load and grid topology data into the prediction model, initial unit parameters are generated as a hot start reference. This eliminates the need for traditional hot start pre-calculation simplification models, saving the computational power and time consumed by pre-solution. It also avoids the hundreds of invalid iterations caused by the lack of a reference reference during cold start. Furthermore, the iteration process dynamically generates regularized penalty terms based on the difference between the intermediate parameters of the previous round and the initial parameters. These penalty terms constrain the deviation of the iterative output from the reference, guiding global optimization to converge quickly. This reduces the number of iteration rounds and constraint verifications, thereby improving the solution efficiency of unit combination.

Description

Technical Field

[0001] This application relates to the field of power grid dispatching technology, and in particular to a method, apparatus, computer equipment, and storage medium for determining unit combinations. Background Technology

[0002] The generating unit is a generator set. The unit combination refers to the start-up and shutdown status (start-up or shutdown) of each unit at each moment during a future scheduling cycle, as well as the specific active power output of the start-up unit, i.e., the active power output. The unit combination needs to meet all safety constraints and achieve the lowest operating cost.

[0003] Traditional techniques typically employ a cold start approach to determine unit configurations. However, this approach lacks any pre-defined reference points and searches for feasible and optimal solutions through iterative iterations from scratch. Due to the lack of effective guidance, hundreds of iterations and constraint checks are usually required to obtain a satisfactory solution, resulting in extremely low efficiency. In large-scale power grid scenarios, this can easily exceed the computation time window specified in the dispatching procedures, preventing the timely generation of dispatching plans and impacting the normal execution of power grid dispatching. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, and storage medium for determining unit combinations to address the aforementioned technical problems, thereby improving the efficiency of solving unit combinations.

[0005] Firstly, this application provides a method for determining unit combination, including:

[0006] Obtain grid topology data of the power grid, as well as predicted electricity load for future scheduling cycles;

[0007] The predicted electricity load and the power grid topology data are input into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0008] In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period for each unit corresponding to the first iteration are preset values.

[0009] Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time in the future scheduling cycle, and obtain the target unit status of each unit.

[0010] Based on the target unit status and the target unit parameters of each unit, the target unit combination of each unit is obtained.

[0011] In one embodiment, in each iteration, with the optimization objective of minimizing the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration, and with load balancing and line transmission power limits as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit, including:

[0012] In the current iteration, with the optimization objective of minimizing the sum of grid operating costs and corresponding regularization penalty terms, and with load balancing and line transmission power limits as constraints, the unit parameters of each unit are solved to obtain the intermediate unit parameters of each unit in the current iteration.

[0013] The intermediate unit parameters are distributed to each unit so that each unit, constrained by its own operating parameter range, determines its corresponding unit combination set and the value score corresponding to each unit's unit combination set based on the intermediate unit parameters. The unit combination set corresponding to each unit includes the active power output and start / stop status at each time point within a future scheduling cycle. The value score is used to measure the degree to which the unit combination set needs optimization for the optimization objective.

[0014] If the iteration stopping condition is met based on the value score of each unit, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the target unit parameters.

[0015] If the value score of each unit determines that the iteration stopping condition is not met, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the intermediate unit parameters for the future scheduling cycle obtained by each unit in the previous iteration round, and the unit parameters are solved again.

[0016] In one embodiment, determining the start / stop status of each unit at each time point within a future scheduling cycle based on the target unit parameters of each unit, and obtaining the target unit status of each unit, includes:

[0017] For each unit at each moment, if the target unit parameter is greater than the unit's minimum active power output, the initial unit state of the unit is determined to be the operating state; if the target unit parameter is less than or equal to the unit's minimum active power output, the initial unit state of the unit is determined to be the shutdown state.

[0018] Determine the total power corresponding to the active power transmission power of each transmission line in each set of transmission lines in the power grid; wherein, the set of transmission lines includes the set of transmission lines used for the same transmission task;

[0019] Determine the difference between the total active power output of each unit in operation and the predicted electricity load; wherein, the total active power output of each unit in operation is determined based on the target unit parameters corresponding to each unit in operation.

[0020] Based on the total power and the difference, the initial unit state of each unit is corrected to obtain the target unit state.

[0021] In one embodiment, the step of correcting the initial unit state of each unit based on the total power and the difference to obtain the target unit state includes:

[0022] If the difference is greater than or equal to a first preset difference, and / or there exists at least one target line set, then the unit to be adjusted is selected from each unit; wherein, the target line set is the transmission line set whose total power is greater than a preset power.

[0023] The initial unit state of the unit to be adjusted is adjusted to obtain the target unit state of the unit to be adjusted.

[0024] In one embodiment, selecting the unit to be adjusted from the units includes:

[0025] For each unit in operation, if the target unit parameter of the unit is greater than the minimum active power output of the unit, and the difference between the target unit parameter and the minimum active power output of the unit is less than a second preset difference, then the unit is designated as a candidate unit.

[0026] For each unit that is in a shutdown state, if the distance between the unit and any node in any target line set is less than a preset distance, then the unit is regarded as a candidate unit.

[0027] Determine the sensitivity coefficients for DC power flow of each candidate generator set and each target line set; whereby the sensitivity coefficients characterize the degree of influence of the active power output adjustment of the candidate generator set on the transmission power of the target line set.

[0028] Candidate units with at least one sensitivity coefficient greater than a preset coefficient will be designated as the first target units.

[0029] Determine the power injection distribution factor of each candidate unit's node to the power imbalance node; where the power imbalance node is the node where the total active power output power does not match the predicted electricity load; the power injection distribution factor characterizes the ability of the candidate unit's active power output power adjustment to regulate the power imbalance.

[0030] Candidate units with power injection distribution factors greater than preset factors are designated as second target units;

[0031] The units existing in the first target unit and the second target unit are designated as units to be adjusted.

[0032] In one embodiment, adjusting the initial unit state of the unit to be adjusted to obtain the target unit state includes:

[0033] For each unit to be adjusted, the start-stop status of the unit at each time in the unit combination set corresponding to the unit to be adjusted is compared with the initial unit status of the unit at each time.

[0034] The initial unit state at the time of inconsistency corresponding to the inconsistent comparison results is adjusted to the start-stop state at the time of inconsistency in the unit combination set, thus obtaining the target unit state of the unit to be adjusted.

[0035] Secondly, this application also provides a unit combination determination device, comprising:

[0036] The acquisition module is used to acquire grid topology data of the power grid, as well as the predicted electricity load in future scheduling cycles;

[0037] The prediction module is used to input the predicted electricity load and the power grid topology data into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0038] The solution module, in each iteration, aims to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration, while constraining load balance and the upper limit of line transmission power. It solves for the unit parameters of each generating unit to obtain the target unit parameters. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period of each unit corresponding to the first iteration are preset values.

[0039] The first determining module is used to determine the start-up and shutdown status of each unit at each time in the future scheduling cycle based on the target unit parameters of each unit, and to obtain the target unit status of each unit.

[0040] The second determining module is used to obtain the target unit combination of each unit based on the target unit status and the target unit parameters of each unit.

[0041] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0042] Obtain grid topology data of the power grid, as well as predicted electricity load for future scheduling cycles;

[0043] The predicted electricity load and the power grid topology data are input into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0044] In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period for each unit corresponding to the first iteration are preset values.

[0045] Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time in the future scheduling cycle, and obtain the target unit status of each unit.

[0046] Based on the target unit status and the target unit parameters of each unit, the target unit combination of each unit is obtained.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0048] Obtain grid topology data of the power grid, as well as predicted electricity load for future scheduling cycles;

[0049] The predicted electricity load and the power grid topology data are input into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0050] In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period for each unit corresponding to the first iteration are preset values.

[0051] Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time in the future scheduling cycle, and obtain the target unit status of each unit.

[0052] Based on the target unit status and the target unit parameters of each unit, the target unit combination of each unit is obtained.

[0053] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0054] Obtain grid topology data of the power grid, as well as predicted electricity load for future scheduling cycles;

[0055] The predicted electricity load and the power grid topology data are input into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0056] In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period for each unit corresponding to the first iteration are preset values.

[0057] Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time in the future scheduling cycle, and obtain the target unit status of each unit.

[0058] Based on the target unit status and the target unit parameters of each unit, the target unit combination of each unit is obtained.

[0059] The aforementioned unit combination determination method, apparatus, computer equipment, and storage medium generate initial unit parameters as a hot-start benchmark by inputting predicted power load and grid topology data into the prediction model. This eliminates the need for traditional hot-start pre-calculation simplification models, saving the computational power and time consumed by pre-solution. It also avoids the hundreds of invalid iterations caused by the lack of a reference benchmark during cold starts. Furthermore, the iteration process dynamically generates regularized penalty terms based on the difference between the intermediate parameters of the previous round and the initial parameters. These penalty terms constrain the deviation of the iterative output from the benchmark, guiding global optimization to converge quickly. This reduces the number of iteration rounds and constraint verifications, thereby improving the solution efficiency of unit combination. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart illustrating a method for determining unit combination in one embodiment;

[0062] Figure 2 This is a schematic diagram of the process for solving the unit parameters of each unit in one embodiment;

[0063] Figure 3 This is a flowchart illustrating the process of determining the target unit status for each unit in one embodiment;

[0064] Figure 4 This is a schematic diagram of the process for selecting the unit to be adjusted from various units in one embodiment;

[0065] Figure 5 This is a structural block diagram of the unit combination determination device in one embodiment;

[0066] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] The unit combination determination method provided in this application can be applied to scenarios involving optimizing the unit combination in a power grid. This method can be executed by a server or a terminal.

[0069] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, and projection devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can include virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses.

[0070] In one exemplary embodiment, such as Figure 1 As shown, a method for determining unit combination is provided. Taking the application of this method to servers as an example, the method includes the following steps:

[0071] S101, obtain the grid topology data of the power grid, as well as the predicted power load in the future dispatch cycle.

[0072] Among them, the power grid topology data includes the spatial connection relationship of generator units, transmission lines, and substations within the power grid, as well as the inherent parameters of the equipment, such as transmission line impedance, rated transmission capacity of the line, substation capacity, unit number, and coordinates of the unit installation location. It fully represents the physical operating architecture of the power grid and is used for power flow calculation, transmission section identification, and unit distance measurement.

[0073] The future scheduling cycle is the time interval for scheduling planning, which can be 24-hour daytime scheduling or 48-hour rolling scheduling. The minimum time granularity is 1 hour (time slice). The cycle contains multiple consecutive moments, and each moment corresponds to a set of planned load and unit output values.

[0074] The predicted electricity load is the total electricity demand of the power grid at each moment within the dispatching cycle, based on historical electricity consumption data, meteorological data, industrial and commercial production plans, and residential electricity consumption patterns. The unit is megawatts (MW).

[0075] For example, the power dispatching master station system synchronizes two types of basic data from the load forecasting subsystem and the power grid model library through the intranet data interaction interface. The data synchronization time is fixed at 10 minutes before dispatching starts to avoid data lag affecting the solution accuracy.

[0076] For predicted electricity load, the load forecasting subsystem integrates the historical load time series of the past 90 days, the weather forecast for the next 24 hours, the maintenance plans of regional industrial and commercial enterprises, and the electricity consumption correction coefficient for holidays. It outputs the load values ​​within the scheduling cycle hourly, forming a one-dimensional load array with a length equal to the number of scheduling time slices. Simultaneously, it completes data preprocessing, uses linear interpolation to fill in the missing load values ​​collected, removes abnormal peak loads caused by lightning strikes and equipment failures, and stores all filtered data in memory cache for easy access by the neural network.

[0077] For the power grid topology data, the full power grid model file is read, and all generator unit numbers, rated output, minimum stable output, ramp limit, minimum continuous start-stop duration, and other inherent parameters of the units are extracted; all transmission line impedances, rated transmission capacity, substation node coordinates, and transmission section division rules are extracted; the association mapping relationship between units, lines, nodes, and sections is established, and a structured topology data table is generated.

[0078] Simultaneously, the topology data undergoes validity verification. If line maintenance or equipment failure is detected, leading to topology changes, the line operational status is automatically updated, and the transmission capacity of out-of-service lines is marked as 0, ensuring a complete match between the topology model and the current actual power grid architecture. After preprocessing, both types of data are aligned according to the time slice dimension to form a one-to-one corresponding input dataset, which is then fed into the neural network prediction model.

[0079] S102, input the predicted power load and grid topology data into the prediction model to obtain the initial unit parameters for each unit in the future scheduling cycle.

[0080] The initial unit parameters include the active power output of each unit at each time point in the future scheduling cycle.

[0081] The prediction model is trained offline based on the historical topology, load, and unit operation data of the power grid over the past 3-5 years. It adopts the decomposition and coordination principle to construct the network structure, with 5-8 hidden layers and 100-200 neurons in each layer. The input is load and topology data, and the forward propagation outputs the predicted output value of each unit for all time periods.

[0082] The initial unit parameters include the predicted active power output of each unit and each time slice within the scheduling cycle. They serve as the global output reference benchmark for the neural network output and are real continuous variables with no start / stop 0 or 1 discrete states.

[0083] Active power output is the active power transmitted by the generator set when it is running stably, measured in MW, and is subject to the upper and lower limits of the generator set's output and the limit of its ramp power.

[0084] The processed standardized load array and structured topology data table are concatenated into a model input tensor and fed into the prediction model. This is the offline-trained neural network prediction model, which then performs forward propagation calculations.

[0085] This neural network is trained based on the principle of decomposition and coordination. The training dataset uses complete operational data of the power grid over the past 3-5 years, containing tens of thousands of samples with different loads and topologies. During training, the unit combination problem is decomposed into a globally coupled optimization sub-problem and a single unit local optimization sub-problem. The network weights are updated in reverse using the lower limit of the total system operating cost as the loss function. The convergence criteria are that the change in loss is less than 0.001 over 100 consecutive iterations and the output prediction error on the validation set is less than 5%. The model network structure has 6 hidden layers, with 150 neurons in each layer. The ReLU function is used as the activation function to adapt to the multi-dimensional input features of a large-scale power grid.

[0086] During the forward propagation process, the input layer receives the load and topology features, performs weight matrix multiplication and activation function mapping layer by layer, and finally outputs a two-dimensional power output matrix. The rows of the matrix correspond to the unit number, the columns correspond to the scheduling cycle time slice, and each value in the matrix is ​​the active power output of the corresponding unit at the corresponding time. This two-dimensional power output matrix is ​​the initial unit parameters.

[0087] After the model output is completed, the initial unit parameters are boundary-checked, and abnormal predicted values ​​that exceed the upper and lower limits of unit output are removed. The upper and lower limits of the rated output of the unit are used for truncation correction to ensure that all initial unit parameters are within the physical feasible range of the unit. The corrected complete output matrix is ​​cached as the reference center for subsequent regularization penalty terms, realizing the generation of neural network hot start benchmark without pre-calculation process, saving the computing power consumption of pre-solving simplified models in traditional hot start mode.

[0088] S103 In each iteration round, with the optimization objective of minimizing the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration round, and with load balance and line transmission power limit as constraints, the unit parameters of each unit are solved to obtain the target unit parameters of each unit.

[0089] The regularization penalty term corresponding to the current iteration round is determined based on the difference between the intermediate unit parameters and the initial unit parameters obtained in the future scheduling period of the previous iteration round for each unit; the intermediate unit parameters obtained in the previous iteration round for each unit corresponding to the first iteration round are preset values.

[0090] The iteration round is a single global solution loop of the column generation decomposition coordination algorithm. Each round includes the entire process of global optimization solution, unit local path search, and convergence determination.

[0091] The power grid operating cost is the total economic cost, which includes the coal or energy consumption cost per unit of power generation, the fixed cost of unit startup, and the cost of unit downtime losses.

[0092] The regularization penalty term is an additional constraint term for the optimization objective. It takes the initial unit parameters (first coupling coordination vector) as the reference center and the weight is the regularization stability coefficient. It is used to penalize solutions where the unit output deviates too much from the initial unit parameters during the iteration process, constrain the iteration direction, and reduce blind search.

[0093] The regular stability coefficient is calculated from the difference between the intermediate unit parameters of the previous round and the initial unit parameters. The greater the difference, the larger the coefficient, and the stronger the penalty constraint; the smaller the difference, the smaller the coefficient, and the weaker the constraint to avoid getting trapped in local optima.

[0094] The intermediate unit parameters are the real-valued solutions of the power output distribution of each unit throughout the time period obtained after a single round of iterative global solution, which serve as the benchmark for comparison in the next round of iteration.

[0095] The degree of difference is the sum of the absolute values ​​of the output difference between the two sets of unit parameters at corresponding times, which quantifies the overall deviation of the two sets of output vectors.

[0096] The default value is that there are no historical intermediate parameters in the first iteration. The benchmark output vector is generated based on the typical daily operating conditions of the power grid in history, which serves as the benchmark for the first round of comparison.

[0097] The load balance constraint is that within each time slice, the sum of the active power output of all operating units equals the predicted power load at that moment, allowing for a small convergence error.

[0098] The upper limit of line transmission power is the maximum active power that a single transmission line or transmission section is allowed to transmit. Exceeding this limit will result in power flow exceeding the limit, which is a hard constraint on power grid safety.

[0099] The target unit parameters are the final coupled coordination vector after iterative convergence, which is the optimal continuous output allocation scheme that satisfies all global and unit constraints.

[0100] Each iteration process is divided into five sub-processes, with the specific steps as follows:

[0101] The first sub-process calculates the difference degree. If it is the first iteration, the intermediate unit parameters of the previous round adopt the preset historical typical output vector. If it is not the first iteration, the two-dimensional output matrix of the intermediate unit parameters output in the previous round is read and compared with the initial unit parameter matrix for each unit and for each time. The absolute value of the output difference at the corresponding position is calculated, and all absolute values ​​are accumulated to obtain the global difference degree value.

[0102] The second sub-process dynamically calculates the regularized stability coefficient based on the difference degree. The larger the difference degree value, the more serious the deviation between the current iteration output and the neural network prediction benchmark. The value of the regularized stability coefficient increases synchronously to strengthen the penalty constraint. When the difference degree approaches 0, the regularized stability coefficient takes the minimum value of 0.1, weakens the penalty and allows for small optimization of the solution, avoiding locking into the local optimum of the neural network prediction.

[0103] The third sub-process constructs a globally coupled optimization task with a regularization penalty term. The optimization objective is set to minimize (total grid operation cost + regularization penalty term). The regularization penalty term is calculated as the difference between the current iteration temporary output matrix and the initial unit parameters × the regularization stability coefficient. The constraints are fixed and include all-time load balance constraints and upper limit constraints on the transmission power of all transmission lines. No additional hard constraints are added. The iteration is guided to approach the initial unit parameters only through the optimization objective.

[0104] The fourth sub-process, the global coupled solution module, uses the initial unit parameters as the initial iteration point of the dual variables to iteratively solve the optimization task, continuously adjusting the output values ​​of each unit. After each adjustment, the load balance and line power flow constraints are checked. If the constraints are not met, the unit output is corrected in the direction of cost reduction until the feasible intermediate unit parameter matrix is ​​obtained in the current round.

[0105] The fifth sub-process distributes intermediate unit parameters to the control unit of each generator unit. Each unit generates a complete state diagram using its own output upper and lower limits, ramp limit, and minimum continuous start-stop duration as the search domain. Non-dominated state nodes are filtered to generate candidate start-stop state chains. After correcting illegal start-stop periods, the unit's start-stop optimization trajectory is obtained, and the local optimization potential value of each unit is calculated simultaneously. All unit local optimization potential values ​​are summarized, and a convergence judgment is performed: if all unit potential values ​​are ≥0, it means each unit has reached a local optimum, the iteration stops, and the current intermediate unit parameters are directly assigned to the target unit parameters; if any unit has a negative potential value, it means there is still room for optimization, the current intermediate unit parameters are updated to the previous round's baseline parameters, a new complete iteration loop is started, and the difference calculation is repeated until the non-negative convergence condition is met, at which point the target unit parameters are output.

[0106] This step uses dynamic regularization penalty to guide the neural network to a warm start, eliminating the shortcomings of traditional cold start with blind iteration without a benchmark and the time-consuming pre-computation of traditional warm start. Each iteration relies on the neural network output benchmark to narrow the search range, reducing the number of iterations and invalid constraint verification calculations.

[0107] S104. Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time point in the future scheduling cycle, and obtain the target unit status of each unit.

[0108] The start-stop status is the unit's operating identifier. The operating status is recorded as 1, and the shutdown status is recorded as 0. It is subject to the minimum continuous operating time and the minimum continuous shutdown time.

[0109] The target unit state is the unit start-stop integer vector obtained by discretization of continuous output, which is the second discrete unit state vector after the feasible integer solution is corrected.

[0110] After iteratively converging the target unit parameters, continuous output discretization is performed to generate initial unit start-up and shutdown vectors. Then, the grid power flow exceeding limits and power imbalance issues are corrected to finally obtain the compliant target unit status.

[0111] Step 1 Discrete Decision: Traverse the active power output values ​​of each unit and each time slice in the target unit parameter matrix, and set the discrete decision threshold to the minimum stable output of the unit; if the continuous output of the unit at a certain moment is greater than the minimum stable output, mark the unit as running (1); if the output is less than or equal to the minimum stable output, mark it as shut down (0), and generate the first discrete unit state vector in batches.

[0112] The second step is power flow over-limit identification: the built-in power flow calculation function of the power grid topology data is called, and the total transmission power is calculated for each transmission section based on the unit output distribution corresponding to the state vector of the first discrete unit. The power flow of the section is compared with the rated transmission capacity. The section with the power flow exceeding the upper limit is marked as the target transmission section, and the section location and the over-limit power value are recorded.

[0113] The third step is to calculate the power imbalance: filter all operating units within the first discrete vector, summarize the output of all operating units at this moment to obtain the total available output, subtract the predicted power load at the corresponding moment from the total available output, and the difference is the power imbalance. A positive number represents a power surplus and a negative number represents a power deficit.

[0114] The fourth step is to determine the state correction: set the power imbalance convergence threshold to 2% of the load forecast value, and at the same time determine whether the target transmission section set is empty; if the absolute value of the imbalance is greater than or equal to the convergence threshold, or if any target transmission section exceeds the power flow limit, then the unit start-up and shutdown state correction is performed; if the imbalance meets the standard and there is no power flow limit exceeding the limit, then the state vector of the first discrete unit is directly used as the state of the target unit.

[0115] During the correction process, adjustable candidate units are screened, and units to be adjusted are selected by combining DC power flow sensitivity and power injection distribution factor. Multiple state reversal schemes are generated, and the minimum continuous start-stop constraints and system ramp-up compensation capabilities of each scheme are verified. The optimal reversal scheme is selected through multi-level sorting, and the discrete vector is updated to obtain the target unit state integer matrix that fully satisfies the unit physical constraints, grid power flow constraints, and power balance constraints.

[0116] This step transforms the continuously optimized solution into an engineering-usable integer start-stop scheme, while taking into account grid safety and the physical limitations of unit operation, thus avoiding problems such as power flow exceeding limits, power imbalance, and frequent start-stop violations of units after discretization.

[0117] S105: Based on the target unit status and target unit parameters of each unit, obtain the target unit combination for each unit.

[0118] The target unit combination is the final output dispatch scheme, which includes the start-stop status sequence of each unit at all times and the optimal active power output plan sequence of each operating unit, and can be directly issued to the power plant for execution.

[0119] The first part generates the start-stop sequence: The corrected two-dimensional integer matrix of the target unit status is directly used as the start-stop status sequence. Each row of the matrix corresponds to one unit and each column corresponds to a scheduling time slice. The matrix element 0 or 1 clearly indicates the start-stop status of the unit at each moment, fully covering the entire scheduling cycle. All continuous start-stop periods meet the constraints of minimum continuous operation and minimum continuous shutdown duration.

[0120] The second part generates the power output plan sequence: traverse the start-stop state sequence and set the power output of all units marked as shutdown state (0) to zero throughout the time period; for units marked as operating state (1), take minimizing the system operating cost as the sub-objective, carry out secondary optimization of economic scheduling based on the target unit parameters, and constrain the upper and lower limits of unit output, ramp power limit, load balance throughout the time period, and upper limit of power flow of transmission section, and solve to obtain the optimal active power output value of each operating unit at each time.

[0121] The secondary optimization of economic dispatch only makes minor adjustments to continuous output without changing the established start-up and shutdown status of the units, ensuring complete power balance and no power flow exceeding the limit at all transmission sections.

[0122] Finally, the start-stop status sequence and the optimal output plan sequence are encapsulated in a structured manner, and unit numbers, timestamps, and constraint verification identifiers are added in a unified way to form a complete target unit combination scheduling scheme. This scheme can be exported to the power grid dispatch automation platform and the power plant remote monitoring system for direct use in actual day-ahead or real-time scheduling execution.

[0123] This step integrates discrete start-stop states with economically optimal output to output a feasible engineering scheduling scheme, balancing the economic efficiency of the solution with the safety of power grid operation.

[0124] For example, a scenario is selected where the scheduling calculation starts at 10:00 on June 20, 2025, with a scheduling cycle of 24 hours (from 10:00 on the same day to 10:00 on the next day). The power grid includes 10 generator sets (numbered 1-10), 20 transmission lines, and 5 substations. The preset time slice granularity is 1 hour, with a total of 24 time points. The implementation process of the above method is introduced.

[0125] Data acquisition: Load forecast data output 24-hour hourly load: 10:00 load 12000MW, 11:00 load 12500MW... the next day 9:00 load 11500MW; Power grid topology data input parameters of 10 generating units (rated output 1500MW, minimum stable output 450MW, ramp limit 150MW, minimum continuous start-stop 4 hours), rated transmission capacity of 20 lines, and information on the division of 3 transmission sections. All lines are in normal operation without maintenance.

[0126] Neural network prediction input: The 24-hour load array and the topology structured data of 10 units are fed into a 6-layer hidden layer neural network. The forward propagation outputs the initial unit parameters (24 rows × 10 columns of power matrix). The baseline output at 10:00 is: Unit 1 = 1200MW, Unit 2 = 1100MW ... Unit 10 = 1000MW. The corresponding output values ​​are generated synchronously at other times.

[0127] Iterative solution for target unit parameters: This is the second iteration. The total difference between the intermediate unit parameters and the initial unit parameters from the previous iteration is 800MW, and the regularization stability coefficient is set to 0.5. The optimization objective is to minimize (system power generation + start-up and shutdown costs + 0.5 × output deviation difference), with constraints that the total hourly output matches the load and the line power flow does not exceed limits. The intermediate unit parameters are obtained through global solution and distributed to 10 units to complete local path search. The trajectories of units 9 and 10 are deemed invalid due to violating the minimum continuous operation time of 4 hours. Units 1-8 output valid start-up and shutdown optimized trajectories and local potential values. All unit potential values ​​are ≥0, the iteration converges, and the intermediate unit parameters are determined as the target unit parameters.

[0128] Target unit status generation: Based on discrete determination of target unit parameters, units 2 and 4 with output ≤450MW are marked as shut down, and the remaining units are marked as running; power flow calculation identifies section 2 with a power flow of 1600MW, exceeding the rated 1500MW, indicating a power flow exceeding the limit; power imbalance is 100MW, with a convergence threshold of 148MW, indicating the imbalance meets the standard but there is a section exceeding the limit. Unit 2 is selected as a candidate for shutdown and unit 3 as a candidate for operation. Based on sensitivity and distribution factor, unit 2 is determined as the unit to be adjusted. A state reversal scheme is generated. After verifying the unit start-up and shutdown duration and system ramp-up compensation capability, the 10:00 state of unit 2 is reversed from shutdown to operation, and the discrete vector is updated to obtain the target unit state matrix.

[0129] Output target unit combination: Encapsulate the target unit status matrix as the start-up and shutdown sequence; the output of the shut-down unit is set to zero, the output of the operating unit is adjusted by economic scheduling, the total output is precisely matched to 12000MW at 10:00, the power flow of section 2 drops to 1450MW without exceeding the limit, forming a complete 24-hour 10-unit scheduling scheme, which is the target unit combination.

[0130] In the above embodiments, the initial unit parameters are generated by inputting the predicted power load and grid topology data into the prediction model as a hot start benchmark. This eliminates the need for the traditional hot start pre-calculation simplified model, saving the computing power and time consumed by pre-solution. At the same time, it avoids the hundreds of invalid iterations caused by the lack of a reference benchmark during cold start. Furthermore, the iteration process dynamically generates regularized penalty terms based on the difference between the intermediate parameters of the previous round and the initial parameters. These penalty terms constrain the deviation of the iterative output from the benchmark, guiding the global optimization to converge quickly. This reduces the number of iteration rounds and constraint verifications, thereby improving the solution efficiency of unit combination.

[0131] In some alternative implementations, see [link to relevant documentation]. Figure 2 , Figure 2 A flowchart illustrating the process of solving for unit parameters of each generating unit is provided, which includes the following steps:

[0132] S201, in the current iteration round, with the optimization objective of minimizing the sum of grid operating cost and corresponding regularization penalty term, and with load balance and line transmission power limit as constraints, the unit parameters of each unit are solved to obtain the intermediate unit parameters of each unit in the current iteration round.

[0133] Upon entering the current iteration, the regularized stability coefficients calculated in this round are retrieved to construct a globally coupled optimization mathematical model with a regularized penalty term. The model's objective function is min(total operating cost + regularized stability coefficient × difference between the current temporary output matrix and the initial unit parameters). Hard constraints are divided into two categories: global grid constraints and basic unit output constraints. Global grid constraints include hourly load balance equation constraints and power transmission inequality constraints for all transmission lines, strictly limiting the power transmission of a single line to not exceed its rated capacity. Basic unit output constraints limit the active power output of each unit to between its minimum stable output and its rated maximum output at any given time.

[0134] The model solution employs a dual decomposition column generation algorithm, using initial unit parameters as the initial iteration points for dual variables, replacing the traditional cold-start random initialization method and significantly shortening the convergence time of dual variables. During the solution process, the output values ​​of each unit are adjusted sequentially throughout the entire time period. After each variable adjustment, all global constraints are simultaneously verified. If constraints are violated, such as exceeding line power flow limits or mismatch between total output and load, the unit output allocation is corrected in reverse along the descent gradient of the objective function.

[0135] The dual variables and unit output are continuously iterated and adjusted until a set of continuous output allocation matrices that satisfies all global constraints is found within the current iteration. This matrix represents the intermediate unit parameters for the current iteration. After the solution is obtained, a basic verification is performed on the intermediate unit parameters, eliminating output values ​​that exceed the upper and lower limits of unit output. The rated boundary of the unit is used for correction to ensure that all intermediate unit parameters fall within the physically feasible range of the unit. After correction, the intermediate unit parameters are fully cached and synchronously distributed to the local control unit of each generator unit as a unified reference benchmark for the local state path search of the unit.

[0136] S202, the intermediate unit parameters are distributed to each unit so that each unit determines its corresponding unit combination set and the value score corresponding to each unit combination set based on the intermediate unit parameters, constrained by its own unit operating parameter range.

[0137] The range of unit operating parameters is the set of physical constraints of the unit, including the upper and lower limits of unit output, the unit ramp power limit, the minimum continuous operating time, and the minimum continuous downtime, which limits the boundaries of all feasible operating conditions of the unit.

[0138] Each generating unit's corresponding unit combination set includes the active power output and start / stop status at each time point within the future scheduling cycle. The value score is used to measure the degree of optimization potential of the unit combination set relative to the optimization objective. The unit combination set consists of multiple complete start / stop output trajectories obtained from the local search of a single generating unit. Each trajectory includes the active power output value and corresponding start / stop status for the entire scheduling cycle, all of which satisfy all physical constraints of the generating unit itself, and are candidate feasible solutions that can be added in the global optimization. The value score is used to quantify the optimization space of the current trajectory of a single generating unit. A value ≥ 0 indicates that the generating unit has reached a local optimum and there is no further optimization space; a value < 0 indicates that there is still optimization space in the trajectory, and the global iteration has not converged.

[0139] The dispatching system sends the complete intermediate unit parameter two-dimensional output matrix point-to-point to the local controller of each generator unit through the dedicated power network communication channel. Each unit only extracts the full-time output data of its own corresponding row as the reference output for local search.

[0140] Each unit controller reads its own complete set of operating parameters (output upper and lower limits, ramp limit, minimum continuous start-stop duration) to define its local status search domain, and only conducts path search within the search domain to exclude all naturally occurring non-compliant operating conditions.

[0141] The local search for a single unit is divided into four stages: The first stage is the construction of the power state diagram. Centered on the reference output at each moment, the hourly power search interval is defined by a preset floating range. Power state nodes are generated according to a fixed discrete step length. If the power difference between nodes in adjacent time slices is less than the ramp limit, a connected edge is established to generate a complete 24-hour state diagram. The second stage is the screening of non-dominated nodes. An output deviation threshold is set to define the expected power domain. Invalid nodes that deviate too much from the reference output are eliminated, while non-dominated power nodes with lower cumulative costs and greater adjustment margins are retained at the same moment. The third stage is the backtracking of candidate start-stop chains. Starting from the non-dominated node with the lowest cost at the time of scheduling termination, backtracking is performed to generate multiple complete candidate unit start-stop state chains. The fourth stage is the correction of compliant trajectories and the calculation of value scores. Each candidate state chain is divided into three categories: safe, critical, and exceeding limits. The safe chain with the highest matching degree with the direction of grid imbalance adjustment is selected as the candidate start-stop trajectory. Violation periods shorter than the minimum continuous start-stop duration within the trajectory are identified. By merging adjacent state segments and iteratively correcting, a fully compliant unit start-stop optimization trajectory is obtained. Multiple compliant trajectories of a single unit together constitute the unit combination set of that unit. After trajectory correction is completed, the computer group adjusts the margin upward (rated maximum output - current output) and downward (current output - minimum stable output) at each time step. The minimum of the two is taken as the potential score at a single time step. The potential scores over the entire time period are combined to obtain the value score (local optimization potential value) of the entire trajectory. The larger the value score, the greater the optimization space of the unit. A negative value indicates that there is still room for optimization and improvement of the current trajectory.

[0142] S203, if the iteration stopping condition is met based on the value score of each unit, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the target unit parameters.

[0143] The iteration stops when all units have non-negative value scores (local optimization potential values) and no unit has negative optimization potential.

[0144] After receiving the value scores returned by all generating units, the iteration stop condition is determined uniformly. The determination criterion is that the iteration stop condition is that the value scores (local optimization potential values) of all generating units are greater than or equal to 0, and there is no unit with a negative value score.

[0145] Establish batch traversal logic, read the value score in order from unit number 1 to N, set a flag bit, if any unit score < 0, mark it as not converged, directly jump out of the judgment process and enter the next round of iteration preparation; if all unit scores are ≥ 0, mark the global solution as converged.

[0146] After convergence is confirmed, it is determined that the start-stop optimization trajectories of each unit corresponding to the intermediate unit parameters in the current round have fully explored the local optimization potential. It is no longer possible to further reduce the total system operating cost by adjusting unit output and start-stop trajectories, and the global optimization objective has reached its theoretical optimum. The intermediate unit parameters obtained in the current iteration are then directly assigned as the target unit parameters. Without initiating a new round of globally coupled solution loops, the process jumps directly to the unit start-stop state discretization step to proceed with subsequent integer unit combination solutions.

[0147] This decision logic relies on the value score of local search quantization of a single unit as the convergence basis. Compared with the traditional method of simply relying on the difference of the objective function to determine convergence, the convergence judgment accuracy is higher, and there will be no problems such as premature termination of iteration and loss of optimal solution or excessive iteration and waste of computing power.

[0148] S204. If the value score of each unit determines that the iteration stop condition is not met, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the intermediate unit parameters of each unit in the future scheduling cycle obtained in the previous iteration round, and the unit parameters are solved again.

[0149] When batch traversing the value scores of the generating units, if at least one generating unit is detected to have a negative value score, it means that there is still room for optimization and adjustment in the current compliant start-stop trajectory of the generating unit. Adjusting the output and start-stop status of the generating unit can further reduce the total operating cost of the system. The global coupling solution has not reached the convergence state, and a new round of iterative loop must be started.

[0150] The baseline parameter update operation includes completely overwriting the intermediate unit parameters output in this iteration and replacing the historical intermediate unit parameters stored in the previous iteration. The updated intermediate unit parameters will be used as the comparison baseline vector for calculating the difference degree and regular stability coefficient in the next iteration.

[0151] After the baseline parameters are updated, all temporary cached data generated in this iteration is cleared, and only the updated intermediate unit parameters, initial unit parameters, basic grid topology and load data are retained. The system automatically jumps back to the starting point of the iteration loop and re-executes the entire process of difference calculation, regular stability coefficient solution, regular penalty term construction, and global coupling optimization solution.

[0152] In the new iteration, the global optimization model will generate new regularized penalty term weights based on the updated intermediate parameters, adjust the iteration constraint strength, guide the dual variables to adjust towards a better output allocation direction, issue new intermediate parameters to each unit to carry out local path search again, recalculate the value score of each unit, and perform convergence judgment again. This process is repeated until the value scores of all units meet the non-negative convergence condition.

[0153] This step achieves closed-loop iterative optimization by dynamically updating the benchmark, accurately capturing the optimization space of the power grid and generating units, ensuring that the final output target unit parameters are the globally optimal continuous power output allocation scheme, and preventing premature termination of the iteration that would result in suboptimal solution.

[0154] For example, taking 10 generating units and a 24-hour scheduling cycle, the second iteration of the global solution obtains the intermediate unit parameters, which are then distributed to the local controllers of the 10 generating units for local search. This illustrates the iterative process, including the following steps:

[0155] Unit local path search: Units 1 to 8 generate safe and compliant start-stop optimized trajectories. The trajectories generated by Units 9 and 10 have continuous start-stop switching for 2 hours, which violates the minimum continuous operation constraint of 4 hours and are judged as invalid trajectories and removed. Units 1-8 calculate their own trajectory value scores (local optimization potential values): Unit 1=280MW, Unit 2=320MW, Unit 3=250MW, Unit 4=0MW, Unit 5=180MW, Unit 6=-50MW, Unit 7=220MW, Unit 8=0MW.

[0156] Convergence criterion: After traversing all unit value scores, the score of unit 6 is -50MW < 0, which does not meet the iteration stopping condition.

[0157] Update the iteration baseline: Replace the historical intermediate parameters of the previous round with the intermediate unit parameters of the current round, and use them as the baseline for calculating the difference in the third round of iterations, and restart the global coupled solution process.

[0158] After the third iteration, Unit 6 resumed local path search, corrected the illegal start-up and shutdown periods, updated its value score to 80MW, kept the scores of the remaining units non-negative, and the value scores of all units were ≥0, thus meeting the iteration stop condition. The parameters of the intermediate units in this round were determined as the parameters of the target units.

[0159] In the above embodiments, a single iteration is broken down into a global coupled solution, a unit local path search, and a convergence determination closed-loop process. First, the intermediate unit parameters are obtained quickly by relying on the dual variables of the neural network hot start. These parameters are then distributed to each unit to carry out a refined state path search with the physical constraints of the complete unit as the boundary. This generates a set of unit combinations that includes all-time output and start / stop states. At the same time, the value score is quantitatively output to intuitively represent the optimization potential of a single unit. The non-negative value score of all units is used as the criterion for stopping the iteration. The determination logic accurately distinguishes between the globally optimal and non-converged conditions. If there is a negative value score, the iteration benchmark is automatically updated and a new round of solution is started until all units have no optimization space. This avoids premature termination of the iteration when there is no convergence, which would lead to suboptimal results. It also stops the iteration in time after global convergence, eliminating meaningless repeated calculations and wasting server computing power. Relying on the global and local collaborative iteration mechanism, the total number of iterations is further reduced while ensuring the optimality of the unit combination solution, thereby improving the overall solution speed of large-scale power grid unit combinations.

[0160] In some alternative implementations, see [link to relevant documentation]. Figure 3 , Figure 3 A flowchart for determining the target unit status of each unit is provided, which specifically includes the following steps:

[0161] S301: For each unit at each moment, if the target unit parameter is greater than the unit's minimum active power output, the initial unit state of the unit is determined to be the operating state; if the target unit parameter is less than or equal to the unit's minimum active power output, the initial unit state of the unit is determined to be the shutdown state.

[0162] Among them, the minimum active power output is the minimum output threshold for stable operation of the unit. If it is lower than this value, the unit cannot continuously generate electricity by connecting to the grid, which is the minimum output of the unit.

[0163] The initial unit state is based on a 0 or 1 start-stop vector obtained by simple discretization of continuous output values.

[0164] After iteratively converging and outputting the two-dimensional output matrix of the target unit parameters, the system traverses all units and all scheduling time slices within the matrix, executes discrete start-stop decision logic point by point, and generates an initial two-dimensional integer matrix of unit states.

[0165] The inherent minimum active power output parameters of each unit are read in advance and cached as a unified judgment threshold. Different units are allowed to have different minimum output values. The judgment logic is executed independently at each time step and is not affected by the output values ​​of adjacent time periods.

[0166] For determining the output value of a single generating unit at a single moment, the process includes retrieving the preset minimum active power output threshold for that unit and comparing it with the current output value of the target unit. When the output value is strictly greater than the minimum active power output, it indicates that the unit can stably connect to the grid and operate under load, and the initial unit state at that moment is assigned the operating state, marked with the number 1. When the output value is less than the minimum active power output, or exactly equal to the minimum active power output, the unit cannot operate stably and continuously, and is determined to be in a shutdown state, marked with the number 0. During the traversal, the duration of continuous operation and continuous shutdown periods for each unit is recorded synchronously. After completing the full matrix discrete determination, the initial unit state vector is output in batches.

[0167] This initial vector completes the discrete partitioning of the output level, but does not consider the power flow constraints of the power grid transmission lines and the power supply and demand balance constraints of the entire network. There will be problems such as the total power of the transmission section exceeding the rated capacity and the mismatch between the total unit output and the load demand in some periods. It cannot be directly used as the final dispatch start-up and shutdown scheme. It is necessary to carry out subsequent correction operations in combination with the total power of the section and the power imbalance difference.

[0168] This step realizes the basic transformation from continuous real-number optimization solutions to discrete states that can be used in engineering (0 or 1). The decision logic is simple and efficient, and batch traversal of millions of output data can be completed in just a few seconds, providing a basic discrete vector for subsequent power grid safety verification and state correction.

[0169] S302, determine the total power corresponding to the active power transmission power of each transmission line in each set of transmission lines in the power grid.

[0170] The transmission line set includes a collection of transmission lines used for the same power transmission task. The transmission line set is essentially a transmission section, a collection of multiple parallel or series transmission lines that undertake the power transmission task in the same area, with the total transmission power of the section being uniformly managed during dispatching.

[0171] The system retrieves the pre-defined transmission line set (transmission section) partitioning rules from the complete power grid topology data. All sections are statically partitioned in advance, and each section is bound to a set of transmission lines responsible for regional power transmission. The line number, impedance, rated transmission capacity, and the upper limit of the section's rated total power are pre-stored in the database. Based on the initial unit state vector and the output distribution of the target unit parameters, the system calls the DC power flow calculation model to calculate the active power transmission of each transmission line time-by-time. The DC power flow model has a fast calculation speed and is suitable for solving the batch power flow problems of large-scale power grids in multiple time periods.

[0172] For each transmission line set, all transmission lines within the set are traversed, and the active power transmission value calculated for each line at that moment is read. The power values ​​of all lines in the set are summed to obtain the total power of the transmission line set at that moment. After calculation, the total power of each section is compared with the rated total transmission power limit of that section. If the total power > the rated limit, the transmission line set is marked as an over-limit section, and the over-limit power difference is recorded simultaneously. If the total power ≤ the rated limit, the section is marked as operating normally. The number, corresponding total power, over-limit identifier, and over-limit value of all transmission line sets are uniformly stored in the power flow verification data table. This serves as the basis for grid safety constraints in subsequent unit initial state corrections, accurately locating transmission sections with power flow overload risks, and providing basic data on spatial location and power flow sensitivity for selecting adjustable units.

[0173] S303 determines the difference between the total active power output of each unit in operation and the predicted electrical load.

[0174] The total active power output of each unit in operation is determined based on the target unit parameters corresponding to each unit in operation. The total active power output is the sum of the output values ​​of the target unit parameters at the corresponding time for all units initially marked as in operation, representing the total power supply capacity of the power grid under the current discrete scheme.

[0175] The difference is the total active power output minus the predicted power load at the corresponding time, i.e., the power imbalance, which reflects the degree of matching between power grid supply and load demand.

[0176] The power imbalance difference is calculated independently for each scheduling time slice, and the calculation process at each time point is independent and does not interfere with each other. For a single time period, all units in the initial unit state vector are first traversed, and all unit numbers marked as operating state (number 1) at the current time are selected. Based on the selected unit numbers, the active power output value of the corresponding unit at that time is extracted from the target unit parameter output matrix. The output values ​​of all operating units are summed to obtain the total active power output of the units in the operating state at that time, which represents the total power supply output of the entire grid under the current discrete start-stop scheme.

[0177] Read the predicted power load value at that moment, subtract the predicted power load from the total active power output, and calculate the power imbalance difference. When the difference is positive, it means that the total output of the currently operating units exceeds the regional power load, the grid has a power surplus, and some units with output close to the minimum threshold need to be shut down to reduce the total output; when the difference is negative, it means that the total output of the operating units cannot meet the load demand, there is a power deficit, and some units that are close to the over-limit section need to be started to supplement the output; when the absolute value of the difference is less than the preset convergence threshold, it means that the power supply and demand are basically balanced, and there is no need to adjust the number of operating units.

[0178] The power imbalance difference, surplus or deficit identifiers for each time period are stored uniformly and merged with the total power data table of the transmission section to form a complete initial discrete scheme defect data table. This clarifies the two types of defects in the current initial unit state vector: power imbalance and power flow exceeding limits, supporting subsequent targeted start-up and shutdown corrections.

[0179] S304, based on the total power and the difference, corrects the initial unit state of each unit to obtain the target unit state.

[0180] The system synchronously reads the total power flow data table and the power imbalance difference data table of the transmission line aggregate, and determines whether the initial unit status needs to be corrected in two levels. The first level determines the power flow defect. If the total power of any transmission line aggregate exceeds the rated upper limit, it is determined that correction is necessary. The second level determines the power balance defect. If the absolute value of the power imbalance difference is greater than the preset convergence threshold at any time, it is determined that correction is necessary. If either of the two types of defects exists, the unit start-up and shutdown status correction process is initiated.

[0181] The correction process consists of six steps: candidate unit screening, unit determination to be adjusted, state reversal scheme generation, compliance verification, optimal scheme selection, and vector update.

[0182] The first step is to screen two types of candidate units: the operating side candidate units are those that are in operation and whose output and minimum active power difference is less than the preset output difference threshold. Shutting down these units has little impact on the total output. The shutdown side candidate units are those that are in shutdown and whose spatial distance from the over-limit transmission section node is less than the preset distance threshold. Starting these units can quickly adjust the power flow of the section.

[0183] The second step is to calculate the DC power flow sensitivity coefficient and power injection distribution factor of all candidate units, and select units that simultaneously meet the sensitivity threshold and distribution factor threshold as units to be adjusted. These units can simultaneously solve the problems of power flow exceeding limits and power imbalance.

[0184] The third step is to compare the start-stop flags of the unit to be adjusted in the feasible integer solution with the initial unit state, generate an initial state reversal scheme, and switch the start-stop state of the unit to compensate for grid defects.

[0185] The fourth step is to verify each set of reversal schemes: First, all units meet the minimum continuous operation and shutdown time constraints during continuous start-up and shutdown periods; second, the combined ramping limit of the remaining non-adjustable units can compensate for the system output jump caused by state switching. Illegal schemes are eliminated to obtain the final set of reversal schemes.

[0186] The fifth step involves multi-level sorting to select the optimal solution, prioritizing power balance and then power flow safety. The solution with the smallest power imbalance difference and the lowest cross-sectional limit is selected first.

[0187] The sixth step is to modify the start-stop identifier of the unit to be adjusted in the initial unit state vector based on the optimal flip-flop scheme, while keeping the states of the other units unchanged. After the correction is completed, the target unit state integer matrix is ​​obtained, which has no power flow over-limit, power balance, and compliant unit start-stop.

[0188] For example, taking 10 generating units and a 24-hour scheduling cycle as an example, the initial unit state vector is generated by discretely generating the target unit parameters as follows: Units 1, 3, 5-10 are running at 10:00 (1), and Units 2 and 4 are shut down (0); the minimum active power output of the units is 450MW. The process of determining the target unit state of each unit is introduced, including the following steps:

[0189] Discrete determination: Unit 2 output 400MW≤450MW represents shutdown; Unit 3 output 480MW>450MW represents operation, and initial unit state vectors are generated in batches.

[0190] Total power calculation of cross section: The power grid consists of 3 transmission lines (cross sections). The total power of cross section 2 is 1600MW, and the rated upper limit is 1500MW. The power flow exceeds the limit.

[0191] Power imbalance difference calculation: The total active power output of the operating units is 7500MW, the load demand is 7400MW, the difference is 100MW, the convergence threshold is 148MW, and the power balance meets the standard.

[0192] State Correction: If there is an over-limit section, the correction process is initiated. Unit 2 is selected as the candidate for shutdown and Unit 3 as the candidate for operation. The sensitivity and distribution factor are calculated to determine that Unit 2 is the unit to be adjusted. A reversal scheme for Unit 2 to switch from shutdown to operation at 10:00 is generated. The start-up and shutdown time of the unit and the system ramp-up compensation capability are verified to be compliant and determined to be the optimal scheme. The initial vector of Unit 2 state is modified to operation to obtain the target unit state matrix.

[0193] In the above embodiments, the basic discretization of the target unit parameters for continuous output is first completed based on the minimum active power output threshold of the unit, and the initial unit start-up and shutdown vector is quickly generated. Then, the power flow exceeding the limit section is accurately identified by calculating the total power of the entire transmission line set. At the same time, the imbalance difference between the total output of the operating units and the load is statistically analyzed to fully locate the two types of defects in the initial discretization scheme: grid safety and power supply and demand. The unit state correction is initiated based on the dual indicators of total power flow and imbalance difference. Adjustable candidate units are selected in a targeted manner. Based on power flow sensitivity and power distribution factor, the units to be adjusted that can simultaneously solve the two types of defects are locked. Multiple start-up and shutdown reversal schemes are generated and the physical constraints of the units and the system ramp-up adjustment capability are verified. The optimal correction scheme is selected through multi-level sorting to update the initial state vector. The final output target unit state not only meets the inherent constraints of the minimum continuous start-up and shutdown of the unit, but also eliminates the problems of power flow exceeding the limit of the transmission section and the power imbalance of the whole network. The unit start-up and shutdown state takes into account both the safety of the unit equipment and the power flow safety of the grid. The correction logic is progressively layered to avoid the violation of another type of grid constraint due to single-dimensional correction.

[0194] In some optional implementations, the initial unit state of each unit is corrected based on the total power and the difference to obtain the target unit state. During the process, if the difference is greater than or equal to a first preset difference, and / or there is at least one target line set, then the unit to be adjusted is selected from each unit; wherein, the target line set is the transmission line set whose total power is greater than the preset power; then the initial unit state of the unit to be adjusted is adjusted to obtain the target unit state of the unit to be adjusted.

[0195] The first preset difference is the power imbalance convergence threshold, which is generally set to 1%-3% of the predicted power load at the corresponding time. If the absolute value of the difference is greater than or equal to this value, it indicates that the power supply and demand are unbalanced and the start-up and shutdown of the units need to be adjusted.

[0196] The target line set is the set of transmission lines whose total power exceeds their own rated transmission power (preset power), i.e., power flow exceeding the limit transmission section.

[0197] The preset power is the maximum total active power that a single transmission line set is allowed to transmit, which is the upper limit of the rated capacity of the cross section.

[0198] For example, the power imbalance difference and the total power data of all transmission lines are read synchronously at each moment, and a dual condition judgment is performed. The condition is an OR logic, and the selection process of the unit to be adjusted is started when either condition is met.

[0199] The first criterion is to take the absolute value of the power imbalance difference and compare it with a first preset difference (2% load). If the absolute value is greater than or equal to the first preset difference, it is determined that the power supply and demand of the power grid is unbalanced, and the start-up and shutdown of the generating units need to be adjusted to change the total power output. The second criterion is to traverse all transmission line sets and compare the real-time total power of each set with the preset power of that set (the upper limit of the rated capacity of the cross section). If the total power of any set is greater than the preset power, the set is marked as the target line set, and it is determined that there is a power flow overload safety hazard in the power grid, and the start-up and shutdown of the generating units need to be adjusted to change the cross section power flow distribution. If either of the two conditions is met, the unit enters the candidate unit and unit to be adjusted screening process; if neither condition is met, it means that the initial unit state has no defects, and the initial unit state is directly assigned to the target unit state without adjustment.

[0200] The screening process consists of two steps. The first step is to screen two types of candidate units: the operating side and the shutdown side. The operating side candidate units are those that are in operation and whose output is close to the minimum active power output, and whose total output fluctuates little after shutdown. The shutdown side candidate units are those that are in shutdown and whose spatial distance from the target line set is relatively close, and whose cross-sectional power flow can be quickly changed after startup.

[0201] The second step involves cross-screening the units to be adjusted based on the power flow sensitivity and power injection distribution factor of the candidate units. Only units that have the ability to regulate both power flow over-limits and power imbalances are retained as the units to be adjusted, thereby reducing meaningless unit start-up and shutdown adjustments and lowering the start-up and shutdown costs of the corrected system.

[0202] This step accurately identifies the operating conditions that need correction through dual defect determination. Unit screening is only initiated when there is power imbalance or power flow exceeding the limit. Operating conditions without defects are directly skipped from the correction process, saving computing resources. At the same time, the units to be adjusted are screened in layers to narrow down the range of units to be adjusted and avoid drastic fluctuations in system output caused by large-scale unit start-up and shutdown changes.

[0203] Furthermore, after locking the unit number to be adjusted, the set of unit combinations (feasible integer solutions) generated in the global iteration phase is retrieved. The feasible integer solutions store the standard start-stop trajectories of all units that satisfy all constraints, which serve as the reference benchmark for state adjustment.

[0204] For each unit to be adjusted, the start / stop flags in the feasible integer solution are compared with the start / stop flags in the initial unit state vector at each time step to identify state conflict moments. Conflict moments are the periods when the start / stop states of the units in the two schemes are inconsistent. For all conflict moments, a state reversal operation is performed, including reversing to stop if the initial unit state is running and the feasible integer solution is stopped; and reversing to running if the initial unit state is stopped and the feasible integer solution is running. This generates multiple initial state reversal schemes.

[0205] After each rollover scheme is completed, a dual compliance check is performed. The first check is the physical constraints of the generating units. This involves iterating through the start-stop states of the units to be adjusted throughout the entire time period after the adjustment. The duration of all continuous operation and shutdown periods must be greater than or equal to the minimum continuous operation and minimum shutdown duration, respectively. Schemes with frequent start-stop violations are eliminated. The second check is the system's ramp-up compensation capability. This involves calculating the jump in total system output between adjacent time periods caused by the state rollover. The maximum ramp-up limit of all non-adjustable generating units must be greater than the jump variable to ensure that the unit output can quickly compensate for power fluctuations and maintain grid stability.

[0206] The scheme that passes the verification is the final state reversal scheme. The optimal reversal scheme is selected by using a multi-level sorting rule. The scheme with the smallest power imbalance difference is selected first. If the differences are the same, the scheme with the lowest cross-sectional power flow limit is selected.

[0207] Based on the optimal flipping scheme, modify the start / stop flags of the unit to be adjusted at conflict times, while keeping the initial vector unchanged for other non-conflict times and the states of the units not to be adjusted. After the modification, the start / stop flags of all times of the unit to be adjusted are the target unit states of that unit. After all units are integrated, a complete target unit state matrix is ​​formed.

[0208] This step relies on feasible integer solutions as the adjustment benchmark to ensure that the corrected unit start-up and shutdown status naturally conforms to the global optimal operating trajectory. Dual compliance verification avoids the risks of illegal unit start-up and shutdown and large fluctuations in system output. Multi-level sorting and screening selects the optimal adjustment scheme, taking into account both the needs of grid power balance and power flow security.

[0209] For example, taking 10 generating units, a 24-hour scheduling cycle, a first preset difference of 148MW, and a preset power of 1500MW for section 2 as an example, the process of determining the status of the target generating units is introduced, including the following steps:

[0210] Defect judgment: The power imbalance difference is 100MW < 148MW (the difference condition is not met); the total power of section 2 is 1600MW > 1500MW, and there is a target line set (the second condition is met), so the screening of units to be adjusted is initiated.

[0211] Screening of units to be adjusted: 3 candidate units on the operating side and 2 candidate units on the shutdown side; cross-judgment based on power flow sensitivity and power distribution factor, unit 2 meets both thresholds and is determined as the unit to be adjusted.

[0212] State Adjustment: Comparing feasible integer solutions, the feasible state of Unit 2 at 10:00 is running, while the initial state is stopped, indicating a state conflict. An adjustment scheme for the reverse operation of Unit 2 is generated, and the start-up and shutdown durations and system ramp-up capabilities are verified to be compliant, thus determining it as the optimal scheme. The state of Unit 2 at 10:00 is adjusted to running, obtaining the target unit state of Unit 2. The states of the other units remain unchanged, and the complete target unit state matrix is ​​obtained by integration.

[0213] In the above embodiments, dual judgment conditions of power imbalance difference and power flow overload section are set. The OR logic is used to identify the operating conditions with scheduling defects. When there is power supply and demand imbalance or power flow overload of the transmission section, the unit adjustment process is initiated. The initial unit state is directly reused for the defect-free operating conditions to reduce unnecessary state correction calculations. When screening the units to be adjusted, only the units that can improve power imbalance and power flow overload are retained to narrow the adjustment range and avoid large-scale unit start-up and shutdown changes from increasing the system start-up and shutdown costs. The adjustment process uses the globally optimal feasible integer solution as the benchmark to flip the unit start-up and shutdown state, and simultaneously verifies the minimum continuous start-up and shutdown physical constraints of the unit and the joint ramp-up compensation capability of the system. Illegal adjustment schemes are excluded. The optimal adjustment strategy that takes into account power balance and power flow safety is selected through multi-level sorting. The final target unit state of the unit to be adjusted not only eliminates the defects of grid operation, but also fully complies with the inherent operating limitations of the generator unit.

[0214] In some alternative implementations, see [link to relevant documentation]. Figure 4 , Figure 4 A flowchart is provided to select the unit to be adjusted from various units, specifically including the following steps:

[0215] S401, for each unit in operation, if the target unit parameter of the unit is greater than the minimum active power output of the unit, and the difference between the target unit parameter and the minimum active power output of the unit is less than the second preset difference, then the unit is designated as a candidate unit.

[0216] The second preset difference is the threshold for judging whether the output of the operating unit is close to the minimum active power output. It is generally set to 5%-10% of the minimum active power output. If the difference is less than this value, it means that the unit output is extremely low and the shutdown has little impact on the total output.

[0217] Traverse all units marked as operating state (1) in the initial unit state vector, and conduct the operation-side candidate unit judgment for each unit. Read the minimum active power output and the preset second difference (minimum output 8%) of each unit in advance. Extract the target unit parameter output value of the unit at the current time. The first layer judges whether the output is strictly greater than the minimum active power output. If the condition is met, the basis for shutdown adjustment is available. The second layer calculates the difference between the output and the minimum active power output. Compare the difference with the second preset difference. If the difference is less than the second preset difference, it means that the current output of the unit is close to the lower limit of stable operation. After shutting down the unit, the total output of the system will decrease by a small margin and will not cause a large-scale power deficit. The adjustment risk is low, and it is marked as an operation-side candidate unit. If the difference is greater than or equal to the second preset difference, the unit output is too high. Shutting down the unit will greatly reduce the total output of the entire network and aggravate the power deficit. It is not included in the candidate units.

[0218] After traversing all operating units, a list of candidate operating units is compiled. All units in this list are low-output operating units. The total system output can be slightly reduced by shutting them down to correct the power surplus condition. At the same time, it can slightly alleviate the power flow pressure on the near-area transmission section, providing a basic candidate pool for subsequent dual screening of power flow and power balance, reducing the number of units to be calculated for subsequent sensitivity and distribution factor calculations, and reducing the amount of calculation.

[0219] S402, for each unit that is in a shutdown state, if the distance between the unit and any node in any target line set is less than a preset distance, then the unit is regarded as a candidate unit.

[0220] The preset distance is the spatial distance threshold between the shut-down unit and the node of the over-limit section, which is usually set to 50-100 kilometers. A distance less than the threshold means that the start-up and shutdown of the unit can quickly change the flow of the section.

[0221] Traverse all outage states (0) of the initial unit state vector, retrieve the planar coordinates of the unit installation nodes and the coordinates of the transmission nodes contained in the entire target line set stored in the grid topology, with a preset distance threshold of 60 kilometers. Using the two-point distance formula in the Cartesian coordinate system, calculate the straight-line distance between the unit itself and all nodes in each target line set for each outage unit. If the distance between the unit and any node of the overload section is less than the preset distance, it means that the unit is geographically close to the power flow overload section. Start the unit to increase the near-area power injection, which can quickly divert the transmission power of the target line set and alleviate the power flow overload. The outage unit is included in the outage side candidate unit. If the distance between the unit and all nodes of the overload section is greater than or equal to the preset distance, the unit is far away from the overload section. Starting and stopping the unit has almost no effect on the power flow of the section and is not included in the candidate list.

[0222] After all the out-of-service units have been traversed, the candidate units on the operating side and the out-of-service side are merged to obtain a complete list of candidate units. The units in the list are divided into two categories: units that can be shut down at low output and units that can be started at near-limit-out sections. This covers two types of regulation needs: adjusting the total output of the system and alleviating the power flow at sections. It accurately identifies units with the potential to correct grid defects and filters out units with no regulation value.

[0223] S403 determines the sensitivity coefficient of DC power flow for each candidate unit and each target line set.

[0224] The sensitivity coefficient is used to characterize the degree of influence of the active power output adjustment of the candidate unit on the transmission power of the target line set. It is the change in the total power flow of the target line set when the unit node power injection changes by 1MW. The larger the value, the stronger the power flow capability of the unit regulating section.

[0225] For each generator unit and each target line set in the candidate generator unit list, the DC power flow sensitivity coefficient is calculated in batches. The DC power flow calculation model keeps the output of other generator units in the grid, the line impedance, and the load demand unchanged. The active power injection power of the current candidate generator unit node is increased by 1MW, and the power flow of all transmission lines is recalculated. The new total transmission power of all lines in the target line set is summarized, and the difference in total power of the cross section before and after the adjustment is calculated. This difference value is the DC power flow sensitivity coefficient of the target line set corresponding to that generator unit.

[0226] A positive coefficient indicates that increasing the unit's output will exacerbate the power flow exceeding the limit at the cross-section, while a negative coefficient indicates that increasing the unit's output can reduce the total power at the cross-section and alleviate overload. The larger the absolute value of the coefficient, the more significant the power flow regulation effect of the unit at that cross-section. The sensitivity coefficient values ​​of all target line sets corresponding to each candidate unit are stored, and a unit-cross-section sensitivity mapping table is established to select the first target unit with the ability to strengthen regulation of power flow exceeding the limit.

[0227] This step quantifies the unit's ability to regulate power flow at overload sections, abandoning the rough selection based solely on spatial distance. Instead, it uses power flow physics calculations to precisely quantify the regulation effect, avoiding the selection of ineffective units that are geographically close but have weak power flow regulation effects.

[0228] S404, select at least one candidate unit with a sensitivity coefficient greater than a preset coefficient as the first target unit.

[0229] The preset coefficient is the sensitivity coefficient screening critical value, which is generally set to 0.1-0.3. A coefficient greater than this value indicates that the unit has a significant effect on regulating the cross-section beyond the limit.

[0230] Read the preset coefficient threshold of 0.2, traverse the candidate unit and section sensitivity mapping table. If the absolute value of the sensitivity coefficient of a single candidate unit corresponding to any target line set is greater than the preset coefficient, it means that the start-up, shutdown and output adjustment of the unit can significantly change the power flow value of the overload section and effectively alleviate the safety hazards of power flow exceeding the limit. The unit is then included in the first target unit set. If the absolute value of the sensitivity coefficient of a candidate unit corresponding to all target line sets is less than or equal to the preset coefficient, the unit has a weak effect on improving power flow exceeding the limit and is not included in the first target unit set.

[0231] A list of all units that meet the criteria is compiled to form the first target unit list. All units on the list have the ability to strongly regulate the power flow of the transmission section and are specifically used to solve the problem of power flow overload of the target line. The power flow regulation capability is screened in a hierarchical manner, and inefficient power flow regulation units are removed from the candidate units to narrow down the scope of units to be adjusted.

[0232] S405, determine the power injection distribution factor of each candidate unit node to the power imbalance node.

[0233] Among them, the power imbalance node is the node where the total active power output power does not match the predicted power load; the power injection distribution factor characterizes the ability of the candidate unit to adjust the power imbalance by adjusting the active power output power, which is the proportion of the power imbalance node supply and demand difference adjusted by the unit node by the unit node per unit power injection. The larger the value, the stronger the ability to adjust the power imbalance.

[0234] The power imbalance node is located, i.e., the load center node with the largest difference between the total power output and load demand in the entire network. Based on the grid topology impedance matrix and node admittance matrix, the improvement ratio of the power imbalance difference when each candidate unit's node delivers a unit of active power to the power imbalance node is calculated. This ratio is the power injection distribution factor. The larger the distribution factor value, the faster the power imbalance difference can be reduced by increasing or decreasing the unit's output, and the higher the efficiency of correcting power surplus or deficit. A value close to 0 indicates that the unit's output adjustment has almost no improvement on the supply and demand imbalance. A preset factor threshold of 0.1 is used. After calculation, the distribution factor values ​​of all candidate units are stored, and a power regulation capability mapping table is established to screen the second target unit that can efficiently correct the power imbalance. This step quantifies the unit's regulation efficiency for the overall network power supply and demand balance, distinguishes between units with high and low regulation efficiency, and provides a screening basis for simultaneously addressing both power flow and power defects.

[0235] S406, select candidate units with power injection distribution factors greater than preset factors as the second target units.

[0236] The preset factor is the critical value for screening distribution factors, which is generally set to 0.05-0.15. A factor greater than this value indicates that the unit can efficiently correct power imbalance.

[0237] The system reads a preset factor threshold of 0.1 and iterates through the power injection distribution factor values ​​of all candidate generating units. If the distribution factor of a single unit is >0.1, it means that the unit's start-up, shutdown, and output adjustment can efficiently reduce the power imbalance difference and quickly correct the power surplus or deficit in the grid, and it is included in the second target generating unit set. If the distribution factor is ≤0.1, the unit's power regulation efficiency is too low, and the adjustment cannot effectively improve the supply-demand imbalance, so it is not included in the second target generating unit list. All qualified generating units are compiled to form the second target generating unit list. The units in the list all have the ability to efficiently regulate the power balance of the entire grid and are specifically used to solve the problem of excessive power imbalance difference, thus completing the hierarchical screening of power balance regulation capabilities.

[0238] S407 designates units existing in the first and second target units as units to be adjusted.

[0239] Perform a set intersection operation on the first target unit list and the second target unit list to filter out units that appear in both lists. Such units meet two conditions: first, they have the ability to strengthen the power flow of the target line and eliminate the power flow exceeding the limit; second, they have the ability to efficiently regulate the supply and demand difference of the power imbalance node and reduce the power imbalance difference.

[0240] Only units within the intersection area are identified as units to be adjusted. Units outside the intersection area, which can only address a single defect or have a minimal improvement effect on two types of defects, are not included in the adjustment scope. The number of units to be adjusted has been significantly reduced, retaining only the optimal units that can simultaneously address both power flow overload and power imbalance defects in one go. This reduces the number of subsequent unit start-up and shutdown adjustments, lowers system start-up and shutdown costs and output fluctuations, and balances grid security with dispatch economy.

[0241] For example, taking a second preset difference of 36MW, a preset distance of 60km, a preset coefficient of 0.2, a preset factor of 0.1, and a target line set of 2 sections as an example, the process of selecting the unit to be adjusted from each unit is introduced, including the following steps:

[0242] Candidate screening on the operation side: Unit 3 with an output of 480MW and a minimum output of 450MW, with a difference of 30MW < 36MW, is marked as a candidate unit on the operation side.

[0243] Candidate units for shutdown: Unit 2, whose distance from node 2 at section 2 is 14.14 km (<60 km), is marked as a candidate unit for shutdown; Unit 4, whose distance exceeds 60 km, is excluded. Complete candidate units: Unit 2 and Unit 3.

[0244] Sensitivity coefficient calculation: Unit 2's sensitivity to section 2 is 0.25 > 0.2; Unit 3's sensitivity is 0.18 < 0.2; The first target unit is only Unit 2.

[0245] Distribution factor calculation: Distribution factor of unit 2 is 0.12 > 0.1, and distribution factor of unit 3 is 0.11 > 0.1; Second target units: unit 2 and unit 3.

[0246] Intersection operation: The first target {unit 2} and the second target {unit 2, unit 3} intersect at unit 2, thus determining unit 2 as the unit to be adjusted.

[0247] In the above embodiments, the units to be adjusted are screened in a hierarchical and step-by-step manner. First, candidate units with adjustment potential are obtained through coarse screening using output difference and spatial distance, filtering out a large number of units without adjustment value, thus reducing the amount of subsequent power flow and power distribution factor calculations. Then, the power flow regulation capability of the units on the over-limit section is quantified by the DC power flow sensitivity coefficient, and the first target unit that can alleviate power flow overload is selected. The power balance regulation efficiency of the units is quantified by the power injection distribution factor, and the second target unit that can correct the supply and demand imbalance is selected. Finally, the intersection of the two types of target units is taken as the units to be adjusted, and only the units that can simultaneously eliminate the power flow over-limit of the transmission section and reduce the power imbalance difference are retained, accurately locking the optimal adjustment target. This avoids the problem that the adjustment can only solve a single defect and the other type of grid constraint still violates the rules after the adjustment. At the same time, it reduces the number of units that need to be started and stopped for adjustment, reduces the equipment loss and system start-up and shutdown costs caused by frequent switching of units, and the correction scheme takes into account the grid power flow security, power supply and demand balance and dispatch economic costs.

[0248] In some optional implementations, the initial unit state of the unit to be adjusted is adjusted to obtain the target unit state of the unit to be adjusted. For each unit to be adjusted, the start-stop state at each time in the unit combination set corresponding to the unit to be adjusted is compared with the initial unit state of the unit to be adjusted at each time. The initial unit state at the inconsistent time corresponding to the inconsistent comparison result is adjusted to the start-stop state at the inconsistent time in the unit combination set to obtain the target unit state of the unit to be adjusted.

[0249] Among them, the start-stop status of the unit combination set is the compliant and feasible integer solution obtained by the local search of the unit in the global iteration stage, and the standard start-stop trajectory of a single unit at all times fully satisfies all physical constraints of the unit and global grid constraints.

[0250] Consistency comparison is performed for the same unit to be adjusted and at the same scheduling time, comparing whether the start / stop identifier of the feasible integer solution is the same as the start / stop identifier of the initial unit state vector. Inconsistency occurs when there is a difference in the unit's start / stop identifier.

[0251] For example, the unit number to be adjusted is read one by one. For a single unit to be adjusted, two types of time-series start-stop data are retrieved. The first type is the set of unit combinations returned to the dispatch master station during the global iteration phase of the unit. The optimal compliant start-stop trajectory (feasible integer solution) within the set is selected, and the standard start-stop identifier for each time period of 24 hours is extracted. The second type is the start-stop identifier of the unit throughout the entire time period within the discretely generated initial unit state vector.

[0252] Using the scheduling time slice as the alignment benchmark, consistency comparisons are performed hourly, and a time-series comparison record table is established. The record table fields include time, feasible integer solution start / stop status, initial unit status, and conflict flag. At each time step, it is determined whether the two sets of flags are completely consistent. If the feasible integer solution flag is 1 and the initial vector flag is also 1, or both are 0, then the time is marked as conflict-free. If one set of flags is 1 and the other is 0, then the time is marked as inconsistent, and the conflict period number is recorded.

[0253] After traversing all 24 scheduling times of the unit to be adjusted, a list of all inconsistent times for the unit is compiled. During non-conflict periods, the original initial start-stop state remains unchanged. State reversal adjustment is only carried out for conflict periods in the list. Units not to be adjusted do not participate in the comparison at all. The start-stop state maintains the initial vector unchanged, avoiding meaningless global state traversal and reducing comparison calculation overhead.

[0254] This step uses the optimal feasible integer solution obtained from the local search of the unit as the standard comparison benchmark. This benchmark naturally satisfies all physical constraints of the unit's minimum continuous start-up and shutdown, ramping, and upper and lower limits of output. Using this as the basis for adjustment can ensure that the corrected start-up and shutdown state will not result in the unit's illegal operation conditions. Only a small number of conflicting time periods are locked for adjustment, which greatly reduces the workload of adjustment.

[0255] Then, the list of inconsistent times of the units to be adjusted is read, and the start / stop identifiers of the feasible integer solutions corresponding to the conflicting time periods are extracted one by one. The original start / stop identifiers of the initial unit state vector for that time period are overwritten and replaced, thus completing the state reversal adjustment. The adjustment rules strictly follow the standard state of the feasible integer solutions: if the feasible integer solution is running (1) at that time and the initial vector is stopped (0), then the initial vector is updated to running; if the feasible integer solution is stopped (0) at that time and the initial vector is running (1), then it is updated to stopped.

[0256] During all conflict-free periods, the initial unit state remains unchanged without any modifications. After all conflict periods have been adjusted, the complete 24-hour start-up and shutdown sequence of the unit to be adjusted is output. This sequence represents the target unit state of the unit to be adjusted.

[0257] After the adjustment is completed, the system automatically verifies the continuous start-stop duration of the unit throughout the entire period, confirming that all continuous operation and shutdown periods meet the minimum duration constraint and there are no violations of frequent start-stop. At the same time, it verifies the power flow of the corresponding transmission section and the power imbalance of the entire network after the adjustment, confirming that the two types of grid defects have been effectively improved.

[0258] After all units to be adjusted have completed their status adjustments, the statuses of each target unit will be spliced ​​and integrated to form a complete target unit status matrix, which will serve as the discrete start-up and shutdown benchmark for subsequent economic dispatch power allocation.

[0259] This step directly uses the globally optimal feasible integer solution as the adjustment standard, only modifying the start-up and shutdown status during conflict periods. The adjustment logic is simple and clear. The corrected unit start-up and shutdown status naturally satisfies both the physical constraints of the unit equipment and the global power flow and power balance constraints. No additional complex iterative optimization is required, resulting in high adjustment efficiency. The output target unit status can be directly used to generate the final unit combination scheduling scheme.

[0260] For example, taking Unit 2 as the unit to be adjusted, and the scheduling time of 10:00 as the only inconsistent time, the process of adjusting the initial unit state of the unit to be adjusted is described, including the following steps:

[0261] Consistency comparison: The unit 2 combination set (feasible integer solution) is in operation (1) at 10:00; the initial unit state vector is out of operation (0) at 10:00, and 10:00 is marked as the inconsistent time; the two sets of states are completely consistent and there is no conflict at the remaining 23 times.

[0262] State adjustment: Modify the initial shutdown state at 10:00 to the running state of a feasible integer solution, and keep the initial state unchanged at other times to obtain the complete 24-hour target unit state sequence of Unit 2. Integrate the unchanged states of the other units to obtain the complete target unit state matrix.

[0263] In the above embodiments, the set of compliant unit combinations generated by the local search of the unit is used as the benchmark for adjusting the start-stop state. This benchmark satisfies all physical constraints of the unit and global constraints of the power grid in advance. By comparing the consistency of the two schemes at each time step, a small number of time periods with state conflicts are accurately identified. The start-stop state of the unit is flipped only for the time of inconsistency. During the time periods without conflict, the original state remains unchanged. There is no need to make large-scale changes to the start-stop identifier of the unit at all times, which reduces the amount of adjustment calculation.

[0264] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0265] Based on the same inventive concept, this application also provides a unit combination determination apparatus for implementing the unit combination determination method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more unit combination determination apparatus embodiments provided below can be found in the limitations of the unit combination determination method described above, and will not be repeated here.

[0266] In one exemplary embodiment, such as Figure 5 As shown, a unit combination determination device is provided, comprising:

[0267] The acquisition module 10 is used to acquire the power grid topology data of the power grid, as well as the predicted power load in the future scheduling cycle;

[0268] The prediction module 20 is used to input the predicted power load and grid topology data into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle;

[0269] The solution module 30 is used in each iteration round to solve for the unit parameters of each unit, with the optimization objective of minimizing the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration round, and with load balance and line transmission power limit as constraints, to obtain the target unit parameters of each unit. Among them, the regularization penalty term corresponding to the current iteration round is determined based on the difference between the intermediate unit parameters of each unit in the future scheduling period obtained in the previous iteration round and the initial unit parameters; the intermediate unit parameters of each unit in the future scheduling period obtained in the previous iteration round corresponding to the first iteration round are preset values.

[0270] The first determining module 40 is used to determine the start-up and shutdown status of each unit at each time in the future scheduling cycle based on the target unit parameters of each unit, and to obtain the target unit status of each unit.

[0271] The second determining module 50 is used to obtain the target unit combination of each unit based on the target unit status and target unit parameters of each unit.

[0272] The aforementioned unit combination determination device generates initial unit parameters as a hot-start benchmark by inputting predicted power load and grid topology data into the prediction model. This eliminates the need for traditional hot-start pre-calculation simplification models, saving the computational power and time consumed by pre-solution. It also avoids hundreds of invalid iterations caused by the lack of a reference benchmark during cold starts. Furthermore, the iteration process dynamically generates regularized penalty terms based on the difference between the intermediate parameters of the previous round and the initial parameters. These penalty terms constrain the deviation of the iterative output from the benchmark, guiding global optimization to converge quickly. This reduces the number of iteration rounds and constraint verifications, thereby improving the solution efficiency of unit combination.

[0273] In some alternative implementations, the solver module 30 is specifically used for:

[0274] In the current iteration, the optimization objective is to minimize the sum of the grid operating cost and the corresponding regularization penalty term. Constrained by load balancing and the upper limit of line transmission power, the unit parameters of each generating unit are solved to obtain the intermediate unit parameters for each unit in the current iteration. These intermediate unit parameters are then distributed to each unit, allowing each unit to determine its own unit combination set and the corresponding value score based on its operating parameter range. The unit combination set for each unit includes future scheduling parameters. The active power output and start / stop status at each moment within the cycle; the value score is used to measure the degree of optimization of the unit combination set relative to the optimization objective; if the iteration stopping condition is met based on the value score of each unit, the intermediate unit parameters obtained by each unit in the current iteration round are used as the target unit parameters; if the iteration stopping condition is not met based on the value score of each unit, the intermediate unit parameters obtained by each unit in the current iteration round are used as the intermediate unit parameters obtained by each unit in the previous iteration round for the future scheduling cycle, and the unit parameters are solved again.

[0275] In some alternative implementations, the first determining module 40 is specifically used for:

[0276] For each generating unit at each moment, if the target generating unit parameter is greater than the unit's minimum active power output, the initial generating unit state is determined to be the operating state; if the target generating unit parameter is less than or equal to the unit's minimum active power output, the initial generating unit state is determined to be the shut-down state. The total power corresponding to the active power transmission power of each transmission line in each set of transmission lines in the power grid is determined; wherein, the set of transmission lines includes the set of transmission lines used for the same transmission task. The difference between the total active power output corresponding to each generating unit in the operating state and the predicted power load is determined; wherein, the total active power output corresponding to each generating unit in the operating state is determined based on the target generating unit parameter corresponding to each generating unit in the operating state. Based on the total power and the difference, the initial generating unit state of each generating unit is corrected to obtain the target generating unit state.

[0277] In some alternative implementations, the first determining module 40 is specifically used for:

[0278] If the difference is greater than or equal to the first preset difference, and / or there exists at least one target line set, then the unit to be adjusted is selected from each unit; wherein, the target line set is the transmission line set whose total power is greater than the preset power; the initial unit state of the unit to be adjusted is adjusted to obtain the target unit state of the unit to be adjusted.

[0279] In some alternative implementations, the first determining module 40 is specifically used for:

[0280] For each operating unit, if the target unit parameter is greater than the unit's minimum active power output, and the difference between the target unit parameter and the unit's minimum active power output is less than a second preset difference, then the unit is designated as a candidate unit. For each shut-down unit, if the distance between the unit and any node in any target line set is less than a preset distance, then the unit is designated as a candidate unit. The sensitivity coefficient of the DC power flow between each candidate unit and each target line set is determined; whereby the sensitivity coefficient characterizes the effect of the candidate unit's active power output adjustment on the target line set. The degree of influence of transmission power; candidate units with at least one sensitivity coefficient greater than a preset coefficient are designated as first target units; the power injection distribution factor of each candidate unit's node on the power imbalance node is determined; wherein, the power imbalance node is the node where the total active power output power does not match the predicted power load; the power injection distribution factor characterizes the ability of the candidate unit's active power output power adjustment to regulate the power imbalance; candidate units with a power injection distribution factor greater than a preset factor are designated as second target units; units existing in the first and second target units are designated as units to be adjusted.

[0281] In some alternative implementations, the first determining module 40 is specifically used for:

[0282] For each unit to be adjusted, the start-stop status of the unit at each time in the unit combination set corresponding to the unit to be adjusted is compared with the initial unit status of the unit at each time. The initial unit status at the inconsistent time corresponding to the inconsistent comparison result is adjusted to the start-stop status at the inconsistent time in the unit combination set, so as to obtain the target unit status of the unit to be adjusted.

[0283] Each module in the aforementioned unit combination determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0284] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a unit combination determination method.

[0285] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0286] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the unit combination determination method described in any of the above embodiments.

[0287] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the unit combination determination method described in any of the above embodiments.

[0288] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the unit combination determination method described in any of the above embodiments.

[0289] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0290] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0291] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0292] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining unit combination, characterized in that, The method includes: Obtain grid topology data of the power grid, as well as predicted electricity load for future scheduling cycles; The predicted electricity load and the power grid topology data are input into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle; In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period for each unit corresponding to the first iteration are preset values. Based on the target unit parameters of each unit, determine the start-up and shutdown status of each unit at each time in the future scheduling cycle, and obtain the target unit status of each unit. Based on the target unit status and the target unit parameters of each unit, the target unit combination of each unit is obtained.

2. The method according to claim 1, characterized in that, In each iteration, the optimization objective is to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration. With load balancing and the upper limit of line transmission power as constraints, the unit parameters of each generating unit are solved to obtain the target unit parameters for each unit, including: In the current iteration, with the optimization objective of minimizing the sum of grid operating costs and corresponding regularization penalty terms, and with load balancing and line transmission power limits as constraints, the unit parameters of each unit are solved to obtain the intermediate unit parameters of each unit in the current iteration. The intermediate unit parameters are distributed to each unit so that each unit, constrained by its own operating parameter range, determines its corresponding unit combination set and the value score corresponding to each unit's unit combination set based on the intermediate unit parameters. The unit combination set corresponding to each unit includes the active power output and start / stop status at each time point within a future scheduling cycle. The value score is used to measure the degree to which the unit combination set needs optimization for the optimization objective. If the iteration stopping condition is met based on the value score of each unit, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the target unit parameters. If the value score of each unit determines that the iteration stopping condition is not met, then the intermediate unit parameters obtained by each unit in the current iteration round are used as the intermediate unit parameters for the future scheduling cycle obtained by each unit in the previous iteration round, and the unit parameters are solved again.

3. The method according to claim 2, characterized in that, The process of determining the start / stop status of each unit at each time point within a future scheduling cycle based on the target unit parameters of each unit, and obtaining the target unit status of each unit, includes: For each unit at each moment, if the target unit parameter is greater than the unit's minimum active power output, the initial unit state of the unit is determined to be the operating state; if the target unit parameter is less than or equal to the unit's minimum active power output, the initial unit state of the unit is determined to be the shutdown state. Determine the total power corresponding to the active power transmission power of each transmission line in each set of transmission lines in the power grid; wherein, the set of transmission lines includes the set of transmission lines used for the same transmission task; Determine the difference between the total active power output of each unit in operation and the predicted electricity load; wherein, the total active power output of each unit in operation is determined based on the target unit parameters corresponding to each unit in operation. Based on the total power and the difference, the initial unit state of each unit is corrected to obtain the target unit state.

4. The method according to claim 3, characterized in that, The step of correcting the initial unit state of each unit based on the total power and the difference to obtain the target unit state includes: If the difference is greater than or equal to a first preset difference, and / or there exists at least one target line set, then the unit to be adjusted is selected from each unit; wherein, the target line set is the transmission line set whose total power is greater than a preset power. The initial unit state of the unit to be adjusted is adjusted to obtain the target unit state of the unit to be adjusted.

5. The method according to claim 4, characterized in that, The selection of the unit to be adjusted from among the various units includes: For each unit in operation, if the target unit parameter of the unit is greater than the minimum active power output of the unit, and the difference between the target unit parameter and the minimum active power output of the unit is less than a second preset difference, then the unit is designated as a candidate unit. For each unit that is in a shutdown state, if the distance between the unit and any node in any target line set is less than a preset distance, then the unit is regarded as a candidate unit. Determine the sensitivity coefficients for DC power flow of each candidate generator set and each target line set; whereby the sensitivity coefficients characterize the degree of influence of the active power output adjustment of the candidate generator set on the transmission power of the target line set. Candidate units with at least one sensitivity coefficient greater than a preset coefficient will be designated as the first target units. Determine the power injection distribution factor of each candidate unit's node to the power imbalance node; where the power imbalance node is the node where the total active power output power does not match the predicted electricity load; the power injection distribution factor characterizes the ability of the candidate unit's active power output power adjustment to regulate the power imbalance. Candidate units with power injection distribution factors greater than preset factors are designated as second target units; The units existing in the first target unit and the second target unit are designated as units to be adjusted.

6. The method according to claim 4, characterized in that, The step of adjusting the initial unit state of the unit to be adjusted to obtain the target unit state includes: For each unit to be adjusted, the start-stop status of the unit at each time in the unit combination set corresponding to the unit to be adjusted is compared with the initial unit status of the unit at each time. The initial unit state at the time of inconsistency corresponding to the inconsistent comparison results is adjusted to the start-stop state at the time of inconsistency in the unit combination set, thus obtaining the target unit state of the unit to be adjusted.

7. A unit combination determination device, characterized in that, The device includes: The acquisition module is used to acquire grid topology data of the power grid, as well as the predicted electricity load in future scheduling cycles; The prediction module is used to input the predicted electricity load and the power grid topology data into the prediction model to obtain the initial unit parameters of each unit in the future scheduling cycle; wherein, the initial unit parameters include the active power output of each unit at each time in the future scheduling cycle; The solution module, in each iteration, aims to minimize the sum of the grid operating cost and the regularization penalty term corresponding to the current iteration, while constraining load balance and the upper limit of line transmission power. It solves for the unit parameters of each generating unit to obtain the target unit parameters. The regularization penalty term corresponding to the current iteration is determined based on the difference between the intermediate unit parameters obtained in the previous iteration for the future scheduling period and the initial unit parameters. The intermediate unit parameters obtained in the previous iteration for the future scheduling period of each unit corresponding to the first iteration are preset values. The first determining module is used to determine the start-up and shutdown status of each unit at each time in the future scheduling cycle based on the target unit parameters of each unit, and to obtain the target unit status of each unit. The second determining module is used to obtain the target unit combination of each unit based on the target unit status and the target unit parameters of each unit.

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

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

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

Images